medical research patient data

Ophthalmology

• Transforming healthcare and research with AI-driven tools

Aug. 28, 2024

Artificial intelligence (AI) has proved to be a useful tool that promises numerous benefits across healthcare and other fields. AI continues to evolve within healthcare with the goal of supporting patients and staff while improving outcomes, potentially lowering healthcare costs and benefiting population health.

Its success, however, is inherently tied to both the quality and quantity of the data used. A massive amount of high-quality data is necessary to train AI algorithms to get accurate results that will work with a broad range of patients.

Through its patient care and research initiatives, Mayo Clinic has been busy building one of the largest repositories of clinical data in the world. With more than 11 million patients with electronic records, Mayo Clinic currently has more than 200 algorithms under development.

One of the AI-driven tools developed within Mayo Clinic Ophthalmology is the Ophthalmology Parametric Universal Search, also known as OPUS. "OPUS is a powerful AI-bioinformatic system that allows us to search for specific patient cohorts in our medical records and build databases for AI training. Further, it strengthens our ability to do retrospective research and identify patients who might qualify for investigative prospective clinical trials," says Raymond Iezzi Jr., M.D. , an ophthalmologist and researcher at Mayo Clinic in Rochester, Minnesota.

"It also allows us to employ AI algorithms, analyze datasets and provide annotations," continues Dr. Iezzi. "By curating annotated datasets, we can better find patterns of disease and assemble cohorts of patients for research."

OPUS draws from 25 different databases, all of which are supported by Mayo Clinic's Center for Digital Health. "Mayo's Center for Digital Health has been critical in supporting and maintaining the OPUS infrastructure and organizing the data," says Dr. Iezzi.

The center is focused on transforming how Mayo Clinic delivers patient-centered care in the digital era. "The Center for Digital Health has the vision of bringing Mayo Clinic to a global community so that we can deliver Mayo Clinic care anywhere in a manner that is streamlined," says Sophie J. Bakri, M.D. , chair of Ophthalmology at Mayo Clinic in Rochester, Minnesota.

Current projects at the Center for Digital Health include improved consumer experience for patients, expanding virtual care, and transforming healthcare delivery through data and analytics.

For Ophthalmology, using Mayo Clinic's resources has been essential to continued development and utilization of OPUS. "OPUS was designed to let us automatically annotate large image sets," says Dr. Iezzi. "For example, our group in Ophthalmology recently annotated 16 million retinal photos. OPUS is one of the most powerful, advanced AI informatic systems in ophthalmology today."

The options for ways to use this technology are seemingly endless. "We're also collaborating with other departments within Mayo Clinic to help determine how oculomics (the analysis of images of the eye) can help us better identify systemic diseases," says Dr. Iezzi.

And the innovative options certainly don't end within Mayo Clinic's walls. "We are at the cusp of a major transition in computer and health information technology," says Dr. Iezzi. "And it presents the opportunity to build a global network of collaborative organizations using research data and patient records."

"Technology and data-driven innovation are making it possible for us to solve some of the most complex medical problems in novel ways," says Dr. Iezzi. "It's expanding our capabilities and transforming the way we cultivate knowledge — with the goal of ultimately enhancing outcomes for our patients."

For more information

Refer a patient to Mayo Clinic.

Receive Mayo Clinic news in your inbox.

Related content.

• Medical Professionals

5X Challenge

Thanks to generous benefactors, your gift today can have 5X the impact to advance AI innovation at Mayo Clinic.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• J Med Internet Res
• v.21(5); 2019 May

Health Data Processes: A Framework for Analyzing and Discussing Efficient Use and Reuse of Health Data With a Focus on Patient-Reported Outcome Measures

Niels henrik ingvar hjollund.

1 Occupational Medicine, University Research Clinic, AmbuFlex/WestChronic, Aarhus University, Herning, Denmark

2 Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark

José Maria Valderas

3 University of Exeter Collaboration for Academic Primary Care, Health Services & Policy Research Group, National Institute for Health Research Collaboration for Leadership in Applied Health Research and Care (South West Peninsula), University of Exeter, Exeter, United Kingdom

4 Centre for Patient Reported Outcomes Research, Institute of Applied Health Research, University of Birmingham, Edgbaston, Birmingham, United Kingdom

5 National Institute for Health Research Birmingham Biomedical Research Centre and National Institute for Health Research Surgical Reconstruction and Microbiology Research Centre, University of Birmingham, Edgbaston, Birmingham, United Kingdom

Melanie Jane Calvert

Associated data.

Examples of basic health data activities.

The collection and use of patient health data are central to any kind of activity in the health care system. These data may be produced during routine clinical processes or obtained directly from the patient using patient-reported outcome (PRO) measures. Although efficiency and other reasons justify data availability for a range of potentially relevant uses, these data are nearly always collected for a single specific purpose. The health care literature reflects this narrow scope, and there is limited literature on the joint use of health data for daily clinical use, clinical research, surveillance, and administrative purposes. The aim of this paper is to provide a framework for discussing the efficient use of health data with a specific focus on the role of PRO measures. PRO data may be used at an individual patient level to inform patient care or shared decision making and to tailor care to individual needs or group-level needs as a complement to health record data, such as that on mortality and readmission, in order to inform service delivery and measure the real-world effectiveness of treatment. PRO measures may be used either for their own sake, to provide valuable information from the patient perspective, or as a proxy for clinical data that would otherwise not be feasible to collect. We introduce a framework to analyze any health care activity that involves health data. The framework consists of four data processes (patient identification, data collection, data aggregation and data use), further structured into two dichotomous dimensions in each data process (level: group vs patient; timeframe: ad hoc vs systematic). This framework is used to analyze various health activities with respect to joint use of data, considering the technical, legal, organizational, and logistical challenges that characterize each data process. Finally, we propose a model for joint use of health data with data collected during follow-up as a base. Demands for health data will continue to increase, which will further add to the need for the concerted use and reuse of PRO data for parallel purposes. Repeated and uncoordinated PRO data collection for the same patient for different purposes results in misuse of resources for the patient and the health care system as well as reduced response rates owing to questionnaire fatigue. PRO data can be routinely collected both at the hospital (from inpatients as well as outpatients) and outside of hospital settings; in primary or social care settings; or in the patient’s home, provided the health informatics infrastructure is in place. In the future, clinical settings are likely to be a prominent source of PRO data; however, we are also likely to see increased remote collection of PRO data by patients in their own home (telePRO). Data collection for research and quality surveillance will have to adapt to this circumstance and adopt complementary data capture methods that take advantage of the utility of PRO data collected during daily clinical practice. The European Union’s regulation with respect to the protection of personal data—General Data Protection Regulation—imposes severe restrictions on the use of health data for parallel purposes, and steps should be taken to alleviate the consequences while still protecting personal data against misuse.

Introduction

Health information is central to all types of activities in the health care system, all of which involve collecting, analyzing, or using health information [ 1 ]. Securing personal data against misuse is the background for several legal initiatives, for instance, the implementation of the European Union’s General Data Protection Regulation (GDPR) [ 2 ]. One key element of this regulation is the principle that personal data collected for one purpose may not be immediately transferred and used for other purposes. However, while misuse of personal data poses a severe ethical problem, so does waste and duplicate collection of the same data from the same patients due to legal, organizational, and technical dysfunction. In addition, from the patient’s perspective, duplicate collection of data may be unnecessarily burdensome and time consuming, and the possibilities and advantages of alternative uses of health data should therefore be considered. We have discussed the patient’s perspective of joint use in more detail elsewhere [ 3 ].

Health information may be generated as an integrated part of health care activities, such as biochemical variables or entries in hospitals’ electronic health record (EHR) system, or it can obtained directly from the patient. The latter is the case for patient-reported outcome (PRO) measures, which have been defined by the US Food and Drug Administration as measurements “of any aspect of a patient’s health that comes directly from the patient, without interpretation of the patient’s responses by a physician or anyone else” [ 4 ]. This definition emphasizes the standardization of PRO data as opposed to unstructured clinician-reported summaries of patient history contained in the notes in patients’ health records.

The evaluation of treatment outcomes for each individual patient is typically captured by a combination of biological data, physical examination, and communication with the patient. However, evaluations of treatment outcomes at a group level (defined geographically, administratively, epidemiologically, or at the facility level) often focus solely on mortality; readmission; and, if available, data such as medicine use and other use of health services. Although these outcomes are undeniably important, they may fail to fully capture treatment outcomes. PRO measures can be used to complement such data as a primary or an additional distal outcome, or even serve as a proxy for an unmeasured clinical variable when collection of the latter is not feasible [ 5 ].

Health informatics aims to respond to the increasing demands of systematic collection and processing of data to inform individual patient care, service improvements, and precision medicine. A lot of effort and resources are expended on collecting, processing, storing, and retrieving health information (both PRO measures and other clinical measurements) such as in hospitals’ EHR systems. In parallel, an increasing number of research projects and initiatives independently collect health information for their own specific objectives. Although health informatics, as a discipline, engages with stakeholders from a wide range of professional backgrounds, roles, and interests, it mostly does so with a focus on one specific single application (clinical practice, clinical research, administrative purposes, surveillance, or computer science), as evidenced in textbooks and the relevant literature [ 1 ]. As a consequence, there is limited literature on the joint use of health information for several purposes.

Technical, legal, organizational, and other types of obstacles to the availability of data for multiple purposes result in inefficient use of resources among patients and clinicians as well as in the health care system and society. Where there is no additional benefit from repeating a measurement, the same health information should be collected only once. A typical example would be laboratory tests, which may be performed by the family doctor before referral to the hospital, but which may be repeated, in many circumstances, unnecessarily, once the patient arrives at the hospital. Similarly, clinicians frequently struggle to retrieve measurements needed for maintaining quality registers even though they may already be recorded in the EHR system. Similarly, PRO measurements may be repeatedly and independently collected in parallel for different purposes, such as clinical management, quality surveillance, and research projects. This may not just mean a waste of resources and an unnecessary burden to patients, but may also have implications for data quality, as response fatigue may lead to reduced response rates.

To qualify the discussion of efficient use of health data, we need a common language usable for all stakeholders, which does not exist. The aim of this paper is to propose a framework for analysis of use and reuse of health information, with a specific focus on the role of PRO measures in order to initiate and facilitate a more precise discussion.

Definitions

There is no consensus on the method to define health information and health data. All definitions rely on the concepts of information (facts about a situation, person, event, etc [ 6 ]) and the organized property of data. Data have been defined accordingly in various ways such as (1) “any organized information collected by a researcher” [ 7 ], (2) “information or knowledge represented or coded in some form suitable for better usage or processing” [ 8 ], or (3) “information, especially facts and numbers, collected to be examined and considered and used to help decision making” [ 6 ]. The first definition focuses on the collection process and excludes purposes other than research, while the second one relies on data structure only. The third one identifies three processes relevant to the health data: collection, examination, and use; it furthermore acknowledges that the nature of data is preserved even if they are only stored and not used, at least not immediately.

In this paper, we use the third definition and differentiate three data processes: data collection , data aggregation , and data use . As patients are the unit of observation for health data, we need to additionally consider a patient-identification process to define whose data will be collected. We will focus on persons who may have, or are under surveillance for, a health condition and use the term “patient” even though some may not have a medical diagnosis. A generic model for health data covering any patient-related health data activity is shown in Figure 1 . Definitions are summarized in Textbox 1 .

The four data processes in the lifespan of patient-related health data. Patient identification process: Identification of patient(s) from whom data are to be collected. Data collection process: The actual collection of health data including logistic procedures. Data aggregation process: Management and organization of collected data for the data use process. Data use process: Use of the health data for the purpose of the specified activity. Each process may be repeated or may take place simultaneously with the previous process. Further information is provided in Textbox 1 .

Definitions of health data terms.

Health data: Health information about individual patients

Health data activity: An activity with a health-related aim that uses or produces health data

Health data processes: Any health data activity includes four processes:

• Patient identification process: Identification of patient(s) from whom data are to be collected
• Data collection process: The collection of health data
• Data aggregation process: Transfer or organization of health data in a way that enables data use
• Data use process: Use of health data for the purpose of a specified health data activity

Timeframe: The timeframe of a health data process:

• Systematic: A planned or repeated health data process
• Ad hoc: A nonplanned process

Level: The level of a health data process:

• Patient level: The individual patient level
• Group level: A level with patients grouped according to some defined criteria

The patient- identification process corresponds to the definition of the patient or the group of patients that will be the ultimate source of data. The subsequent data collection process contains measurement methods for generating data for that patient or population of patients as well as logistic issues. In the data aggregation process, data are transferred, organized, and transformed to enable their subsequent use. Aggregation may include data logistic procedures like transmission, data reformatting, and data management procedures such as combining and merging with other data. The aggregation may be explicit during data management (eg, a specific data manager making the dataset ready for the researcher’s use) or implicit (eg, such as a clinical summary based on patient data in an emergency room). In the data use process, the aim for the actual health data activity is fulfilled (eg, publish the results or a clinical decision of a treatment plan). Any pair of consecutive data processes may be repeated and make take place simultaneously.

Two dimensions can be recognized across all four data processes: level and timeframe . Level may be either the individual patient or a defined patient group level (eg, patients admitted to a hospital department or patients with a specific health condition). Timeframe considers the scope of the health data activity and may be either ad hoc or part of a systematic planned process. Examples are provided below.

Based on these dimensions, 2 × 2 tables with four cells may be constructed for each of the four processes. Four basic health data activities may be defined, where the same cell is used in all the four data processes in Figure 1 . Figure 2 shows such examples.

Examples of basic health data activities, where the same cell is used in all the four health data processes. I: The patient makes an appointment, and during the consultation, data are collected and aggregated to make a clinical decision and treatment plan (all four data processes ad hoc at the patient level). II: The target population is identified and data are collected, managed, analyzed, and published (all data processes ad hoc at a group level). III: An inpatient is discharged and referred for continuous planned outpatient follow-up and data are collected during follow-up, aggregated at each visit, and used at the visit (all data processes are systematic at the patient level) IV: Patient groups are identified repeatedly (eg, once a year) based on some criteria and data are collected, managed, and analyzed/reported (all data processes take place systematic at a group level).

In most heath data activities, different cells are applied in the four health data processes, and these patterns will be analyzed to highlight their properties and differences.

Basic Health Data Activities

Each data process of basic health data activities ( Figure 2 ) is described below and displayed in Table 1 . A description of the contents of each process is shown in the Multimedia Appendix 1 .

Examples of basic and complex health data activities divided by level and timeframe. In basic health data activities, all four processes are in the same level/timeframe cell.

a Sys: Systematic or repeated data process.

b The most frequently applied data processes.

The Single-Episode Clinical Contact Activity

The patient makes an appointment with the general practitioner. During the consultation, data are collected by medical history and physical examination. These data and a general view of the patient and his/her resources are aggregated by the general practitioner into a conclusion and used for a clinical decision. Another single-episode example is an emergency room visit. The timeframe for all the four data processes listed in Figure 1 is ad hoc, and all take place at the patient level . The content of the data collection process may include systematic methods such as standardized blood tests or use of a specific validated questionnaire, in which case, the timeframe is ad hoc .

The Planned Outpatient Follow-Up Activity

Many patients with a chronic disease have systematic follow-ups in an outpatient clinic. The patients are referred to outpatient follow-up in a systematic manner based on written formal or local informal guidelines. The data needed for the outpatient consultation (eg, medical history, laboratory tests, PRO measures, and physical examination) are collected, aggregated, and used at the patient level in relation to each visit. The timeframe for all the four data processes is systematic and takes place at the patient level .

The Cross-Sectional Clinical Research Activity

In a cross-sectional study, the group of patients is defined once, the data are collected and analyzed once, and the results based on condensed data are published once. Another example is a registry-based study. The timeframe for all four data processes is ad hoc and takes place at a group level .

The Quality Surveillance Program Activity

The quality surveillance program is an ongoing activity, where at a defined timepoint, (eg, once a year), data are sampled and subsequently analyzed with respect to differences between departments and hospitals. Reports are published and used for optimizing quality of care or to inform the patient’s choice of health care provider. The timeframe for all four data processes is systematic with predefined intervals and take place at a group level .

Complex Health Data Activities

Although the abovementioned activities apply data processes in the same cell in all four data processes, most health data activities combine different cells. Table 1 (lower part) shows examples of such activities. The list is not comprehensive but represents examples of the possible combinations of data process patterns.

The Longitudinal Clinical Research Activity

The patient-identification process—the recruitment (eg, for a clinical trial)—takes place once, or in the case of an open cohort study, systematically over a long inclusion period. Data are collected systematically over time according to a defined study protocol. Aggregation (data management and analysis) and use (publication) take place only once. All data processes are systematic and take place at a group level . As discussed later, these data can be used for a range of other purposes.

The Clinical Guideline Activity

Clinical guidelines are based on meta-analyses of clinical trials and longitudinal studies collected at a group level . Data are aggregated to inform the guidelines and are published and implemented once or at regular intervals. The use of clinical guidelines is, however, most often ad hoc at the individual patient level when clinical decisions about diagnostic procedures and treatment are made at the “bedside” together with that specific patient. In many countries, the traditional ad hoc use of guidelines is being replaced by quality programs or pay-per performance systems with the purpose of implementing the guidelines for all relevant patients. This will move the data use process from ad hoc to systematic at the patient level .

The Individual Prognosis Activity

Like treatment guidelines, prognostic indicators rely on information collected at the group level . Prognostic forecasts are used at the individual patient level and use the experiences of cohorts of patients to provide information on individual prognosis. Prognostic information may also be used as decision support together with the patient, for example, to choose between two treatments. Two approaches that differ with respect to the aggregation process—model based or data based—may be distinguished. In the model-based approach, the data are aggregated once at the group level and published as, for example, an equation based on regression coefficients, while in the data-based approach, data are aggregated from the cohort data each time the prognosis is asked for, and the prognosis for a subgroup with characteristics similar to the patient is selected and displayed [ 9 ]. Traditionally, prognosis has been expressed in terms of clinical outcomes (survival, readmission etc), but PRO measures may be used to include outcomes such as symptom burden and functioning.

The Screening Program Activity

In a population screening program, citizens or patients to be invited are identified systematically based on risk factors such as age; gender; and at times, disease-specific risk factors. The data collection takes place at the group level , but the aggregation and use processes occur at the individual patient level , since each screening-positive citizen is referred and further diagnosed and treated individually.

The Disease Surveillance Activity

Registers for monitoring diseases have been known since the middle of the 19th century, when the first-known registry was established with the purpose of monitoring leprosy at the population level [ 10 ]. Relevant patients are preferably identified based on diagnosis codes in existing registers, but a number of disease registers still rely on reports from the individual clinician, as do etiological registers like worker’s compensation registers. Secondary collection of data (eg, histologic type of cancer or treatment) is organized systematically at the patient level .

The Health Care Error Surveillance Activity

Health care error is, by nature, an ad hoc event at the patient level . In surveillance, patients are identified and data are collected ad hoc at the patient level and aggregated to statistics and reports at the group level (eg, hospital, department, or physician). In case of a serious error, the data may also be used at the individual level as a basis for audit, compensation, or even legal action.

Activities in Primary Health Care

Traditionally, all four data processes in primary health care have been ad hoc at the patient level , except for systematic, group-level programs like vaccination, pregnancy, and maternal care as well as some mandatory reporting of summary statistics to medical authorities. However, in some countries, primary care activities go from ad hoc to systematic ally framed processes at the group level (eg, chronic care programs), where the general practitioner is expected to identify patients with certain profiles, and primary health care quality surveillance programs based on group-level aggregation of clinical data are also being implemented.

Patient-Reported Outcome Measures in the Data Collection Process

PRO-based health data are not essentially different from other sources of information with respect to the data processes of identification, aggregation, and use, but the data collection process has a number of features that are specific for PRO measures. First, without PRO measures, health data on symptoms and functioning are difficult to collect systematically and will be limited to observations and unsystematic clinician-reported subjective summaries of patient history, which frequently underestimate patient problems [ 11 ]. Second, PROs are often the only way to collect data from a patient at home (telePRO). A number of telehealth projects have tried to collect data from home with various hi-tech methods with impact limited to few specific diseases, whereas telePRO has shown robustness and been used in a range of chronic diseases [ 12 ]. In the following section, PRO-specific aspects of activities listed in Table 1 will be highlighted.

Patient-Reported Outcome in Patient-Level Activities

Paper-based patient questionnaires have been used in the clinical setting for decades to support the communication between the patient and physician. The PRO data are aggregated and used during the consultation as a tool to screen for a priori defined, critically important symptoms (red flags) and to prioritize issues based on the patient’s preferences. This use of PRO measures has increased with the introduction of Web-based questionnaires, patient kiosks in the waiting area, etc. The effects on the consultation processes have been reviewed elsewhere [ 13 - 15 ]. During patient follow-up, PRO data are collected in connection with each scheduled visit and used to support a longitudinal overview of symptoms and functions over time and to provide real-time warnings of deterioration aimed at facilitating a prompt response from the care team ( Figure 3 ). If patients complete the PRO remotely online, usually at home ( telePRO ), this information may be used as the base for demand-driven outpatient follow-up without prebooked visits, where disease-relevant PRO questionnaires filled in at home at fixed intervals are aggregated by a disease-specific algorithm that semiautomatically decides whether there is a need or wish for an outpatient visit [ 12 ]. This may solve the paradox that outpatient clinics may be drowning in patients even though a substantial part of the visits turn out to be unnecessary from both the patient’s and clinician’s point of view [ 16 - 18 ]. In Denmark, this principle has been implemented in chronic and malignant diseases including asthma, chronic obstructive pulmonary diseases, epilepsy, sleep apnea, prostatic cancer, and chemotherapy for a number of malignant diseases [ 19 ]. A national implementation of the principle is underway in Denmark for selected diagnostic groups.

Longitudinal overview of patient-reported outcome and self-reported measurements in outpatient follow-up (translated from Danish) [ 13 ].

Patient-Reported Outcome in Group-Level Activities

The clinical research, quality surveillance, clinical guideline, and individual prognosis activities ( Table 1 ) rely on similar data and will be discussed together. PRO data collection has been applied for decades in clinical research based on the belief that outcomes cannot be evaluated on the basis of clinical measures only. Ideally, most group-based activities need longitudinal data with a long follow-up period, often beyond the time span of outpatient follow-up. Due to the increasing use of PRO measures for clinical purposes, isolated collection becomes problematic because the patient is often reluctant to answer more than one questionnaire, especially when the relevance is not clear and questions across measures overlap, leading to repeated questions with similar content [ 20 ].

The demand for data by the health care system will undoubtedly increase in the future for all described activities, with the cross-sectional study as a possible exception. Most of the listed data-demanding activities focus on longitudinal data, and the following discussion will focus on this and the role of PRO measures in a longitudinal follow-up.

Multiple Use of Data Collected as Part of Clinical Follow-Up

Of the four data processes, the data collection process is the main challenge with respect to costs as well as logistics. To reduce costs and workload among patients and clinicians, it is essential to focus on joint efforts of data collection with subsequent use in other health data activities. To some extent, this is already happening (eg, clinical research based on clinical quality databases).

The basic example of longitudinal activity is the patient follow-up, where information on the course of treatment, symptoms, and effect of the intervention is monitored, and, if necessary, treatment is adjusted. This activity is systematic at the patient level , and data are already stored for documentation purposes and may therefore potentially be reused in other activities. A schematic overview of principles in joint efforts where data collected from patient follow-up are used in other activities is shown in Table 2 . For the activities listed in Table 1 , data aggregation and data use are unchanged; only the processes in the alternative patient identification and data collection processes differ.

Examples of joint use of health data based on reuse of data routinely collected during patient follow-up with alternative patient identification, complementary data collection, alternative aggregations, and uses of data.

b All check marks indicate unchanged activity-specific processes (see Table 1 ).

c Basis: The routine collected follow-up data are the base for alternative uses.

d Reuse: Direct reuse of data collected in the cell above.

e Comp: Complementary data collection.

A model for joint use of health data based on data collected during patient follow-up with secondary identification of missing patients, observations, and variables for the alternative use is shown in Figure 4 . The methods used for identification of missing patients, observations, and variables for the alternative use ( Figure 4 ) depend on the timeframe of the alternative use. If the ad hoc method is used, data are exported to an external system where the completeness is analyzed with record linkage methods similar to those used in normal registry-based research, followed by additional ad hoc data collection. If the alternative use is to take place repeatedly in a systematic manner, this detection of missing data should preferably take place with online access to the environment in which the clinical data reside. In the Central Denmark Region, a central data warehouse has been established, which now contains clinical information on medication, diagnoses, and procedures, and more information is being collected [ 21 ]. These data are available for use in quality-improvement projects, but according to the GDPR, the use of data for research requires the patient to provide explicit permission, which reduces the possibility for joint use significantly.

Joint use of health data based on data collected during patient follow-up. The oblique arrow indicates identification of missing patients, observations, and variables for alternative use.

Group-level ad hoc procedures are applied to identify missing patients, observations, and variables, with subsequent complementary data collection. In quality surveillance programs, identification takes place as a systematic group-level process, but data collection should rely mainly on data collected as part of normal clinical activity. In clinical research, the identification of patients to be included will be ad hoc , based on the specific research protocol, while data collection should rely on data collected as part of normal clinical activity as the primary source of data, supplemented with additional ad hoc data collection, when supplementary outcome assessment is needed. Individual prognosis will most often rely solely on data collection in research activities, but as the only other activity, data-based individual prognosis may entirely rely on data collected as part of clinical follow-up, given that the data are available for instantaneous on-the-fly aggregation and use [ 9 ].

Patient-Reported Outcome Data Collection Supplemental to Data Collected During Follow-up

Different health data activities may have different data demands with respect to timing and PRO content, which makes supplemental PRO data collection necessary. PRO collection in clinical practice does not cover the whole population of patients, since some patients may not attend follow-up and some patients may, for one reason or another, not complete a questionnaire. For clinical use, response rates over 90% are obtainable in telePRO [ 12 , 22 ]. For the traditional use of clinical PRO measures, typically collected in the waiting room area, PRO data are obtained only from patients who turn up at follow-up visits, and the response rate is dependent on local commitment, influenced by local population characteristics and configuration of the system (user interface, accessibility and compatibility) and integrated within the existing EHR, clinical pathways, and workflow.

Clinical data collection stops when the patient follow-up ends. For reuse in other activities, it may be necessary to apply supplemental data collection. It is possible to incorporate supplementary research items into PRO questionnaires used for clinical purposes as well as to extend the follow-up period longer than clinically relevant for data use in the activities of clinical research, quality surveillance, and individual prognosis. With respect to content, PRO questionnaires used for a clinical purpose may not be appropriate for other activities, but with respect to domains to be covered, the common set between the activities is often substantial. For example, the European Organisation for Research and Treatment of Cancer scales created for use at group level are often usable for telePRO [ 23 ] when supplemented with a few items, most importantly, the patient’s preference for contact with the health care provider.

High response rates are crucial for PRO data collection in any health data activity, and PRO data collection in clinical practices often has higher response rates than PRO data collected for use at group level [ 22 ], where the response rate is dependent on local coordination and commitment [ 24 ]. Response rates are dependent on how relevant the data appear to the patient, and clinical use seems nearest to the patient. PRO data collected in clinical practice may therefore yield higher quality of data than traditional surveillance studies. A known problem when PRO data are collected at a group level is what to do with alarming answers, the so-called PRO “alerts” such as high depression scores or signs of suicidal ideation [ 25 ]. This is feasible to deal with in a clinical setting but is very difficult when collecting data only for group-level use. While supplemental data collection of clinical data may be troublesome and expensive due to several reasons such as extra follow-up visits, PRO data collection processes may be centralized and automated if the relevant infrastructure is available [ 22 ].

Challenges in Reuse of Patient-Reported Outcome Data Collected During Follow-up

In order to achieve the anticipated potential of joint use of PRO data, some critical challenges should be addressed. The psychometric requirements may vary depending on the specific use (eg, level of reliability and sensitivity to change), but other requirements for the data collection process, such as high response rate, low attrition rate, and high data completeness, remain essentially the same regardless of the activity and type of data. All activities must meet challenges in terms of data collection logistics and management, and the demands for data security are typically also identical. Supplemental data collection requires close cooperation between PRO activities with real-time access to data, which raises some issues. The challenges are divided into three types.

Legal Challenges

Legislation issues have a bearing on all four data processes; therefore, the legal framework has to be precisely specified before any data collection can begin. Activities with systematic data collection may typically benefit from permanent permission from national data protection agencies, while ad hoc projects must apply for permission for a specified period. The fundamental problem is that all approvals are only valid for the specific activity (eg, quality surveillance or clinical research), which means that data cannot be used for other activities. The implementation of the European Union’s regulation with respect to protection of personal data—GDPR [ 2 ]—will make it even more difficult to use data for other purposes without a specific consent from each patient. This will have a serious impact on joint use unless health data are given a differentiated treatment, such that the requirements for confidentiality can be maintained and individual approval can be collected in an efficient way (eg, through some form of umbrella approval process). For group-level use, analyses of personal data may be performed on a remote server where the researcher may upload a dataset and merge it with personal data using a unique personal identifier. The researcher has access to only aggregated data such as tables and outputs from statistical analyses [ 26 ]. Such a method of accessing personal data is available in Denmark and the Netherlands, but for now, few health data such as those on diagnoses and procedures are available for merging.

Technical Challenges

The principle of supplemental identification and data collection described above presupposes real-time access to relevant patient databases in the patient-identification process and in most cases, in the data collection process. Apart from that, there are substantial technical issues related to the aggregation process. Data may be collected and stored, but not available for the relevant alternative aggregation. A typical example is the quality surveillance activity, where the needed data may already exist in the patient’s EHR, but an automated process of extracting and transporting data is not possible due to inadequate and incompatible information technology systems or a lack of relevant expertise. PRO data may already be collected but stored in a different system or format. A possible solution to the latter is proposed by the international Health Level Seven standards for transfer of clinical and administrative data between software apps used by various health care providers [ 27 ]. A special Health Level Seven section for PRO measures has recently been adopted.

Challenges Related to Content and Timing of Data Collection

The need for valid, reliable, and responsive measurement scales is common for data for any health data activity. For PRO data collected for making individual clinical decisions, measurement error is of particular importance. Although scales that have acceptable psychometric properties at the patient level will normally also perform well at a group level , the opposite is not true and the desirable content and length of a PRO questionnaire are likely to differ between group-level and patient-level activities. In routine patient follow-up visits, short instruments are often preferred and procedures that the clinician finds irrelevant for the actual patient may not be collected as prescribed (eg, a comprehensive time-consuming test of performance in a patient who has clinically completely recovered or a depression score in a patient who is clinically obviously not depressed). A possible solution to these contradictive interests may be application of item banks and computer-adaptive testing, which can achieve high reliability with the lowest-possible administration burden [ 28 ]. Timing of data collection poses another challenge for joint use, and the optimal timing of data collection may differ between activities. Quality surveillance and clinical research may prefer that data collection follow a fixed scheduled in compliance with a protocol, while outpatient clinical practice is focused on the practical arrangement of follow-up, and visits often have to be postponed for various reasons. Although from a clinician’s perspective, it might be acceptable that patients who are doing well cancel their appointments, this may result in devastating selective attrition in group-level activities. From a resource point of view and the patient’s perspective, a patient who does not need or want clinical attention should not go to follow-up visits just to deliver data for other purposes. A rational approach for addressing these problems with missing data for the alternative activity could be a supplemental real-time identification of patients with missing data combined with collection of PRO data on proxy variables.

Conclusions

We have introduced a model for health data with four data processes, each dividable with respect to timeframe and group level, which distinguishes properties relevant to the discussion of joint use across different purposes and supports consideration of the associated organizational and technological challenges. Based on this, we propose a model for joint use of health data, with data collected during follow-up as the backbone. In the future, clinical settings will be a prominent source of PRO data and data collection for research and quality surveillance will have to adapt to this circumstance and design ways of complementary data collection as and when necessary. Demands for health data will continue to increase, which will further add to the need for the concerted use and reuse of PRO data for parallel purposes due to financial, logistical, and ethical reasons. A number of legal, technical, and organization challenges must be addressed.

The risk of patients’ information being accessed and used by people for whom it was not initially intended is real. For example, the use of health data by private insurance companies might restrict access to health coverage for vulnerable patients and those with a precondition. Additionally, access to private medical information by law enforcement agencies could be a risk for individuals and society. However, the current legal restriction on the joint use of health data imposed by the GDPR makes no distinction between these misuses and the uses described in this paper. Steps should be taken to alleviate the current legal restriction on the joint use of health data imposed by the GDPR while still protecting patient data against misuse.

Acknowledgments

We acknowledge Professor Henrik Toft Sørensen, Department of Clinical Epidemiology, Aarhus University Hospital, Denmark, for his valuable input and comments to an earlier version of the manuscript.

MC and DK are funded by the NIHR Birmingham Biomedical Research Centre and the NIHR Surgical Reconstruction and Microbiology Research Centre at the University Hospitals Birmingham NHS Foundation Trust and the University of Birmingham. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health. MC has received personal fees from Astellas, Takeda, and Glaukos outside the submitted work and grants from the NIHR and is a coinvestigator at Health Data Research UK Midlands. DK is supported by the NIHR postdoctoral fellowship scheme (PDF-2016-09-009). JV has received grants and fellowships from NIHR and Instituto de Salud Carlos III (Spain); grants from MRC, CRUK, Royal Devon, and Exeter Trust, Fondo de Investigaciones Sanitarias (Spain); and consultancy fees from the World Health Organization and Technische Krankenkasse (Germany) and currently is chairman (unpaid) of the International Consortium for Health Outcomes Measurement panel for the development of core sets of outcomes for Health Overall Adult Health. NH is the founder of AmbuFlex.

Abbreviations

Multimedia Appendix 1

Conflicts of Interest: None declared.

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here .

Loading metrics

Open Access

Peer-reviewed

Research Article

The state of artificial intelligence in medical research: A survey of corresponding authors from top medical journals

Contributed equally to this work with: Michele Salvagno, Alessandro De Cassai

Roles Conceptualization, Data curation, Methodology, Writing – original draft, Writing – review & editing

Affiliation Department of Intensive Care, Hôpital Universitaire de Bruxelles (HUB), Brussels, Belgium

Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review & editing

* E-mail: [email protected] , [email protected] (AC)

Affiliation Sant’Antonio Anesthesia and Intensive Care Unit, University Hospital of Padua, Padua, Italy

Roles Writing – original draft, Writing – review & editing

Roles Visualization, Writing – original draft, Writing – review & editing

Affiliations Department of Mathematical Modelling and Artificial Intelligence, National Aerospace University “Kharkiv Aviation Institute”, Kharkiv, Ukraine, Ubiquitous Health Technologies Lab, University of Waterloo, Waterloo, Canada

Affiliation Department of Clinical Sciences and Community Health, Università degli Studi di Milano, Milan, Italy

Affiliation Department of Anesthesiology and Critical Care Medicine, Johns Hopkins University School of Medicine, Baltimore, MD, United States of America

Roles Supervision, Writing – original draft, Writing – review & editing

• Michele Salvagno, 
• Alessandro De Cassai, 
• Stefano Zorzi, 
• Mario Zaccarelli, 
• Marco Pasetto, 
• Elda Diletta Sterchele, 
• Dmytro Chumachenko, 
• Alberto Giovanni Gerli, 
• Razvan Azamfirei, 
• Fabio Silvio Taccone

• Published: August 23, 2024
• https://doi.org/10.1371/journal.pone.0309208
• Peer Review
• Reader Comments

Natural Language Processing (NLP) is a subset of artificial intelligence that enables machines to understand and respond to human language through Large Language Models (LLMs)‥ These models have diverse applications in fields such as medical research, scientific writing, and publishing, but concerns such as hallucination, ethical issues, bias, and cybersecurity need to be addressed. To understand the scientific community’s understanding and perspective on the role of Artificial Intelligence (AI) in research and authorship, a survey was designed for corresponding authors in top medical journals. An online survey was conducted from July 13 th , 2023, to September 1 st , 2023, using the SurveyMonkey web instrument, and the population of interest were corresponding authors who published in 2022 in the 15 highest-impact medical journals, as ranked by the Journal Citation Report. The survey link has been sent to all the identified corresponding authors by mail. A total of 266 authors answered, and 236 entered the final analysis. Most of the researchers (40.6%) reported having moderate familiarity with artificial intelligence, while a minority (4.4%) had no associated knowledge. Furthermore, the vast majority (79.0%) believe that artificial intelligence will play a major role in the future of research. Of note, no correlation between academic metrics and artificial intelligence knowledge or confidence was found. The results indicate that although researchers have varying degrees of familiarity with artificial intelligence, its use in scientific research is still in its early phases. Despite lacking formal AI training, many scholars publishing in high-impact journals have started integrating such technologies into their projects, including rephrasing, translation, and proofreading tasks. Efforts should focus on providing training for their effective use, establishing guidelines by journal editors, and creating software applications that bundle multiple integrated tools into a single platform.

Citation: Salvagno M, Cassai AD, Zorzi S, Zaccarelli M, Pasetto M, Sterchele ED, et al. (2024) The state of artificial intelligence in medical research: A survey of corresponding authors from top medical journals. PLoS ONE 19(8): e0309208. https://doi.org/10.1371/journal.pone.0309208

Editor: Sanaa Kaddoura, Zayed University, UNITED ARAB EMIRATES

Received: November 22, 2023; Accepted: August 8, 2024; Published: August 23, 2024

Copyright: © 2024 Salvagno et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All relevant data are within the manuscript and its Supporting Information files.

Funding: The author(s) received no specific funding for this work.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Artificial intelligence (AI) and machine learning systems are advanced computer systems designed to emulate human cognitive functions and perform a wide range of tasks independently. The giant leaps these systems provide are the possibility to learn and solve problems through autonomous decision-making if an adequate initial database is provided [ 1 ]. Natural Language Processing (NLP) represents a field within AI focused on enabling machines to understand, interpret, and respond to human language meaningfully.

One intriguing advancement within the realm of AI is the development of Large Language Models (LLMs), which are a subset of NLP technologies. They are characterized by billions of parameters, which allows them to process and generate human-like text, understanding and producing language across a wide range of topics and styles.Generative chatbots, like ChatGPT(Generative Pre-trained Transformer), Microsoft Copilot, or Google Gemini,enhance these models and offer an easy-to-use interface. These LLMs excel in natural language processing and text generation, making them invaluable for diverse applications. Specifically, they have been used in medical research for estimating adverse effects and predicting mortality in clinical settings [ 2 – 4 ], as well as in scientific writing and publishing [ 5 ]. Finally, domain-specific or fine-tuned modelsare models that undergo additional training on a specialized dataset and are tailored to specific areas of expertise. This allows these models to develop a deeper understanding of terminology, concepts, and contexts, making them more adept at handling tasks ina specific field.

Potential applications of AI, and more precisely LLMs, in scientific production, are vast and multi-faceted. These applications range from automated abstract generation to enhancing the fluency of English prose for non-native speakers and even streamlining the creation of exhaustive literature reviews [ 6 , 7 ]. However, AI output is far from being perfect, as AI hallucination has been well described and documented in the current literature [ 8 , 9 ]. Additional concerns include ethical, copyright, transparency, and legal issues, the risk of bias, plagiarism, lack of originality, limited knowledge, incorrect citations, cybersecurity issues, and the risk of infodemics [ 9 ].

In light ofAI’s novel application in scientific production, it remains unclear to what extent the scientific community understands its inherent potentials, limitations, and potential applications. To address this, the authors designed a survey to examine the level of familiarity, understanding, and perspectives among contributing authors in premier medical journals regarding the role and impact of artificial intelligence in top scientific research and authorship. We hypothesize that, given the novelty of large language models (LLMs), researchers might not be familiar with their use and may not have implemented them in their daily practice.

Survey design

An online survey in this study was conducted using the SurveyMonkey web instrument ( https://www.surveymonkey.com , SurveyMonkey Inc., San Mateo, California, USA). The survey protocol (P2023/262) was approved by the Hospitalo-FacultaireErasme–ULB ethical commission(Comitéd’Ethiquehospitalo-facultaireErasme–ULB, chairman: Prof. J.M. Boeynaems) on July 11 th , 2023.

Two members of the survey team (M.S. and A.D.C.) performed a bibliographic search on April 19, 2023, on PubMed and Scopus, to retrieve any validated questionnaire on the topic using the following search string: [((Artificial Intelligence) OR (ChatGPT) OR (ChatBot)) AND ((scientific production) OR (scientific writing)) AND (survey)]. No existing surveys on the specific topic were found.

Therefore, the research team constructed the questionnaire under the BRUSO acronym to create a well-constructed survey [ 10 ]. The survey consisted of 20 single-choice, multiple-choice, and open-ended questions investigating individuals’ perceptions of using Artificial Intelligence (AI) in scientific production and content. The full list of questions is available for consultation in English ( S1 Appendix Content 1, Survey Questionnaire in English).

Population of interest

The population of interest in this survey consisted of corresponding authors who published in 2022 in the 15 highest-impact medical journals ( S2 Appendix Content 2), as ranked by the Journal Citation Report from Clarivate. In this survey, we used the Journal Impact Factor (JIF) as a benchmark to target leading publications in the research field. Originally developed by Eugene Garfield in the 1960s, the JIF is frequently employed as a proxy for a journal’s relative importance within its discipline. It is calculated by dividing the number of citations in a given year to articles published in the preceding two years by the total number of articles published in those two years. The focus on the corresponding authors aimed to access a segment of the research community that is potentially at the forefront of research publishing and scientific production. For this survey, only the email addresses of the corresponding authors listed in the manuscript were sought and collected. Whenmultiple emails were listed as corresponding, only the first email for each article was collected.When no email addresses were found, no further steps were taken to retrieve them.No differentiation was made regarding the type of published article, except for excluding memorial articles dedicated to deceased colleagues. All other articles were included. The authenticity of the email addresses or their correspondence with the author’s name was not verified. As a result, it was not possible to calculate the a priori sample size.

Survey distribution plan

To enhance the survey’s effectiveness, a pretest was performed in two phases. In the first phase, the survey team reviewed the entire survey, with particular attention to the flow and the order of the questions to avoid issues with “skip” or “branch” logic. The time required to complete the survey was estimated to be around four minutes. In the second phase,the survey was distributed for validation to a small subset of participants, which included researchers working at the Erasme Hospital, to identify any issues before distributing it to the general population of interest. Their answers were not included in the final data analysis.

UsingSurveyMonkey’s email distribution feature, the survey link was disseminated to all collected email addresses of the corresponding authors. To minimize the ratio of non-responders, reminder emails were sent one, two, and three weeks after the initial contact, with a final reminder sent one month later. Responses were collected from July 13 th , 2023, to September 1 st , 2023. SurveyMonkey’s web instrument automatically identifies respondents and non-respondents through personalized links, allowing for targeted reminders to only those who had not yet completed the survey. This system also automatically prevents duplicate responses.

Statistical analysis

Descriptive statistics was used to provide an overview of the dataset. Depending on the nature of the variables the results are reported either as percentages or as medians with interquartile range (IQR). Comparison among percentages were performed with the chi-square test with a p-values significance threshold at 0.05. All statistical analyses were performed using Jamovi (Jamovi, Sydney, NSW Australia, Version 2.3) and GraphPad Prism (GraphPad Software, Boston, Massachusetts USA,Version 10).

A total of 4,302 email addresses for inclusion in the survey were collected from the list of journals in the appendix. Survey data were collected from 13 th July to 1 st September 2023. Following the initial email outreach and four subsequent reminders, 222 emails bounced back, and 142 recipients actively opted out of participating.Of those who opened the survey link, 266 respondents answered the initial questions. However, some immediately declined to continue, resulting in 236(5.5% of the emails sent) participants who started the survey and were included in the final analysis upon response.

The geographical distribution and demographic data of 229 respondents are depicted in Table 1 ,.The United States and the United Kingdom were most prominently represented, accounting for 57 (24.9%) and 41 (17.9%) of respondents, respectively. In total, English-speaking nations (USA, UK, Canada, and Australia) accounted for 124 (54.1%) of respondents.

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

https://doi.org/10.1371/journal.pone.0309208.t001

The role of 229 responders is represented in Fig 1 . Physicians, research academics and research clinicians were equally represented, with 64 (27.9%), 65 (28.4%) and 67 (29.2%) responders, respectively. The other responders declared not to be classified as the aforementioned and explained themselves mainly as journalists, students, veterinarians, editors, and pharmacists.

Proportion of respondents in various professional roles as a percentage of the total respondent pool.

https://doi.org/10.1371/journal.pone.0309208.g001

Most of the respondents to this question reported moderate 93 (40.6%) or little 60 (26.2%) familiarity with AI tools. Only 13 (5.7%) indicated extensive familiarity.Following questions up to Q14 were answered by all participants except for the 10 individuals (4.4%) who indicated no prior knowledge of AI (resulting in their automatic exclusion from answering those specific questions). Notably, 9 (69.2%)out of 13 with extensive familiarity reported AI tool usage, compared to lower rates among 20 out of 93 (21.5%)with moderate and 5 out of 60 (8.3%)minimal familiarity (p < 0.001).

More than half of 229 respondents (130, 55%) published their first medical article over 15 years ago, while 31 (13.5%) did so within the last five years. The median Scopus H-index among respondents was 24 (IQR 13–42). No statistically significant correlations were identified between H-index, AI familiarity and AI usage (p > 0.05).

Only 2 participants (< 1%), reported receiving specific training in AI for scientific production. Despite this, 55 (24.02%) out of 229 responders usedAI tools in scientific content creation.Of these, the majority (67.3%) used ChatGPT. Interestingly, among participants from the US(n = 57), a notable difference exists between those who have used AI for scientific production(n = 8, 14%) and those who have not (n = 49, 86%).Those who published the first medical article more than 15 years ago, also declared to have ever used AI tools for scientific production in a lesser amount than the ones who published the first medical article less than 15 years ago(23/130 [17.7%] vs. 32/99 [32.3%], p = 0.01).

As shown in Fig 2 , besides ChatGPT, among the 55 responders who have already published using the aid of AI during the scientific production,Microsoft Bing and Google Bard were used by 8 (14.5%) and 2 (3.6%) of respondents, respectively. Other large language models comprised 5.0% of the usage. Various software tools, including image creation and meta-analysis assistant tools, were also reported to be used by 7 (12.7%) and 6 (10.9%), respectively. Other AI tools reported are mainly Grammarly, Image Analysis tools, and plagiarism-checking tools.

The Y-axis lists the AI tools reported by respondents, while the X-axis shows their stated usage as a percentage. The total percentage exceeds 100% as respondents could report using multiple tools. LLM: Large Language Models; AI: Artificial Intelligence.

https://doi.org/10.1371/journal.pone.0309208.g002

When the 55 respondents who already used AI tools were asked about the primary applications of AI, 55.6% reported using AI for rephrasing text, 33.3% for translation, and 37.78% for proofreading. The rate of AI usage for language translation was consistent across English and non-English-speaking countries (94.4% vs 92.4%,p = 0.547). Additional applications such as draft writing, idea generation, and information synthesis were each noted by 24.4% of respondents.

In the survey, 8 of the 51 who answered this question (15.7%) admitted to using a chatbot for scientific work without acknowledgment.By contrast, 27 (11.9%)out of 226 are certain they will employ some form of Artificial Intelligence in future scientific production. The complete set of responses is summarized in Table 2 .

https://doi.org/10.1371/journal.pone.0309208.t002

The primary challenges associated with utilizing AI in scientific research are outlined in Table 3 .

https://doi.org/10.1371/journal.pone.0309208.t003

The medical fields that respondents anticipate will gain the most from AI applications are Big Data Management and Automated Radiographic Report Generation. Additionalareas are detailed in Table 4 .

https://doi.org/10.1371/journal.pone.0309208.t004

When asked about their ability to distinguish between text written by a human and text generated by AI, 7 (3.1%) out of 226 respondents believed they could always tell the difference. Meanwhile, 120 (53.1%) felt they could only sometimes discern the difference. A total of 59 (26%)were uncertain, and a small fraction, 3 (1.3%), reported it is never possible to distinguish between the two.

Over 80% of respondents (n = 226) do not foresee AI supplanting the role of medical researchers in the future, with 81 (35.8%)strongly disagreeing and 106 (46.9%)disagreeing. A small fraction, 10 responders (4.4%), either somewhat or strongly agree that AI could take on the role of medical researchers. Meanwhile, 29 (12.8%)remain uncertain. By contrast, when it comes to the impact on clinical physicians,among the 226 responders to this last question, 177(78.3%) anticipate that AI will partially alter the nature of their work within the next two decades. A minority of 18 responders (8.0%) foresee no change at all, and a very small fraction, 2 (0.9%), predict a complete transformation in the role of clinical physicians. To conclude, 14 (6.0%)are still unsure about the future impact of AI on clinical practice.

The present study aimed to explore the perceptions and utilization of Artificial Intelligence (AI) tools in scientific production among corresponding authors who published in the 15 most-impacted factor medical journals in 2022.

Familiarity and training in AI

Intriguingly, this survey indicated that less than 1% of respondents had undergone formal training specifically designed for the application of AI in scientific research. This highlights a critical need for educational programs tailored to empower researchers with the necessary skills for effective AI utilization. The dearth of formal training may also contribute to the observed "limited" to "moderate" familiarity with AI concepts and tools among most survey participants, without a difference among ages and genders.Generally, AI tools are user-friendly and straightforward, requiring no specialized skills for basic usage. This could account for the lack of a significant difference between younger and older users.However, even though the basic use appears straightforward, a lack of comprehension may lead individuals to commit unnoticed errors with these tools, stemming from an unawareness of their own knowledge gaps [ 11 ].

Although beyond the primary focus of this study, we find it noteworthy to comment on the responses concerning the Scopus H-index. This score remains a subject of debate and is fraught with limitations, including self-citation biases, equal attribution regardless of author order and academic age,as well as gender-based disparities other than topic-specific biases. In our survey, the responders presented a median H-index of 24 (IQR 13–42), without statistically significant correlationsbetween H-index values and the variables of interest. Remarkably, two respondents indicated a lack of interest in monitoring their H-index. One respondent, a journal editor, expressed outright indifference with the remark "Who cares", probably echoing a sentiment that could be ascribed to Nobel Laureate Tu Youyou, whose current relatively low Scopus H-index of 16 belies her groundbreaking work on artemisinin, a treatment for malaria that has saved millions of lives.

Applications of AI in scientific production

The survey results underscore a paradoxical relationship between familiarity with AI concepts and its actual utilization in scientific production. While many respondents indicated a “limited” to “moderate” familiarity with AI, around 25% reported employing AI tools in their research endeavors. This suggests that while the theoretical understanding of AI might be limited among the surveyed population, its practical applications are cautiously being explored. It is plausible that the rapid advancements in AI, coupled with its increasing accessibility, have allowed researchers to experiment with these tools without necessarily delving deep into the underlying algorithms and principles.Notably, the preponderance of the surveyed gravitated toward ChatGPT, suggesting a proclivity for natural language processing applications. Indeed, ChatGPT could assist scientists in scientific production in several ways [ 12 ].

The principal tasks for which AI was employed encompassed rephrasing, translation, and proofreading functions. AI tools, especially natural language processing models like ChatGPT, can significantly improve the fluency and coherence of scientific texts, especially for non-native English speakers. This is crucial in the globalized world of scientific research, where effective communication can determine the reach and impact of a study. Interestingly, the rates of AI use for language translation were quite similar between English-speaking and non-English-speaking countries, at 94.4% and 92.4%, respectively. This is unexpected since English is often the preferred language for communication in scientific fields, diminishing the perceived need for translation tools. Several factors could explain this trend. First, these countries have a high proportion of expatriates, leading to many non-native English speakers in the workforce. One limitation of our study is that we did not inquire about the respondents’ countries of origin, so we cannot provide further insights. Another possible explanation could be the selectivity of our respondent pool, which may not be sufficiently representative to show a difference in this variable.Nevertheless, ifthe predominant use of AI for tasks such as rephrasing, translation, and proofreading underscores its potential to enhance the quality of research output, it is essential to strike a balance to ensure that the essence and originality of the research are maintained in the pursuit of linguistic perfection.

This pattern intimates that, in its current stage, AI is predominantly perceived as a facilitator for enhancing the textual quality of scholarly work, rather than as an instrument for novel research ideation or data analysis. In response to this evolving landscape, academic journals, for example, JAMA and Nature, have issued guidelines concerning the judicious use of large language models (LLMs) and generative chatbots [ 13 , 14 ]. Such guidelines often stipulate authors’ need to disclose any AI-generated content explicitly, including the specification of the AI model or tool deployed.

While the survey highlighted the use of LLMs predominantly in textual enhancements, the potential of other AI in data analysis still needs to be explored among the respondents. Indeed, LLM and NLP, in general, currently have a very weak theoretical basis for data prediction.Nevertheless, longitudinal electronic health record (EHR) data have been effectively tokenized and modeled using transformer approaches, to integrate different patient measurements, as reported in the field of Intensive Care Medicine [ 15 ], even if this field is still insufficiently explored. Advanced AI algorithms can process vast datasets, identify patterns, and even accurately predict future trends, often beyond human capabilities. For instance, in biomedical research, numerous machine learning applications tailored to specific tasks or domains can assist in analyzing complex genomic data, predicting disease outbreaks, or modeling the effects of potential drugs. As indicated by the survey, the limited utilization of AI in these areas may be due to the lack of specialized training or apprehensions about the reliability of AI-generated insights.

Future prospects

Most respondents were optimistic about the future role of AI in scientific production, with nearly 12% stating they would "surely" use AI in the future. This optimism towards integrating AI in scientific production can be attributed to the numerous advancements and breakthroughs in AI in recent years. As AI models become more sophisticated, their potential applications in research expand, ranging from data analysis and visualization to hypothesis generation and experimental design. The increasing availability of open-source AI tools and platforms makes it more accessible for researchers to incorporate AI into their work, even without extensive technical expertise.

However, most respondents (> 80%) did not believe that AI would replace medical researchers, suggesting a balanced view that AI will serve as a complementary tool rather than a replacement for human expertise. The sentiment that AI will augment rather than replace human expertise aligns with the broader perspective in the AI community, often termed “augmented intelligence” [ 16 ]. This perspective emphasizes the synergy between human intuition and AI’s computational capabilities. While AI can handle vast amounts of data and rapidly perform complex calculations, human researchers bring domain expertise, critical thinking, and ethical considerations [ 17 ]. This combination can lead to more robust and comprehensive research outcomes [ 16 , 18 ].

Moreover, the evolving landscape of AI in research also presents opportunities for interdisciplinary collaboration [ 19 ]. As AI becomes more integrated into scientific research, there will be a growing need for collaboration between AI specialists and domain experts. Such collaborations can ensure that AI tools are developed and applied in contextually relevant and scientifically rigorous ways. This interdisciplinary approach can lead to novel insights and innovative solutions to complex research challenges.

Ethical and technical concerns

This survey identified a wide range of concerns regarding the integration of Artificial Intelligence (AI) into the realm of scientific research. Among these, content inaccuracies emerged as the most salient, flagged by over 80% of respondents. The risks associated with AI-generated content include creating ostensibly accurate but factually erroneous data, such as fabricated bibliographic references, a phenomenon described as "Artificial Intelligence Hallucinations"[ 20 ]. It has already been proposed that the Dunning-Kruger effect serves as a pertinent framework to consider the actual vs. the perceived competencies that exist regarding the application of AI in research [ 21 ]. Furthermore,the attitudes and expectations surrounding such technologies, just one year following the release of OpenAI’s ChatGPT, can be aptly illustrated by the Gartner Hype Cycle [ 22 ]. Consequently, it is imperative that content generated by AI algorithms, even translations, undergo rigorous validation by subject matter experts.

Moreover, the rapid evolution of AI models, especially deep learning architectures, has created ’black box’ systems where the decision-making process is not transparent [ 23 ]. This opacity can further exacerbate researchers’ trust issues towards AI-generated content. The lack of interpretability can hinder the widespread adoption of AI in scientific research, as researchers might be hesitant to rely on tools they need to understand fully. Efforts are being made in the AI community to develop more interpretable and explainable AI models, but the balance between performance and transparency remains a challenge [ 24 ].

Beyond the ethical implications, another emerging concern is the potential for AI to perpetuate existing biases in the training data or continue "citogenesis"[ 25 ], which represents an insidious form of error propagation within the scientific corpus [ 26 ]. If AI models are trained on biased datasets, they can produce skewed or discriminatory results, leading to flawed conclusions and the perpetuation of systemic inequalities in research. This is particularly concerning in social sciences and medicine, where biased conclusions can have far-reaching implications [ 27 ]. For this reason, researchers must be aware of these pitfalls and advocate for the usage of data that is as unbiased and representative as possible in training AI models. The full spectrum of potential negative outcomes remains largely unquantified. Furthermore, using AI complicates the attribution of accountability, particularly in clinical settings. Ethical concerns, echoed by most of our respondents, coexist with legal considerations [ 28 ].

Additionally, integrating AI into scientific research raises data privacy and security questions [ 29 ]. As AI models often require vast amounts of data for continued training,there is the risk of submitted sensitive information being unintentionally exposed or misused during the process.This is one of the main reasons why several AI companies recently came out with enterprise and on-premise software versions.Such measures are especially pertinent in medical research, where patient data confidentiality is paramount [ 23 , 30 ]. Ensuring robust data encryption and adhering to stringent data handling protocols becomes crucial when incorporating AI into the research workflow.

Various policy options have been tabled to govern the use of AI in the production and editing of scholarly texts. These range from a complete prohibition on using AI-generated content in academic manuscripts to mandates for clear disclosure of AI contributions within the text and reference sections [ 31 ]. Notably, accrediting AI systems as authors appear to be universally rejected.Given these challenges, the concerns identified are legitimate and necessitate comprehensive investigation, particularly as AI technologies continue to advance and diversify in application.

A collaborative approach that includes AI experts, ethicists, policymakers, and researchers is crucial to manage the ethical and technical complexities and fully leverage AI in a responsible and effective manner. Furthermore, it is advisable for journal editors to establish clear guidelines for AI use, as some have already begun [ 14 ], including mandating the disclosure of AI involvement in the research process. Strict policies should be implemented to safeguard the data utilized by AI systems. Human oversight is necessary to interpret the data and results produced by AI. Additionally, an independent group should assess the impact of AI on research outcomes and ethical issues.

Lastly, attention must be paid to the energy consumption of AI systems and their consequent carbon footprint, which can be considerable, especially in the case of large-scale computational models [ 32 ]. AI and machine learning models, particularly those utilizing deep learning, require extensive computational resources and use significant amounts of electricity. To minimize this footprint, researchers should focus on optimizing AI algorithms to increase their energy efficiency and employ these systems only when absolutely necessary. It is essential for researchers to consider the environmental impact of their AI usage, treating ecological sustainability as a critical factor in today’s world.

Future in healthcare

The advent of AI in healthcare is rapidly evolving, and our responders anticipate Big Data Management [ 33 ] and Automated Radiographic Report Generation [ 34 ] to be the most impactful areas influenced by AI applications in the next few years. These results underline the growing recognition of AI’s transformative potential in these domains [ 35 ]. Indeed, the current healthcare landscape generates massive amounts of data from diverse sources, including electronic health records, diagnostic tests, and patient monitoring systems [ 36 ]. AI-powered analytics tools could revolutionize how we understand and interpret this data, thus aiding in more accurate diagnosis and personalized treatment protocols. Similarly, medical imaging studies require considerable time and expertise for interpretation, representing a potential bottleneck in clinical workflow. Automated systems powered by AI can analyze images and rapidly generate reports with a speed and consistency that could vastly improve throughput and possibly contribute to improved patient outcomes, bolstering the assumption that AI-assisted radiologists work better and faster [ 37 ]. By contrast, these systems have been demonstrated to generate more incorrect positive results compared to radiology reports, especially when dealing with multiple or smaller-sized target findings [ 38 ]. Despite these and other limitations such as privacy security concerns, computer-aided diagnosis is promising and could impact several specialties [ 39 ]. In the market, there are already various user-friendly and easy-to-use mobile apps available, designed for healthcare professionals as well as patients, that offer quick access to artificial intelligence tools for obtaining potential diagnoses.Nevertheless, AI currently lacks the precision and capability to make clinical diagnoses, and thus cannot be a substitute for a doctor.

Finally, the development of AI in diagnosis and drug development was also highly rated in the survey. These results mirror current research trends, where AI has been applied for early disease detection and drug discovery processes, significantly cutting down time and costs. Even so, the essential human interaction between patient and clinician remains a core aspect of medical care, making it unlikely that AI will soon replace the need for in-person connection [ 40 ]. Our survey respondents echo this sentiment, as the majority believe clinical doctors will only be partially replaced by technological advancements. Interestingly, in the open-ended responses, among the others, we found this comment “Humans do not want an AI-doctor”. Even though literature tells us that AI could be more empathetic than human doctors [ 41 ], for the moment, everyone agrees.

Limitations

While this study provides valuable insights into the understanding and utilization of Artificial Intelligence (AI) in scientific research, there are some noteworthy limitations. First, the study sample focuses exclusively on corresponding authors from high-impact medical journals. Although this allows us to capture perspectives from researchers at the forefront of scientific advancements, it may limit the generalizability of our findings to the broader scientific and medical community, including early-career researchers and students. Future surveys should aim to include a more diverse range of participants for a fuller picture.

Second, the survey had a low response rate. Physicians are generally challenging to be involved in survey research, and web-based surveys often yield lower participation rates [ 42 ]. Additionally, the accuracy of the email addresses is not guaranteed in email surveys, as evidenced by the emails that were bounced back, likely due to outdated or incorrect institutional email addresses. Nevertheless, although we didn’t conduct an a priori sample size calculation, our aim was to collect responses from at least 300 participants to obtain a substantial perspective on the subject.

Third, the data was gathered through an online survey, which might introduce selection bias as those who are more comfortable with technology and AI may have been more inclined to participate.

Fourth, there was no verification process for the authenticity of the email addresses used in our study, which leaves room for potential inaccuracies in the data collected.

Conclusions

This survey revealed varying degrees of familiarity with AI tools among researchers, with many in high-impact journals beginning to integrate AI into their work. The majority of respondents were from the USA and UK, with 54.1% from English-speaking countries. Only 5.7% indicated extensive familiarity with AI, and 24% used AI tools in scientific content creation, predominantly ChatGPT. Despite low training rates in AI (less than 1%), its use is gradually becoming more prevalent in scientific research and authorship.

Supporting information

S1 appendix. survey questionnaire..

https://doi.org/10.1371/journal.pone.0309208.s001

S2 Appendix. List of the leading 15 medical journals by impact factor.

https://doi.org/10.1371/journal.pone.0309208.s002

• View Article
• PubMed/NCBI
• Google Scholar
• 14. Artificial Intelligence (AI) | Nature Portfolio n.d. https://www.nature.com/nature-portfolio/editorial-policies/ai (accessed April 15, 2024).
• 25. 978: Citogenesis ‐ explain xkcd n.d. https://www.explainxkcd.com/wiki/index.php/978:_Citogenesis (accessed September 3, 2023).

• Alzheimer's disease & dementia
• Arthritis & Rheumatism
• Attention deficit disorders
• Autism spectrum disorders
• Biomedical technology
• Diseases, Conditions, Syndromes
• Endocrinology & Metabolism
• Gastroenterology
• Gerontology & Geriatrics
• Health informatics
• Inflammatory disorders
• Medical economics
• Medical research
• Medications
• Neuroscience
• Obstetrics & gynaecology
• Oncology & Cancer
• Ophthalmology
• Overweight & Obesity
• Parkinson's & Movement disorders
• Psychology & Psychiatry
• Radiology & Imaging
• Sleep disorders
• Sports medicine & Kinesiology
• Vaccination
• Breast cancer
• Cardiovascular disease
• Chronic obstructive pulmonary disease
• Colon cancer
• Coronary artery disease
• Heart attack
• Heart disease
• High blood pressure
• Kidney disease
• Lung cancer
• Multiple sclerosis
• Myocardial infarction
• Ovarian cancer
• Post traumatic stress disorder
• Rheumatoid arthritis
• Schizophrenia
• Skin cancer
• Type 2 diabetes
• Full List »

share this!

August 27, 2024

This article has been reviewed according to Science X's editorial process and policies . Editors have highlighted the following attributes while ensuring the content's credibility:

fact-checked

peer-reviewed publication

trusted source

Novel algorithm improves intracranial EEG accuracy to enhance future patient care

by University of Minnesota Medical School

Published in the Journal of Neural Engineering , a research team led by the University of Minnesota Medical School has evaluated the reliability of human experts in comparison to an automated algorithm in assessing the quality of intracranial electroencephalography (iEEG) data. This research hopes to pave the way for more accurate and efficient seizure detection and localization, ultimately improving outcomes for epilepsy patients.

iEEG is a procedure that measures brain activity by placing electrodes directly on or inside the brain. This detailed information is crucial for diagnosing and treating conditions like epilepsy, where pinpointing the exact source of seizures is essential for effective treatment.

For this study, the research team enlisted 16 experts, including EEG technologists and fellowship-trained neurologists, to rate 1,440 iEEG channels as "good" or "bad." In this study, good meant recording brain activity and bad meant not recording brain activity. Their evaluations were compared to themselves, each other and against the Automated Bad Channel Detection (ABCD) algorithm, which was developed by the Herman Darrow Human Neuroscience Lab at the University of Minnesota.

The ABCD algorithm demonstrated higher accuracy (95.2%) and better overall performance compared to human raters, particularly in identifying channels with high-frequency noise.

"Our findings highlight potential biases and limitations in human-based EEG assessments. The ABCD algorithm's performance suggests a future where automated methods can aid clinicians in improving the accuracy and efficiency of seizure detection, ultimately enhancing patient care," said Alexander Herman, MD, Ph.D., an assistant professor at the U of M Medical School and attending psychiatrist with M Health Fairview.

This research underscores the potential of automated solutions to enhance the reliability and efficiency of iEEG data interpretation—critical for seizure localization and improved patient outcomes.

"This research demonstrates the potential of automated algorithms to outperform human experts in identifying bad EEG channels. By reducing the workload and variability in assessments, we can focus more on clinical decision-making and patient care," said David Darrow, MD, MPH, an assistant professor at the U of M Medical School and neurosurgeon with M Health Fairview

Future research should aim to refine these automated methods further and explore their application in real-time clinical settings .

Explore further

Feedback to editors

Cochlear implant users reveal basic approaches for how people recognize words

Fluid 'stickiness' in female reproductive tract could influence fertility

4 hours ago

Certain diabetes drugs may reduce risk of dementia, Korean study reveals

11 hours ago

New insights into the pathogenesis of amyotrophic lateral sclerosis

12 hours ago

Small RNAs found to boost immune response to tuberculosis

Medication may stop migraines before headache starts, study shows

14 hours ago

Study reveals molecular mechanism behind multiple sclerosis and other autoimmune diseases

Collaborative research cracks the autism code, making the neurodivergent brain visible

Autistic traits, behavioral problems in 7-year-olds linked with gender nonconforming play

16 hours ago

People with mild cases of mental ill-health may be perceived differently depending on presence of diagnostic labels

Related stories.

Researchers introduce enhanced brain signal analysis technique

Mar 27, 2024

Tracking seizures: Brain research aims to improve epilepsy treatment

Jan 19, 2023

Finding epilepsy hotspots before surgery: A faster, non-invasive approach

Aug 27, 2024

What is a 'brain pacemaker?'

Feb 23, 2022

Neural network enables objective assessment of breast symmetry

Jan 26, 2024

New AI tool for PET imaging allows for high-quality, fully automated evaluation of brain tumors

Oct 17, 2023

Recommended for you

Treating depression with psilocybin or escitalopram found to result in different hierarchical brain reconfigurations

21 hours ago

Prioritizing the unexpected: New brain mechanism uncovered

19 hours ago

How zebrafish map their environment: Spatial orientation mechanisms surprisingly similar to our own

18 hours ago

Uncovering the mechanics behind ketamine's rapid antidepressant effects

How humble cells in a little-known organ manage brain inflammation

Let us know if there is a problem with our content.

Use this form if you have come across a typo, inaccuracy or would like to send an edit request for the content on this page. For general inquiries, please use our contact form . For general feedback, use the public comments section below (please adhere to guidelines ).

Please select the most appropriate category to facilitate processing of your request

Thank you for taking time to provide your feedback to the editors.

Your feedback is important to us. However, we do not guarantee individual replies due to the high volume of messages.

E-mail the story

Your email address is used only to let the recipient know who sent the email. Neither your address nor the recipient's address will be used for any other purpose. The information you enter will appear in your e-mail message and is not retained by Medical Xpress in any form.

Newsletter sign up

Get weekly and/or daily updates delivered to your inbox. You can unsubscribe at any time and we'll never share your details to third parties.

More information Privacy policy

Donate and enjoy an ad-free experience

We keep our content available to everyone. Consider supporting Science X's mission by getting a premium account.

E-mail newsletter

• Clinical Research
• Interventional
• Radiation Oncology
• Women's
• Artificial Intelligence
• Enterprise Imaging
• Imaging Informatics
• Informatics
• Compensation
• Education & Training
• Patient Care
• Policy & Regulations
• Practice Management
• Professional Associations
• Experience Stories
• Webinars & Videos

Search form

Nearly half of fda cleared ai medical devices have not been validated on patient data.

New research into the validation processes of artificial intelligence-enabled medical devices has some experts suggesting that their deployment could potentially “pose risks to patient care.” 

The study, which is the product of a collaboration between a team of researchers at the UNC School of Medicine, Duke University, Ally Bank, Oxford University, Colombia University and University of Miami, revealed that many of the AI-enabled medical products approved by the U.S. Food and Drug Administration have not been validated on real patients. The team cautioned that utilizing such tools in medical settings without clinical validation could be putting patients at risk. 

“Although AI device manufacturers boast of the credibility of their technology with FDA authorization, clearance does not mean that the devices have been properly evaluated for clinical effectiveness using real patient data,” first author of the study Sammy Chouffani El Fassi, a MD candidate at the UNC School of Medicine and research scholar at Duke Heart Center, said in a release on the findings. 

The team published their analysis in Nature Medicine . It contains a thorough overview of more than 500 FDA-cleared AI algorithms (each tailored for use in the medical field) and details on how the products were validated. 

They found that approximately 28% of the algorithms had been retrospectively analyzed, with another 28% being prospectively validated. Just 22 of the 521 algorithms were validated using randomized controlled trials—the gold standard for clinical validation. What’s more, 43% did not have publicly available published clinical validation data, and some devices were not validated on human data at all, but rather “phantom images.” 

A call for increased credibility

Currently, the FDA’s 2023 draft guidance on the approval process for AI devices does not differentiate between the type of validation the agency recommends manufacturers use. 

Although the agency has worked to make improvements regarding AI regulation and updates, the authors of this latest study believe that more needs to be done to promote the public’s trust in the manufacturers developing these products. They suggested that the agency start by clearly distinguishing between the type of processes used to validate AI products in its manufacturer guidelines. 

“We shared our findings with directors at the FDA who oversee medical device regulation, and we expect our work will inform their regulatory decision making,” Chouffani El Fassi said. “We also hope that our publication will inspire researchers and universities globally to conduct clinical validation studies on medical AI to improve the safety and effectiveness of these technologies. We’re looking forward to the positive impact this project will have on patient care at a large scale.” 

There are currently more than 900 AI medical devices that have received the FDA’s official stamp of approval, with products tailored to radiology accounting for over 70% of the clearances. 

The full list of AI medical device clearances can be accessed here . 

FDA clears AI-powered POCUS platform for structural heart disease, heart failure

Ai rules out abnormal findings on chest x-rays, significantly reducing workloads, fda adds dozens of ai-enabled radiology applications to list of clearances.

In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She joined Innovate Healthcare in 2021 and has since put her unique expertise to use in her editorial role with Health Imaging.

Related Content

What is Patient Data? Types, Uses & Hospital Patient Databases

What is patient data.

Patient data refers to any information related to an individual’s health and medical history. It includes personal details, such as name, age, and contact information, as well as medical records, diagnoses, treatments, and test results. This data is collected and stored by healthcare providers to facilitate patient care, research, and administrative purposes.

Best Patient Data Databases & Datasets

Here is Datarade's curated selection of top Patient Data. These trusted databases and datasets offer high-quality, up-to-date information.

Data Collection by Shaip: Text, Audio, Image, Video for AI & ML Training

Syntegra Synthetic EHR Data | Structured Healthcare Electronic Health Record Data

Diaceutics Repository of Diagnostic Testing Data

Syntegra Synthetic Claims Data | Medicare Claims | Multiple Formats

Monetize data on datarade marketplace, patient data explained.

Safeguarding patient data is crucial to ensure privacy, confidentiality, and compliance with data protection regulations.

Examples of patient data include demographic information (such as name, age, and gender), medical history, laboratory test results, diagnoses, treatments, medications, and imaging reports. Patient data is used by healthcare professionals to provide appropriate care, make informed decisions, monitor patient progress, and conduct research. In this page, you’ll find the best data sources for patient data.

Patient Data Collection

Anytime you go to a medical professional your data will be recorded by the doctors and nurses for their internal databases. This could either be by tests and observations they make on you or by information that you freely provide. It is the latter which especially gives an insight into the patient’s behaviour and lifestyle. Similarly, the rise in health apps and products, such as smart watches, means that individual behavior is being more constantly recorded and from this patient data has evolved to include behavior data as well as basic medical information. Research projects also provide patient data as well as historical records and family medical information.

Main Attributes

The attributes of patient data can be split into two sections: Traditional medical data - Traditional data includes information such as health history, diagnoses, current medications, vaccination record, and ongoing or historical treatments. Newer patient information - Contemporary patient data has evolved to include more information which is less medically-focussed, so that it includes information about people who aren’t necessarily sick. New patient information includes demographic and behavioral intelligence, such as lifestyle, eating habits or patient behaviour.

Patient data is used by both medical professionals and businesses: Healthcare professionals and scientisits - Data records help manage and diagnonse individuals with medical issues. It also provides reference information for researchers to study diseases or the human body in general. Businesses - Pharmaceutical companies rely on patient data for market research to identify gaps in the market for healthcare and medical solutions. Marketers for these companies can also use patient data for audience segmentation to target specific patients suffering the same sickness with ads for a suitable treatment.

How can a user assess the quality of Patient Data?

Because it’s classed as PII (personally identifiable information) patient data is protected by strict privacy laws. Moreover, there are doctor-patient confidentiality laws which mean it’s essential that your patient dataset is privacy-compliant. Always check that your patient data provider is GDPR and CCPA-compliant, and that their data sources follow ethical guidelines.

Due to the wide variety of patient data available, it is important to pick a dataset that is suited to your personal needs. Ask for a sample of data from the data provider before buying to ensure it matches what you are looking for and always check the data provider’s reviews before you make any purchase.

Frequently Asked Questions

Where can i buy patient data.

Data providers and vendors listed on Datarade sell Patient Data products and samples. Popular Patient Data products and datasets available on our platform are Data Collection by Shaip: Text, Audio, Image, Video for AI & ML Training by ShAIp , Syntegra Synthetic EHR Data | Structured Healthcare Electronic Health Record Data by Syntegra , and Diaceutics Repository of Diagnostic Testing Data by Diaceutics .

How can I get Patient Data?

You can get Patient Data via a range of delivery methods - the right one for you depends on your use case. For example, historical Patient Data is usually available to download in bulk and delivered using an S3 bucket. On the other hand, if your use case is time-critical, you can buy real-time Patient Data APIs, feeds and streams to download the most up-to-date intelligence.

Users also searched for

Medical School

• MD Students
• Residents & Fellows
• Faculty & Staff
• Duluth Campus Leadership
• Department Heads
• Dean's Distinguished Research Lectureship
• Dean's Tribute to Excellence in Research
• Dean’s Tribute to Excellence in Education
• Twin Cities Campus
• Office of the Regional Dean
• Office of Research Support, Duluth Campus
• CentraCare Regional Campus St. Cloud
• Office of Diversity, Equity & Inclusion
• Office of Faculty Affairs
• Grant Support
• Medical Student Research Opportunities
• The Whiteside Institute for Clinical Research
• Office of Research, Medical School
• Graduate Medical Education Office
• Office of Graduate and Postdoctoral Studies
• Departments
• Centers & Institutes
• Dr. James E. Rubin Medical Memorial Award
• Fisch Art of Medicine Student Awards
• Graduating Medical Student Research Award
• Veneziale-Steer Award
• Dr. Marvin and Hadassah Bacaner Research Awards
• Excellence in Geriatric Scholarship
• Distinguished Mentoring Award
• Distinguished Teaching Awards
• Cecil J. Watson Award
• Exceptional Affiliate Faculty Teaching Award
• Exceptional Primary Care Community Faculty Teaching Award
• Herz Faculty Teaching Development Awards
• The Leonard Tow Humanism in Medicine Awards
• Year 1+2 Educational Innovative Award
• Alumni Philanthropy and Service Award
• Distinguished Alumni Award
• Early Distinguished Career Alumni Award
• Harold S. Diehl Award
• Nomination Requirements
• Duluth Scholarships & Awards
• Account Management
• Fundraising Assistance
• Research & Equipment Grants
• Medical School Scholarships
• Student Research Grants
• Board members
• Alumni Celebration & Reunions
• Alumni Relations - Office of Alumni Relations
• Maximizing Medical Practice Conference
• Diversity & Inclusion
• The Bob Allison Ataxia Research Center
• Children's Health
• Orthopaedics
• Regenerative Medicine
• Scholarships
• Transplantation
• MD Student Virtual Tour
• Why Minnesota?
• Facts & Figures
• Annika Tureson
• Caryn Wolter
• Crystal Chang
• Emily Stock
• Himal Purani
• Jamie Stang
• Jordan Ammons
• Julia Klein
• Julia Meyer
• Kirsten Snook
• Madison Sundlof
• Rashika K Shetty
• Rishi Sharma
• Savannah Maynard
• Prerequisite Courses
• Student National Medical Association Mentor Program
• Minority Association of Pre-Med Students
• Twin Cities Entering Class
• Next Steps: Accepted Students
• Degrees Offered
• Preceptor Resources (MD)
• Doctor of Physical Therapy (DPT)
• Message from the Director
• Clinical Training
• Curriculum & Timeline
• Integrated Physician Scientist Training
• MSTP Student Mentoring and Career Development
• External Fellowships
• Student Directory
• Student Advisory Committee
• Staff Directory
• Steering Committee
• Alumni Directory
• MSTP Student First Author Papers
• Thesis Defense
• MSTP Graduate Residency Match Information
• MSTP Code of Ethics
• Research Experience
• International Applicants
• Prerequisites
• Applicant Evaluation
• The Interview Process
• Application Process
• Financial Support
• Pre-MSTP Summer Research Program
• MSTP Virtual Tour
• Physician Scientist Training Programs
• Students with Disabilities
• Clinical Continuity & Mentoring Program
• Mental Health and Well-Being
• Leadership in Diversity Fellowship
• M1 and M2 Research Meeting
• Monthly Student Meetings
• MSTP Annual Retreat
• Preparation
• Women in Science & Medicine
• Graduate Programs/Institute Activities & Seminars
• CTSI Translational Research Development Program (TRDP)
• Grant Opportunities
• Graduate Programs
• Residencies & Fellowships
• Individualized Pathways
• Continuing Professional Development
• CQI Initiative
• Improvement Summary
• Quality Improvement Communication
• Student Involvement
• Student Voice
• Contact the Medical School Research Office
• Funding Opportunities
• Research Ethics
• Research Support
• Training Grants
• Veteran's Affairs (VA)
• ALPS COVID For Patients
• 3D/Virtual Reality Procedural Training
• Addressing Social Determinants of Health in the Age of COVID-19
• Augmented Reality Remote Procedural Training and Proctoring
• COVID19: Outbreaks and the Media
• Foundations in Health Equity
• Global Medicine in a Changing Educational World
• INMD 7013: COVID-19 Crisis Innovation Lab A course examining the COVID-19 crisis
• Intentional Observational Exercises - Virtual Critical Care Curriculum
• Live Suture Sessions for EMMD7500
• M Simulation to Provide PPE training for UMN students returning to clinical environments
• Medical Student Elective course: COVID-19 Contact Tracing with MDH
• Medical Students in the M Health Fairview System Operations Center
• Telehealth Management in Pandemics
• Telemedicine Acting Internship in Pediatrics
• The Wisdom of Literature in a Time of Plague
• Virtual Simulations for EMMD7500 Students
• COVID-19 Publications
• Blood & Marrow Transplant
• Genome Engineering
• Immunology & Infectious Disease
• Magnetic Resonance Imaging
• News & Events
• Support the Institute
• Lung Health
• Medical Discovery Team on Addiction Faculty
• Impact on Education
• Meet Our Staff
• Donation Criteria
• Health Care Directive Information
• Inquiry Form
• Other Donation Organizations
• Death Certificate Information
• Social Security Administration Notification and Death Benefit Information
• Grief Support
• Deceased Do Not Contact Registration and Preventing Identity Theft
• Service of Gratitude
• For Educators and Researchers
• Clinical Repository
• Patient Care
• Blue Ridge Research Rankings
• Where Discovery Creates Hope

Cutting-edge algorithm improves intracranial EEG accuracy to improve future patient care

MINNEAPOLIS/ST. PAUL (08/27/2024) —  Published in the  Journal of Neural Engineering , a research team led by the University of Minnesota Medical School evaluated the reliability of human experts in comparison to an automated algorithm in assessing the quality of intracranial electroencephalography (iEEG) data. This research hopes to pave the way for more accurate and efficient seizure detection and localization, ultimately improving outcomes for epilepsy patients.

iEEG is a procedure that measures brain activity by placing electrodes directly on or inside the brain. This detailed information is crucial for diagnosing and treating conditions like epilepsy, where pinpointing the exact source of seizures is essential for effective treatment.

For this study, the research team enlisted 16 experts, including EEG technologists and fellowship-trained neurologists, to rate 1,440 iEEG channels as “good” or “bad.” In this study, good meant recording brain activity and bad meant not recording brain activity. Their evaluations were compared to themselves, each other and against the Automated Bad Channel Detection (ABCD) algorithm, which was developed by the  Herman Darrow Human Neuroscience Lab at the University of Minnesota. 

The ABCD algorithm demonstrated higher accuracy (95.2%) and better overall performance compared to human raters, particularly in identifying channels with high-frequency noise.

"Our findings highlight potential biases and limitations in human-based EEG assessments. The ABCD algorithm's performance suggests a future where automated methods can aid clinicians in improving the accuracy and efficiency of seizure detection, ultimately enhancing patient care,” said  Alexander Herman, MD, PhD , an assistant professor at the U of M Medical School and attending psychiatrist with M Health Fairview. 

This research underscores the potential of automated solutions to enhance the reliability and efficiency of iEEG data interpretation — critical for seizure localization and improved patient outcomes.

"This research demonstrates the potential of automated algorithms to outperform human experts in identifying bad EEG channels. By reducing the workload and variability in assessments, we can focus more on clinical decision-making and patient care," said  David Darrow, MD, MPH , an assistant professor at the U of M Medical School and neurosurgeon with M Health Fairview 

Future research should aim to refine these automated methods further and explore their application in real-time clinical settings.

Funding was provided by the  Institute for Translational Neuroscience and  MnDRIVE Brain Conditions .

About the University of Minnesota Medical School The University of Minnesota Medical School is at the forefront of learning and discovery, transforming medical care and educating the next generation of physicians. Our graduates and faculty produce high-impact biomedical research and advance the practice of medicine.  We acknowledge that the U of M Medical School is located on traditional, ancestral and contemporary lands of the Dakota and the Ojibwe, and scores of other Indigenous people, and we affirm our commitment to tribal communities and their sovereignty as we seek to improve and strengthen our relations with tribal nations. For more information about the U of M Medical School, please visit  med.umn.edu . 

• Article Information

Data Sharing Statement

• As Ozempic’s Popularity Soars, Here’s What to Know About Semaglutide and Weight Loss JAMA Medical News & Perspectives May 16, 2023 This Medical News article discusses chronic weight management with semaglutide, sold under the brand names Ozempic and Wegovy. Melissa Suran, PhD, MSJ
• Patents and Regulatory Exclusivities on GLP-1 Receptor Agonists JAMA Special Communication August 15, 2023 This Special Communication used data from the US Food and Drug Administration to analyze how manufacturers of brand-name glucagon-like peptide 1 (GLP-1) receptor agonists have used patent and regulatory systems to extend periods of market exclusivity. Rasha Alhiary, PharmD; Aaron S. Kesselheim, MD, JD, MPH; Sarah Gabriele, LLM, MBE; Reed F. Beall, PhD; S. Sean Tu, JD, PhD; William B. Feldman, MD, DPhil, MPH
• What to Know About Wegovy’s Rare but Serious Adverse Effects JAMA Medical News & Perspectives December 12, 2023 This Medical News article discusses Wegovy, Ozempic, and other GLP-1 receptor agonists used for weight management and type 2 diabetes. Kate Ruder, MSJ
• GLP-1 Receptor Agonists and Gastrointestinal Adverse Events—Reply JAMA Comment & Response March 12, 2024 Ramin Rezaeianzadeh, BSc; Mohit Sodhi, MSc; Mahyar Etminan, PharmD, MSc
• GLP-1 Receptor Agonists and Gastrointestinal Adverse Events JAMA Comment & Response March 12, 2024 Karine Suissa, PhD; Sara J. Cromer, MD; Elisabetta Patorno, MD, DrPH
• GLP-1 Receptor Agonist Use and Risk of Postoperative Complications JAMA Research Letter May 21, 2024 This cohort study evaluates the risk of postoperative respiratory complications among patients with diabetes undergoing surgery who had vs those who had not a prescription fill for glucagon-like peptide 1 receptor agonists. Anjali A. Dixit, MD, MPH; Brian T. Bateman, MD, MS; Mary T. Hawn, MD, MPH; Michelle C. Odden, PhD; Eric C. Sun, MD, PhD
• Glucagon-Like Peptide-1 Receptor Agonist Use and Risk of Gallbladder and Biliary Diseases JAMA Internal Medicine Original Investigation May 1, 2022 This systematic review and meta-analysis of 76 randomized clinical trials examines the effects of glucagon-like peptide-1 receptor agonist use on the risk of gallbladder and biliary diseases. Liyun He, MM; Jialu Wang, MM; Fan Ping, MD; Na Yang, MM; Jingyue Huang, MM; Yuxiu Li, MD; Lingling Xu, MD; Wei Li, MD; Huabing Zhang, MD
• Cholecystitis Associated With the Use of Glucagon-Like Peptide-1 Receptor Agonists JAMA Internal Medicine Research Letter October 1, 2022 This case series identifies cases reported in the US Food and Drug Administration Adverse Event Reporting System of acute cholecystitis associated with use of glucagon-like peptide-1 receptor agonists that did not have gallbladder disease warnings in their labeling. Daniel Woronow, MD; Christine Chamberlain, PharmD; Ali Niak, MD; Mark Avigan, MDCM; Monika Houstoun, PharmD, MPH; Cindy Kortepeter, PharmD

See More About

Select your interests.

Customize your JAMA Network experience by selecting one or more topics from the list below.

• Academic Medicine
• Acid Base, Electrolytes, Fluids
• Allergy and Clinical Immunology
• American Indian or Alaska Natives
• Anesthesiology
• Anticoagulation
• Art and Images in Psychiatry
• Artificial Intelligence
• Assisted Reproduction
• Bleeding and Transfusion
• Caring for the Critically Ill Patient
• Challenges in Clinical Electrocardiography
• Climate and Health
• Climate Change
• Clinical Challenge
• Clinical Decision Support
• Clinical Implications of Basic Neuroscience
• Clinical Pharmacy and Pharmacology
• Complementary and Alternative Medicine
• Consensus Statements
• Coronavirus (COVID-19)
• Critical Care Medicine
• Cultural Competency
• Dental Medicine
• Dermatology
• Diabetes and Endocrinology
• Diagnostic Test Interpretation
• Drug Development
• Electronic Health Records
• Emergency Medicine
• End of Life, Hospice, Palliative Care
• Environmental Health
• Equity, Diversity, and Inclusion
• Facial Plastic Surgery
• Gastroenterology and Hepatology
• Genetics and Genomics
• Genomics and Precision Health
• Global Health
• Guide to Statistics and Methods
• Hair Disorders
• Health Care Delivery Models
• Health Care Economics, Insurance, Payment
• Health Care Quality
• Health Care Reform
• Health Care Safety
• Health Care Workforce
• Health Disparities
• Health Inequities
• Health Policy
• Health Systems Science
• History of Medicine
• Hypertension
• Images in Neurology
• Implementation Science
• Infectious Diseases
• Innovations in Health Care Delivery
• JAMA Infographic
• Law and Medicine
• Leading Change
• Less is More
• LGBTQIA Medicine
• Lifestyle Behaviors
• Medical Coding
• Medical Devices and Equipment
• Medical Education
• Medical Education and Training
• Medical Journals and Publishing
• Mobile Health and Telemedicine
• Narrative Medicine
• Neuroscience and Psychiatry
• Notable Notes
• Nutrition, Obesity, Exercise
• Obstetrics and Gynecology
• Occupational Health
• Ophthalmology
• Orthopedics
• Otolaryngology
• Pain Medicine
• Palliative Care
• Pathology and Laboratory Medicine
• Patient Care
• Patient Information
• Performance Improvement
• Performance Measures
• Perioperative Care and Consultation
• Pharmacoeconomics
• Pharmacoepidemiology
• Pharmacogenetics
• Pharmacy and Clinical Pharmacology
• Physical Medicine and Rehabilitation
• Physical Therapy
• Physician Leadership
• Population Health
• Primary Care
• Professional Well-being
• Professionalism
• Psychiatry and Behavioral Health
• Public Health
• Pulmonary Medicine
• Regulatory Agencies
• Reproductive Health
• Research, Methods, Statistics
• Resuscitation
• Rheumatology
• Risk Management
• Scientific Discovery and the Future of Medicine
• Shared Decision Making and Communication
• Sleep Medicine
• Sports Medicine
• Stem Cell Transplantation
• Substance Use and Addiction Medicine
• Surgical Innovation
• Surgical Pearls
• Teachable Moment
• Technology and Finance
• The Art of JAMA
• The Arts and Medicine
• The Rational Clinical Examination
• Tobacco and e-Cigarettes
• Translational Medicine
• Trauma and Injury
• Treatment Adherence
• Ultrasonography
• Users' Guide to the Medical Literature
• Vaccination
• Venous Thromboembolism
• Veterans Health
• Women's Health
• Workflow and Process
• Wound Care, Infection, Healing

Others Also Liked

• Download PDF
• X Facebook More LinkedIn

Sodhi M , Rezaeianzadeh R , Kezouh A , Etminan M. Risk of Gastrointestinal Adverse Events Associated With Glucagon-Like Peptide-1 Receptor Agonists for Weight Loss. JAMA. 2023;330(18):1795–1797. doi:10.1001/jama.2023.19574

Manage citations:

© 2024

• Permissions

Risk of Gastrointestinal Adverse Events Associated With Glucagon-Like Peptide-1 Receptor Agonists for Weight Loss

• 1 Faculty of Medicine, University of British Columbia, Vancouver, British Columbia, Canada
• 2 StatExpert Ltd, Laval, Quebec, Canada
• 3 Department of Ophthalmology and Visual Sciences and Medicine, University of British Columbia, Vancouver, Canada
• Medical News & Perspectives As Ozempic’s Popularity Soars, Here’s What to Know About Semaglutide and Weight Loss Melissa Suran, PhD, MSJ JAMA
• Special Communication Patents and Regulatory Exclusivities on GLP-1 Receptor Agonists Rasha Alhiary, PharmD; Aaron S. Kesselheim, MD, JD, MPH; Sarah Gabriele, LLM, MBE; Reed F. Beall, PhD; S. Sean Tu, JD, PhD; William B. Feldman, MD, DPhil, MPH JAMA
• Medical News & Perspectives What to Know About Wegovy’s Rare but Serious Adverse Effects Kate Ruder, MSJ JAMA
• Comment & Response GLP-1 Receptor Agonists and Gastrointestinal Adverse Events—Reply Ramin Rezaeianzadeh, BSc; Mohit Sodhi, MSc; Mahyar Etminan, PharmD, MSc JAMA
• Comment & Response GLP-1 Receptor Agonists and Gastrointestinal Adverse Events Karine Suissa, PhD; Sara J. Cromer, MD; Elisabetta Patorno, MD, DrPH JAMA
• Research Letter GLP-1 Receptor Agonist Use and Risk of Postoperative Complications Anjali A. Dixit, MD, MPH; Brian T. Bateman, MD, MS; Mary T. Hawn, MD, MPH; Michelle C. Odden, PhD; Eric C. Sun, MD, PhD JAMA
• Original Investigation Glucagon-Like Peptide-1 Receptor Agonist Use and Risk of Gallbladder and Biliary Diseases Liyun He, MM; Jialu Wang, MM; Fan Ping, MD; Na Yang, MM; Jingyue Huang, MM; Yuxiu Li, MD; Lingling Xu, MD; Wei Li, MD; Huabing Zhang, MD JAMA Internal Medicine
• Research Letter Cholecystitis Associated With the Use of Glucagon-Like Peptide-1 Receptor Agonists Daniel Woronow, MD; Christine Chamberlain, PharmD; Ali Niak, MD; Mark Avigan, MDCM; Monika Houstoun, PharmD, MPH; Cindy Kortepeter, PharmD JAMA Internal Medicine

Glucagon-like peptide 1 (GLP-1) agonists are medications approved for treatment of diabetes that recently have also been used off label for weight loss. 1 Studies have found increased risks of gastrointestinal adverse events (biliary disease, 2 pancreatitis, 3 bowel obstruction, 4 and gastroparesis 5 ) in patients with diabetes. 2 - 5 Because such patients have higher baseline risk for gastrointestinal adverse events, risk in patients taking these drugs for other indications may differ. Randomized trials examining efficacy of GLP-1 agonists for weight loss were not designed to capture these events 2 due to small sample sizes and short follow-up. We examined gastrointestinal adverse events associated with GLP-1 agonists used for weight loss in a clinical setting.

We used a random sample of 16 million patients (2006-2020) from the PharMetrics Plus for Academics database (IQVIA), a large health claims database that captures 93% of all outpatient prescriptions and physician diagnoses in the US through the International Classification of Diseases, Ninth Revision (ICD-9) or ICD-10. In our cohort study, we included new users of semaglutide or liraglutide, 2 main GLP-1 agonists, and the active comparator bupropion-naltrexone, a weight loss agent unrelated to GLP-1 agonists. Because semaglutide was marketed for weight loss after the study period (2021), we ensured all GLP-1 agonist and bupropion-naltrexone users had an obesity code in the 90 days prior or up to 30 days after cohort entry, excluding those with a diabetes or antidiabetic drug code.

Patients were observed from first prescription of a study drug to first mutually exclusive incidence (defined as first ICD-9 or ICD-10 code) of biliary disease (including cholecystitis, cholelithiasis, and choledocholithiasis), pancreatitis (including gallstone pancreatitis), bowel obstruction, or gastroparesis (defined as use of a code or a promotility agent). They were followed up to the end of the study period (June 2020) or censored during a switch. Hazard ratios (HRs) from a Cox model were adjusted for age, sex, alcohol use, smoking, hyperlipidemia, abdominal surgery in the previous 30 days, and geographic location, which were identified as common cause variables or risk factors. 6 Two sensitivity analyses were undertaken, one excluding hyperlipidemia (because more semaglutide users had hyperlipidemia) and another including patients without diabetes regardless of having an obesity code. Due to absence of data on body mass index (BMI), the E-value was used to examine how strong unmeasured confounding would need to be to negate observed results, with E-value HRs of at least 2 indicating BMI is unlikely to change study results. Statistical significance was defined as 2-sided 95% CI that did not cross 1. Analyses were performed using SAS version 9.4. Ethics approval was obtained by the University of British Columbia’s clinical research ethics board with a waiver of informed consent.

Our cohort included 4144 liraglutide, 613 semaglutide, and 654 bupropion-naltrexone users. Incidence rates for the 4 outcomes were elevated among GLP-1 agonists compared with bupropion-naltrexone users ( Table 1 ). For example, incidence of biliary disease (per 1000 person-years) was 11.7 for semaglutide, 18.6 for liraglutide, and 12.6 for bupropion-naltrexone and 4.6, 7.9, and 1.0, respectively, for pancreatitis.

Use of GLP-1 agonists compared with bupropion-naltrexone was associated with increased risk of pancreatitis (adjusted HR, 9.09 [95% CI, 1.25-66.00]), bowel obstruction (HR, 4.22 [95% CI, 1.02-17.40]), and gastroparesis (HR, 3.67 [95% CI, 1.15-11.90) but not biliary disease (HR, 1.50 [95% CI, 0.89-2.53]). Exclusion of hyperlipidemia from the analysis did not change the results ( Table 2 ). Inclusion of GLP-1 agonists regardless of history of obesity reduced HRs and narrowed CIs but did not change the significance of the results ( Table 2 ). E-value HRs did not suggest potential confounding by BMI.

This study found that use of GLP-1 agonists for weight loss compared with use of bupropion-naltrexone was associated with increased risk of pancreatitis, gastroparesis, and bowel obstruction but not biliary disease.

Given the wide use of these drugs, these adverse events, although rare, must be considered by patients who are contemplating using the drugs for weight loss because the risk-benefit calculus for this group might differ from that of those who use them for diabetes. Limitations include that although all GLP-1 agonist users had a record for obesity without diabetes, whether GLP-1 agonists were all used for weight loss is uncertain.

Accepted for Publication: September 11, 2023.

Published Online: October 5, 2023. doi:10.1001/jama.2023.19574

Correction: This article was corrected on December 21, 2023, to update the full name of the database used.

Corresponding Author: Mahyar Etminan, PharmD, MSc, Faculty of Medicine, Departments of Ophthalmology and Visual Sciences and Medicine, The Eye Care Center, University of British Columbia, 2550 Willow St, Room 323, Vancouver, BC V5Z 3N9, Canada ( [email protected] ).

Author Contributions: Dr Etminan had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Sodhi, Rezaeianzadeh, Etminan.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Sodhi, Rezaeianzadeh, Etminan.

Critical review of the manuscript for important intellectual content: All authors.

Statistical analysis: Kezouh.

Obtained funding: Etminan.

Administrative, technical, or material support: Sodhi.

Supervision: Etminan.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was funded by internal research funds from the Department of Ophthalmology and Visual Sciences, University of British Columbia.

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement .

• Register for email alerts with links to free full-text articles
• Access PDFs of free articles
• Manage your interests
• Save searches and receive search alerts

Ohio State nav bar

The Ohio State University

• BuckeyeLink
• Find People
• Search Ohio State

A human-centered AI tool to improve sepsis management

Proposed model’s features based on clinician feedback.

A proposed artificial intelligence tool to support clinician decision-mak ing about hospital patients at risk for sepsis has an unusual feature: accounting for its lack of certainty and suggesting what demographic data, vital signs and lab test results it needs to improve its predictive performance. 

The system, called SepsisLab, was developed based on feedback from doctors and nurses who treat patients in the emergency departments and ICUs where sepsis , the body’s overwhelming response to an infection, is most commonly seen. They reported dissatisfaction with an existing AI-assisted tool that generates a patient risk prediction score using only electronic health records, but no input data from clinicians. 

Scientists at The Ohio State University designed SepsisLab to be able to predict a patient’s sepsis risk within four hours – but while the clock ticks, the system identifies missing patient information, quantifies how essential it is, and gives a visual picture to clinicians of how specific information will affect the final risk prediction. Experiments using a combination of publicly available and proprietary patient data showed that adding 8% of the recommended data improved the system’s sepsis prediction accuracy by 11%.

“The existing model represents a more a traditional human-AI competition paradigm, generating numerous annoying false alarms in ICUs and emergency rooms without listening to clinicians,” said senior study author Ping Zhang , associate professor of computer science and engineering and biomedical informatics  at Ohio State.  

“The idea is we need to involve AI in every intermediate step of decision-making by adopting the ‘AI-in-the-human-loop’ concept. We’re not just developing a tool – we also recruited physicians into the project. This is a real collaboration between computer scientists and clinicians to develop a human-centered system that puts the physician in the driver’s seat.” 

The research was published Aug. 24 in KDD ’24: Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining and will be presented orally Wednesday (Aug. 28) at SIGKDD 2024 in Barcelona, Spain.

Sepsis is a life-threatening medical emergency – it can rapidly lead to organ failure – but it’s not easy to diagnose because its symptoms of fever, low blood pressure, increasing heart rate and breathing problems can look like a lot of other conditions. This work builds upon a previous machine learning model developed by Zhang and colleagues that estimated the optimal time to give antibiotics to patients with a suspected case of sepsis. 

SepsisLab is designed to come up with a risk prediction quickly, but produces a new prediction every hour after new patient data has been added to the system . 

“When a patient first comes in, there are many missing values, especially for lab tests,” said first author Changchang Yin , a computer science and engineering PhD student in Zhang’s Artificial Intelligence in Medicine lab.  

In most AI models, missing data points are accounted for with a single assigned value – a process called imputation – “but the imputation model could suffer from uncertainty that can be propagated to the downstream prediction model,” Yin said. 

“If the imputation model cannot accurately impute the missing value and it’s a very important value, the variable should be observed. Our active sensing algorithm aims to find such missing values and tell clinicians what additional variables they might need to observe – variables that can make the prediction model more accurate.” 

Equally important to removing uncertainty from the system over the passage of time is providing clinicians with actionable recommendations. These include lab tests rank-ordered based on their value to the diagnostic process and estimates of how a patient’s sepsis risk would change depending on specific clinical treatments. 

Experiments showed adding 8% of the new data from lab tests, vital signs and other high-value variables reduced the propagated uncertainty in the model by 70% – contributing to its 11% improvement in sepsis risk accuracy. 

“The algorithm can select the most important variables, and the physician’s action reduces the uncertainty,” said Zhang, also a core faculty member in Ohio State’s  Translational Data Analytics Institute . “This fundamental mathematics work is the most important technical innovation – the backbone of the research.” 

Zhang sees human-centered AI as part of the future of medicine – but only if AI interacts with clinicians in a way that makes them trust the system. 

“This is not about building an AI system that can conquer the world,” he said. “The center of medicine is hypothesis testing and making decisions minute after minute that are not just ‘yes’ or ‘no.’ We envision a person at the center of the interaction using AI to help that human feel superhuman.”

This research was supported by the National Science Foundation, the National Institutes of Health and an Ohio State President’s Research Excellence Accelerator Grant. Zhang has received additional NIH funding to continue collaborating with clinicians on this work. 

Additional co-authors include Jeffrey Caterino of The Ohio State University Wexner Medical Center, Bingsheng Yao and Dakuo Wang of Northeastern University, and Pin-Yu Chen of IBM Research.

More Ohio State News

Ohio state skull session livestream season kicks off saturday.

The Buckeyes kick off their 2024 football season Saturday in Ohio Stadium. For the fourth consecutive year, The Ohio State University Marching Band’s pregame Skull Session will be livestreamed. 

Bringing environmental justice to disadvantaged communities

Not all communities in the United States face the same risks for environmental problems such as air pollution, noise and wastewater.  But how can federal agencies fairly identify which areas deserve the most help?

Locked in a glacier, viruses adapted to survive extreme weather

Ancient viruses preserved in glacial ice hold valuable information about changes in Earth’s climate, a new study suggests.

Contact: Admissions | Webmaster | Page maintained by University Communications

Request an alternate format of this page | Web Services Status | Nondiscrimination notice

43% of FDA-Approved AI-based Medical Devices Are Trained Using Fake Patient Data

Close to half of the AI devices authorized for use by the U.S. Food and Drug Administration (FDA) are trained without using clinical data from actual patients, according to new findings by an international team of researchers.

Artificial intelligence offers many promising benefits to the medical field, from creating and updating patient charts as needed to diagnosing health issues with unprecedented accuracy. However, implementing these tools is becoming a challenge. The adoption of AI in medicine has been met with skepticism, mainly due to concerns about patient privacy, bias, and device accuracy.

A multi-institutional team of researchers from the University of North Carolina (UNC) School of Medicine, Duke University, Ally Bank, Oxford University, Columbia University, and the University of Miami is addressing these concerns. Led by Sammy Chouffani El Fassi, an MD candidate at UNC, and Gail E. Henderson, Ph.D., a professor at UNC, the team conducted a thorough analysis of clinical validation data for over 500 AI medical devices.

Their findings, published in Nature Medicine, revealed that nearly half of the AI devices authorized by the U.S. Food and Drug Administration (FDA) lacked reported clinical data from actual patients.

“Although AI device manufacturers boast of the credibility of their technology with FDA authorization, clearance does not mean that the devices have been properly evaluated for clinical effectiveness using real patient data,” Chouffani El Fassi said in a recent statement . “With these findings, we hope to encourage the FDA and industry to boost the credibility of device authorization by conducting clinical validation studies on these technologies and making the results of such studies publicly available.”

The Rise of AI Use in Medicine

For the past decade, the use of AI in medicine and healthcare has been steadily increasing. Part of this demand comes from the shortage of medical professionals and the rise of telehealth, where patients are diagnosed remotely instead of going in person. The COVID-19 pandemic significantly increased telehealth visits, driving companies to invest in virtual software to host these visits. AI has been used to help manage this software, update patient forms , and assist with scheduling and other logistics.

AI’s use in medicine can not only save time but could also help save significant costs in the long term. According to a 2020 paper : “It is estimated that AI applications can cut annual US healthcare costs by $150 billion in 2026.”

AI-based medical devices, in particular, can streamline processes, reduce the need for expensive diagnostic tests, and minimize human error, translating into substantial cost savings for healthcare systems. However, the cost-effectiveness of these AI tools hinges on their accuracy and reliability in real-world clinical settings.

Training with Fake Patient Data

As the number of FDA authorized AI-based medical devices has risen from two to 69 since 2016 , these approved devices should have been trained on real patient data to insure the best accuracy in their applications.

Yet, the study found that 43% (or 226) of these devices (our of 521 approved devices) lacked published clinical validation data, and some even relied on “phantom images” rather than real patient data.

Bizarre Tachyons That May Be Able to Send Data Back in Time Could be Reconciled with Special Relativity

“A lot of the devices that came out after 2016 were created new, or maybe they were similar to a product that already was on the market,” Henderson explained in a recent statement . “Using these hundreds of devices in this database, we wanted to determine what it really means for an AI medical device to be FDA-authorized.”

Lack of Reinforcement

To make matters worse, the team found that in the September 2023 FDA guidance document, the most recent version, there is no distinction between different types of clinical validation studies for recommendations to device manufacturers, meaning that there’s weaker reinforcement in making sure these devices are trained on real patient data.

In the study, the researchers advocated for clearer distinctions between different types of clinical validation—retrospective, prospective, and randomized controlled trials—within FDA guidelines. They hope their findings will influence FDA regulatory decisions and inspire global research efforts to improve the safety and effectiveness of AI medical technologies.

“We shared our findings with directors at the FDA who oversee medical device regulation, and we expect our work will inform their regulatory decision-making,” said Chouffani El Fassi. “We also hope that our publication will inspire researchers and universities globally to conduct clinical validation studies on medical AI to improve the safety and effectiveness of these technologies. We’re looking forward to the positive impact this project will have on patient care at a large scale.”

Kenna Hughes-Castleberry is the Science Communicator at JILA (a world-leading physics research institute) and a science writer at The Debrief. Follow and connect with her on X or contact her via email at [email protected]

Putting people at the centre of decisions about patient data

Research, advocacy and communication to make the way patient data is used more trustworthy

• Learn about how patient data is used and why it's important
• Use our research and resources to explain patient data
• Find out how we work and ways to partner with us

We work with patient groups, charities, NHS organisations and policymakers to bring transparency, accountability and public involvement to the way patient data is used. Get in touch to partner with us.

Latest news

Understanding patient data writes to new health and science ministers, what might the new government mean for health data policy and practices, blog: collaborating well with patients when developing healthcare ai.

Masks Strongly Recommended but Not Required in Maryland, Starting Immediately

Due to the downward trend in respiratory viruses in Maryland, masking is no longer required but remains strongly recommended in Johns Hopkins Medicine clinical locations in Maryland. Read more .

• Vaccines  
• Masking Guidelines
• Visitor Guidelines  

Institutional Review Board

Participants' access to study records.

January 2005

Under federal and state law, patients generally have a right to a copy of their medical records, while participants in a research study generally do not have a right to a copy of their study records.   However, there are exceptions to this general no-access rule to study records.   Because the participant’s right to a copy of their study records often depends on whether the study record is also maintained as a medical record, it is important that researchers understand the rules which govern access to each type of record.

This guidance explains medical records and research records, the access rules for each and the four exceptions to the general no-access rule for study records.

  General Access Rules to Medical and Study Records

  Under Maryland law, in most circumstances, a health care provider must provide a copy of the medical record to the patient.   A medical record is one which contains information identifying the patient and relates to the health care of the person.   We interpret “health care” to mean care, treatment or procedures by a physician to address a physical or mental condition of a patient or recipient.    If the study record does not relate to “health care” for the individual, the study record would not be a medical record, and the patient would not have a right to access the study record.

  Under federal privacy regulations a patient has the right to a copy of their records that contain protected health information (“PHI”) if those records pertain to the provision, coordination or management of health care for that individual.   If their records contain PHI but do not pertain to health care, it appears that the individual would not have a right to a copy of those records.   This is so because the Office of Civil Rights, the federal agency that interprets these privacy regulations, has stated that if the study record and the medical record are kept separate, the study record would not be a record to which the individual would have a right of access under the privacy regulations.

Four Exceptions to the General No Access to Study Records Rule

1. Granting an Access Right in the Informed Consent

  A research study may involve clinical procedures such as an MRI, blood tests, hearing tests, etc. even though the purpose of the study is not to treat the individual for any health condition and does not involve treatment of any kind.   In these situations, the general rule of no right to access study records would be applicable.   However, the researcher may decide that he/she wants participants to have an enforceable right to receive specific test information.   (For pathology tests, only results from tests performed in CLIA approved labs may be disclosed unless the IRB specifically approves the disclosure of tests results for tests performed in non-CLIA approved labs.)   In such cases, the researcher may include a provision in the research study informed consent advising participants that they will be given the results of these tests.   The informed consent then becomes the vehicle that gives the participant a right to their test results.   Language in the consent form advising the participant that they will receive test results would not entitle the participant to a copy of the entire study record unless the consent form specifically states that the participant has a right to the entire record.

2. Voluntary Disclosure of Medical Information in the Study Record:

As was the case in Exception 1 discussed above, a study may involve clinical procedures such as an MRI, blood tests, hearing tests, etc. even though the purpose of the study is not to treat the individual for any health condition and does not involve treatment of any kind.   Again, in these situations, the general rule of no right to access study records would be applicable.   However, the researcher may decide, even if there is nothing in the study consent form that grants such a right, that he/she wants voluntarily to provide the test results to the participants.   Subject to any contractual obligations, discussed below, this voluntary disclosure may be accomplished in three ways.   First, the researcher may extract these results from the study record and give them to the participant.   Second, if the researcher is not the treating physician and the participant gives their consent, the researcher could extract these results from the study record and provide them to the participant’s treating physician.   Third, if the researcher is also the treating physician, the researcher may extract these results from the study record and enter the medical information in the participant’s medical record.

In all these voluntary situations, so long as only the medical information is supplied to the participant, and the study analysis and other data are not provided to the participant, this disclosure of medical information should not violate any confidentiality agreements with sponsors, collaborators, etc.   Of course, if the disclosure of medical information is expressly prohibited by such an agreement, the researcher could not make this voluntary disclosure.

3. Keeping a Combined Research/Medical Record:

Some studies, e.g., therapeutic trials, combine a research study with treatment (or possible treatment).   In such cases the study information and the medical information may be combined in one record.   If the research record and the medical record are kept as one, and are not separate, the right of a patient to access the medical record supersedes the general rule of no access by a participant to the study record.   In such a case, the patient/ participant would have a right of access to the medical/study record.

Under federal privacy regulations, it is possible in research studies related to treatment to notify the participant in the study informed consent that they will not have a right of access to the study record during the course of the study.   However, if a researcher is going to use this exception, two separate records, i.e., a study record and a medical record, must be kept (see the last exception below).

4. Entering Medical Information from the Study Record in the Medical Record:  

In studies which combine research and treatment (or possible treatment) if there is a desire not to provide study information to the participant, then the researcher must keep two separate records.   The study record would have both the medical information necessary for the trial and any study evaluation/observations/analysis, etc.   The medical record would have only the complete medical information and any relevant clinical treatment entries.   If this separate record keeping is done, then when patients ask for a copy of their medical records, only the medical record would need to be provided to the patient.   The participant would not have a right to access to the study record.

Asia & Oceania

Middle east & africa.

• United States
• Asia Pacific
• Australia & NZ
• Southeast Asia
• Czech Republic
• Deutschland
• España
• Switzerland
• United Kingdom

EMEA Thought Leadership

Developing IQVIA’s positions on key trends in the pharma and life sciences industries, with a focus on EMEA.

• Middle East and Africa
• United Kingdom (English)

NHS Solutions

Life sciences industry, advancing uk healthcare, pharmacy communications.

Utilising the extensive range of data assets exclusively available to IQVIA, combined with our market leading expertise, we are able to support the NHS meet the challenge of delivering high quality, efficient and cost effective healthcare by identifying actionable insight and realising the value in data.

• Insight & Feedback
• Costing Solutions
• Clinical Coding Analytics
• Benchmarking
• Pathway Transformation

We believe in the UK’s potential to become a global leader in the life sciences sector, and our investments announced in the 2018 Second Life Sciences Sector Deal further support the UK Government and the industry’s aim to make that potential a reality.

Life Sciences Industry Solutions

• Real-World Evidence
• Virtual Care
• IQVIA Medical Research Extraction Scheme
• Market Intelligence Solutions
• Market Access and Patient Activation

We are committed to providing solutions that inform and enable all stakeholders across UK healthcare, to innovate with confidence, maximise opportunities, and, ultimately, drive human health outcomes forward.

Advancing UK Healthcare Solutions

• Providing Integrated Solutions

IQVIA’s Pharmacy Communications offers pharma companies a simple and cost-effective channel to communicate product changes, updates and patient advice directly to patients via pharmacists.

BLOGS, WHITEPAPERS & CASE STUDIES

Explore our library of resources spanning the healthcare & life science ecosystem on the latest topics, trends, and market developments that matter to you.

Stay up to date on the latest news and press releases from IQVIA UK & Ireland.

"Our commitment to sustainability.

IQVIA Medical Research Data Privacy Notice

• Information for Members of the Public
• Medical Research Data

IQVIA Ltd is part of the IQVIA group of companies serving the combined industries of health information technologies and clinical research worldwide. We specialise in the provision of products and services including medical research and analytical solutions to healthcare, life sciences  and other organisations with an interest in medical and health research. In the UK, IQVIA has collected and supported the research use of non-identified patient data for over 20 years.

IQVIA Medical Research Data (IMRD) is non-identified electronic patient health record data supplied from UK General Practioner (GP) clinical systems via the IQVIA Medical Research Extraction Scheme and used for medical and public health research and treatment analysis.

Pursuant to data protection legislation, IQVIA is the data controller of the IQVIA Medical Research Data .

We are registered in England and Wales as: IQVIA Ltd, registration number: 03022416 and the registered address is 3 Forbury Place, 23 Forbury Road, Reading, United Kingdom, RG1 3JH.

What information is included in the IQVIA  Medical Research Data?

The  IQVIA Medical Research Data   is non-identified, consisting of data such as:

• Patient details: year of birth, sex, GP practice registration date, practice de-registration date
• Morbidity data: symptoms, diagnoses with dates, referrals to hospitals
• Prescribed medication: all prescriptions with date issued, drug name, formulation, strength, quantity, dosing instructions
• Immunisations: all in-practice immunisations
• Lab tests and and other health data: smoking status, height, weight, blood pressures, pregnancy, birth, death

The IQVIA Medical Research Data does NOT include any direct patient identifiers such as names, addresses, NHS numbers, or full dates of birth, nor any direct identifiers of GP practices participating in this data collection scheme.  Data can be extracted from a number of clinical systems via the schemes listed below:

IQVIA Medical Research Extraction Scheme (MRES)

IQVIA has set up the Medical Research Extraction Scheme (MRES) to collect the non-identified coded data extracted from GP practices in the UK that use EMIS Health and SystmOne clinical management systems. GP practices who have agreed to participate in MRES contribute data to IQVIA Medical Research Data which is used for medical and public health research and treatment analysis.

The purpose for which the data is processed and used

The IQVIA Medical Research Data may be processed and used for the purpose of medical and public health research in the following areas but not limited to:

• Epidemiology
• Pharmacoepidemiology
• Drug safety
• Public health research (including clinical audit)
• Drug Utilisation Studies (DUS)
• Post Authorisation Safety Studies (PASS)
• Outcomes research
• Health economics research
• Resource utilisation
• Morbidity & mortality research
• Risk management
• Treatment analysis and comparison

IQVIA processes the data on the basis of its legitimate interests in supporting medical and public health research and, because it is necessary for the above scientific research purposes, subject to appropriate safeguards.  Technical and organisational measures are in place to ensure only non-identified data is used.

IQVIA faciltates medical and public health research conducted by academic and commercial researchers in the disciplines listed above. IQVIA also uses aggregated forms of the data for treatment analysis, to provide insights into patient, disease and prescribing profiles.

Ethics approval

The use of IQVIA Medical Research Data for the purpose of medical and public health research and for supplying the data to external researchers for scientifically approved studies under Data Sharing Agreements has been approved by the NHS Health Research Authority (NHS Research Ethics Committee ref 23/EM/0151.

https://www.hra.nhs.uk/planning-and-improving-research/application-summaries/research-summaries/iqvia-medical-research-data/

A list of publications relating to studies that have used IQVIA Medical Research Data is available on the IQVIA Real-World Insights bibliography: http://www.rwebibliography.com/

Who is the IQVIA Medical Research Data made available to?

IQVIA carries out research on behalf of, or makes the data available to researchers from academic, public health, research establishment, charitable, commercial and regulatory bodies.

The IQVIA Medical Research Data may only be made available to external researchers for scientific studies for medical and public health research purposes under specific Data Sharing Agreement (DSA) terms that restrict use of the data.

Study Protocol Review

All proposed medical research studies using IQVIA Medical Research Data are subject to scientific review and all publications MUST be based on protocols approved by a recognised scientific review board.   The Committee that reviews IQVIA Medical Research Data protocols is the Scientific Review Committee (SRC).

All studies are required to show scientific merit, fulfil the research purpose outlined, and demonstrate potential benefit to health and social care.

Data Retention

IQVIA Medical Research Data consists of all available non-identified patient electronic health records. In order for studies using patient data to be scientifically sound, all information relating to a patient’s past medical events should be considered as this will influence their doctor’s decision and affects their current care. Longitudinal historical records on patient contact with their GP are important because the lead up to diagnosis of many conditions, particularly rare diseases, can be complex and lengthy. In addition, longitudinal data is increasingly valuable, for example, in studying chronic conditions, when a long history is required so that trends and signals can be detected and early risk prediction tools can be designed. Availability of sufficient data and research can drive health policy changes or adoption of new guidelines for healthcare providers, thus leading to improved outcomes for patients.

IQVIA retains data in accordance with UK Medical Research Council (MRC) guidelines which recommend basic research data and related material be retained for a minimum of 10 years after the study has been completed.  This is to support good research practice which requires research studies to be reproducible.

https://mrc.ukri.org/documents/pdf/retention-framework-for-research-data-and-records/

IQVIA has implemented a global Integrated Information Security Framework policy in order to govern and protect the confidentiality, integrity, and availability of data. In the UK, IQVIA Ltd is certified as being compliant with the ISO 27001 information security standard.

IQVIA has put in place appropriate technical and organisational security measures to ensure a level of security suitable to the risk and to prevent accidental or unlawful destruction, loss, alteration, unauthorised access to or disclosure of the data. Access to the data is strictly controlled and limited internally to authorised personnel. External use and access is governed by Data Sharing Agreements which are legally binding agreements detailing confidentiality and the terms on which the data is shared, used, stored and accessed. IQVIA stores the IQVIA Medical Research Data on secure servers located in the United Kingdom (UK) and the European Economic Area (EEA). If information is required to be transferred outside the UK or the EEA to a country that is not subject to an adequacy decision by the UK Information Commissioner or the EU Commission, data is adequately protected by UK Information Commissioner or EU Commission-approved standard contractual clauses.

Consent and patient rights

Data is collected in line with the requirements of applicable data protection law.

Information on the processing of patients´ data will be made available from the GP practices that are contributing data in accordance with applicable legislation, including the Data Protection Act 2008 and the UK General Data Protection Regulation (UK GDPR). The GP practices are also responsible for obtaining any necessary patient consents, if required by UK GDPR, for the processing of their personal data for the anticipated research purposes.

Each GP practice will additionally provide for the exercise of patients´ rights, including access, rectification, objection to processing and erasure, in accordance with UK GDPR requirements.  Patients have the right to opt out of their data being used for purposes beyond their individual care and treatment.  Patients are informed by means of posters in their GP practice that their data is collected for scientific research and they can opt-out at any time by contacting their GP practice. Alternatively patients can directly opt out of sharing their data via the National Data Opt-out Policy service. https://digital.nhs.uk/services/national-data-opt-out . Any patients that have opted-out from THIN or the IQVIA Medical Research Extraction Scheme will no longer be included in the data extracts from that point forward.

Patients have the right to request rectification of their personal data, or restriction of processing of personal data or to object to the processing of such personal data, as well as the right to data portability. This can only be done by the patients contacting their GP to exercise these rights.

Changes to this Notice

We may update this notice from time to time.  We encourage you to review this notice periodically to stay informed about how we are using and protecting IQVIA Medical Research Data.  Any changes to this notice take effect immediately after being posted on this webpage or otherwise provided by us.

Contact details

Questions and comments regarding this notice should be addressed to Louise Pinder at  IQVIA The Point, 37 North Wharf Road, Paddington, London, W2 1AF or emailed to [email protected] .

For this browsing session please remember my choice and don't ask again.

SGLT-2 inhibitors show potential in preventing dementia in diabetes patients

• Download PDF Copy

Sodium-glucose cotransporter-2 (SGLT-2) inhibitors used to treat type 2 diabetes might prevent dementia, providing greater benefits with longer treatment, suggests a large study from Korea published by  The BMJ  today. 

As this study was observational, the researchers note that the effect size could have been overestimated and say randomized controlled trials are now needed to confirm these findings.

According to the World Health Organization, the number of people with dementia globally is expected to reach 78 million by 2030 and type 2 diabetes is associated with a greater risk of developing dementia.

A recent study of over 65s with type 2 diabetes suggested a decreased risk of dementia associated with SGLT-2 inhibitors versus another type of diabetes drug, dipeptidyl peptidase-4 (DPP-4) inhibitors. But the effects on younger people and specific types of dementia (eg, Alzheimer's disease, vascular dementia) remains unclear.

To address this, researchers used the Korea National Health Insurance Service database to identify 110,885 pairs of adults with type 2 diabetes aged 40-69 years who were free of dementia and started taking either an SGLT-2 inhibitor or a DPP-4 inhibitor between 2013 and 2021. 

All participants (average age 62; 56% men) were matched by age, sex, use of the diabetes drug metformin, and baseline cardiovascular risk and were followed up for an average of 670 days to see who developed dementia.

Potentially influential factors including personal characteristics, income level, underlying risk factors for dementia, other conditions and related medicine use, were also taken into account.

Over the follow-up period, a total of 1,172 participants with newly diagnosed dementia were identified.

Dementia rates per 100 person years were 0.22 for those using SGLT-2 inhibitors and 0.35 for those using DPP-4 inhibitors, corresponding to a 35% reduced risk of dementia associated with use of SGLT-2 inhibitors compared with DPP-4 inhibitors.

The researchers also found a 39% reduced risk for Alzheimer's disease, and a 52% reduced risk for vascular dementia associated with SGLT-2 inhibitors compared with DPP-4 inhibitors.

Related Stories

• Promising new approach to treating type 2 diabetes
• Diabetes rates surge 18.6% in the U.S., highlighting racial and economic disparities
• Dementia Australia welcomes UK approval of lecanemab to tackle Alzheimer’s

What's more, the effect of SGLT-2 inhibitors seemed more pronounced with longer treatment duration. A 48% reduced risk of dementia was seen for more than two years of treatment versus a 43% reduced risk for two years or less.

This is an observational study so no firm conclusions can be drawn about cause and effect, and the authors note that details of health behaviors (eg, smoking and alcohol consumption) and duration of type 2 diabetes were not fully available.

However, they point out that this was a large study based on nationally representative data that included relatively younger people with type 2 diabetes , and results were highly consistent across subgroups. 

As such, they say SGLT-2 inhibitors might prevent dementia, providing greater benefits with longer treatment, and they call for randomized controlled trials to confirm these findings.

This study reports promising results that have important implications for clinical practice as well as from a public health perspective, say researchers from Taiwan in a linked editorial.

They agree that further trials are needed to confirm these findings, and suggest that studies are also needed "to explore the underlying mechanisms of any neuroprotective effects of SGLT-2 inhibitors."

As no cure currently exists for dementia and few effective treatment options are available, strategies that can potentially prevent onset are critically important, they write. 

Given the substantial socioeconomic and public health burdens associated with both dementia and type 2 diabetes, they also recommend that clinical guidelines and healthcare policies should be updated regularly to incorporate latest best evidence on the potential benefits of SGLT-2 inhibitors, including reduced dementia risk.

Shin, A., et al . (2024). Risk of dementia after initiation of sodium-glucose cotransporter-2 inhibitors versus dipeptidyl peptidase-4 inhibitors in adults aged 40-69 years with type 2 diabetes: population based cohort study.  BMJ . doi.org/10.1136/bmj-2024-079475 .

Posted in: Medical Research News | Medical Condition News | Pharmaceutical News

Tags: Alcohol , Alzheimer's Disease , Chemicals , Dementia , Diabetes , Glucose , Health Insurance , Healthcare , Hospital , Medicine , Metformin , Public Health , Research , Smoking , Type 2 Diabetes , Vascular

Suggested Reading

Cancel reply to comment

• Trending Stories
• Latest Interviews
• Top Health Articles

Global and Local Efforts to Take Action Against Hepatitis

Lindsey Hiebert and James Amugsi

In this interview, we explore global and local efforts to combat viral hepatitis with Lindsey Hiebert, Deputy Director of the Coalition for Global Hepatitis Elimination (CGHE), and James Amugsi, a Mandela Washington Fellow and Physician Assistant at Sandema Hospital in Ghana. Together, they provide valuable insights into the challenges, successes, and the importance of partnerships in the fight against hepatitis.

Addressing Important Cardiac Biology Questions with Shotgun Top-Down Proteomics

In this interview conducted at Pittcon 2024, we spoke to Professor John Yates about capturing cardiomyocyte cell-to-cell heterogeneity via shotgun top-down proteomics.

A Discussion with Hologic’s Tim Simpson on the Future of Cervical Cancer Screening

Tim Simpson

Hologic’s Tim Simpson Discusses the Future of Cervical Cancer Screening.

Latest News

Newsletters you may be interested in

Your AI Powered Scientific Assistant

Hi, I'm Azthena, you can trust me to find commercial scientific answers from News-Medical.net.

A few things you need to know before we start. Please read and accept to continue.

• Use of “Azthena” is subject to the terms and conditions of use as set out by OpenAI .
• Content provided on any AZoNetwork sites are subject to the site Terms & Conditions and Privacy Policy .
• Large Language Models can make mistakes. Consider checking important information.

Great. Ask your question.

Azthena may occasionally provide inaccurate responses. Read the full terms .

While we only use edited and approved content for Azthena answers, it may on occasions provide incorrect responses. Please confirm any data provided with the related suppliers or authors. We do not provide medical advice, if you search for medical information you must always consult a medical professional before acting on any information provided.

Your questions, but not your email details will be shared with OpenAI and retained for 30 days in accordance with their privacy principles.

Please do not ask questions that use sensitive or confidential information.

Read the full Terms & Conditions .

Provide Feedback

• Health Equity
• Managed Care
• Maternal & Early Childhood Health
• Rural Health
• Unwinding the PHE
• State Data Hub
• Research & Reports
• Comments on Federal Regulations
• Faculty and Staff
• LinkedIn LinkedIn

Biden Administration Proposes Rule To Ban Medical Debt From Credit Reporting

• share this post on Facebook
• share this post on Twitter
• share this post on LinkedIn
• share this post through Email

By Sheela Ranganathan and Maanasa Kona, Georgetown University Center on Health Insurance Reforms

Amidst the growing interest among policymakers to protect patients from medical debt and its negative downstream effects, in April 2023, the three credit reporting agencies (CRAs)—Equifax, Experian, and TransUnion— voluntarily agreed  to stop reporting any medical debt under $500. This April, the Consumer Financial Protection Bureau (CFPB)  found that , despite these changes, 15 million Americans still have $49 billion worth of medical bills on their credit reports. In particular, CFPB found that medical debt disproportionately burdens older Americans as well as low-income and rural communities.

In an effort to protect these patients, CFPB issued a  proposed rule  in June 2024 seeking to ban medical debt from certain credit reports. If finalized, the rule would also stop creditors from relying on medical bills for underwriting decisions and stop debt collectors from engaging in certain coercive collection practices.

In a  press release  accompanying the proposed rule, Vice President Kamala Harris and CFPB Director Rohit Chopra highlighted the importance of ensuring that getting sick or taking care of loved ones does not result in financial hardship. The proposed rule is aimed at protecting consumers from  repercussions of medical debt like  restricted access to loans and increased risk of bankruptcy, many of which disproportionately burden Black, Hispanic, and low-income communities.

The Impact Of Medical Debt On Credit Reports And Policy Responses To Date

Medical debt is a  leading cause of bankruptcy  in the United States and affects about  100 million Americans . Medical debt can impact patients’ well-being, making them  less likely  to seek future medical care and negatively affecting their financial stability. One of ways in which medical debt can negatively affect a patient’s financial health is through its impact on their credit report. Hospitals and debt collectors often provide information about medical debt to credit reporting agencies, which then include this information in the credit reports they provide to prospective lenders, employers, and landlords. This practice can make it much harder for patients to obtain  loans, employment, and housing .

At the federal level, the tax code defines the practice of non-profit hospitals reporting medical debt to credit reporting agencies as an “ extraordinary collections action .” Before taking this action, federal law requires non-profit hospitals to wait at least 120 days from the day of providing the first post-discharge bill, and to notify patients at least 30 days before reporting that debt to credit reporting agencies. This requirement only applies to non-profit hospitals and falls far short of actually prohibiting the practice entirely.

Further, at the federal level, the CFPB has additional oversight authority over credit reporting through the Fair Credit Reporting Act (FCRA). Under FCRA, CFPB oversees credit reporting agencies, entities that provide information to them (debt collectors and hospitals), and entities that use credit reports in their decision making (creditors, employers, landlords). In 2003, FCRA was amended to  prohibit creditors from using medical information  in their decision making, but  subsequent regulations  created an exemption that allows the consideration of financial medical information or information related to medical debt, a subcategory of medical information.

States have been active in this space as well. In the past year alone, a number of states have enacted legislation that will prevent medical debt from harming patient credit reports. For example,  Rhode Island  implemented a “belt-and-suspenders” approach, which (1) prohibits providers from reporting medical debt to credit reporting agencies, and (2) prohibits credit reporting agencies from retaining or reporting on any medical debt information.

New CFPB Rule Seeks To Remove Medical Debt Data From Certain Credit Reports

In the  preamble to the proposed rule , CFPB notes that medical debt has limited predictive value for credit underwriting purposes given the unique circumstances that cause an individual to go into medical debt. Further, CFPB finds that medical debt information can be riddled with inaccuracies. According to the agency, many industry participants have stopped or reduced their reliance on medical debt information, “casting doubt on its value.”

Given the limited utility of using medical debt to make credit decisions, CFPB’s proposed rule would amend Regulation V, which implements FCRA, to incorporate three main changes. First, it would remove the financial information exception that currently allows creditors to use medical information related to medical debt when making credit eligibility determinations. The preamble explains:

Medical information related to medical debt includes, for example, “[t]he dollar amount, repayment terms, repayment history, and similar information regarding medical debts to calculate, measure, or verify the repayment ability of the consumer, the use of proceeds, or the terms for granting credit” and “[t]he identity of creditors to whom outstanding medical debts are owed in connection with an application for credit, including but not limited to, a transaction involving the consolidation of medical debts.”

CFPB would apply this requirement to any medical debt owed directly to a health care provider, sold to a debt buyer, assigned to a third-party debt collector who has been assigned the debt by a health care provider, or that is the subject of a civil judgment related to a debt collection action.

Second, the proposed rule would prohibit consumer reporting agencies from including medical debt information in credit reports provided to creditors, when it believes that creditors are prohibited from considering it. CFPB states that it intends for this requirement “to operate alongside Federal and State-level efforts to increase consumer protections around medical debt consumer reporting.” While the proposed rule falls short of a full prohibition, it would significantly limit the appearance of medical debt on credit reports.

Lastly, the proposal would ban repossession of medical devices. For example,  CFPB provides  that lenders would be prohibited from “taking medical devices as collateral for a loan” and “repossessing medical devices, like wheelchairs or prosthetic limbs, if people are unable to repay the loan.” If finalized, the rule would be effective 60 days after publication in the Federal Register.

While this proposed regulation represents a significant step forward in protecting patients from the negative downstream effects of medical debt, there are some gaps in the proposed rule that are worth noting. First, the proposed rule only prohibits the inclusion of medical debt information in credit reports generated for creditors making lending decisions. It does not prohibit credit reporting agencies from including information about medical debt in credit reports issued to others who use credit report information, such as prospective employers or landlords. Second, the proposed rule’s protections would not extend to patients who pay for their medical bills through either general purpose or medical credit cards.

Recent Litigation Allays Concerns About CFPB’s Constitutionality

CFPB’s ability to issue rules like the proposed rule on medical debt hinges on its authority and funding to do so. In 2020 and again this term, the Supreme Court considered broad constitutional attacks seeking to stop CFPB from conducting its work, ultimately rejecting such claims and permitting the agency to continue to issue regulations and bring enforcement actions.

The first existential lawsuit threatening CFPB was decided in 2020. That case was brought by a law firm in California that was being investigated by CFPB for alleged violations of telemarketing laws. The law firm asserted that CFPB’s demand for certain documents in its investigation process was invalid because CFPB’s leadership structure was unconstitutional under separation of powers principles. In a  5-4 opinion , the Court held that the agency’s single-Director configuration was incompatible with the Constitution, especially because the Director was not removable at will by the President. However, finding that CFPB’s leadership structure provisions were severable from the rest of the statute granting CFPB its authority, the Court found that the agency could continue to exercise its authority under a Director that was removable at the President’s discretion.

Again in its most recent term, the Court considered whether the structure of CFPB was constitutional—this time, evaluating whether the agency’s funding mechanism—separate from the annual appropriations process by Congress, though consistent with the model used for the Federal Reserve and other financial regulators—violated the Appropriations Clause. In May 2024, in a  7-2 decision  written by Justice Clarence Thomas, the Court held that CFPB’s funding structure did not violate the Appropriations Clause because a valid appropriation only needed to identify a source of public funds and authorize the expenditure of those funds for designated purposes. In a  press release  following the decision, Director Rohit Chopra stated that the ruling “makes clear that the CFPB is here to stay,” noting that the agency would resume its enforcement actions and rulemakings that were on pause while the case was heard.

Whether the CFPB issues a final rule on medical debt may depend on the upcoming presidential election and potential shifts in policy that could result from a change in administration. Under the Trump administration, CFPB was less engaged in both rulemaking and enforcement, consistent with the administration’s overall deregulatory efforts. In referring to the CFPB’s strategic plan for 2018 to 2022, the agency’s acting director at the time  stated that the administration was “committed to fulfilling the Bureau’s statutory responsibilities, but go no further.” Further, even if the rule is finalized, it might have to face and survive legal challenges.

Given the repeated challenges to CFPB’s authority, the uncertainty around the upcoming election, and the high probability of litigation if the rule is finalized, further state action could ensure that at least some patients are protected from the impact of medical debt on their credit reports. Even if the rule is finalized as proposed and survives legal challenges, state action can address some key gaps in the rule. Notably, the rule does not limit the use of medical debt information in employment and tenant screening or protect patients who pay for medical care using general purpose or medical credit cards. State action prohibiting providers from supplying information about medical debt to credit reporting agencies in the first place, or prohibiting credit reporting agencies from including medical debt information on any credit report they generate, could significantly expand protections for patients.

Sheela Ranganathan, Maanasa Kona, “Biden Administration Proposes Rule to Ban Medical Debt from Credit Reporting,” Health Affairs Forefront, August 9th, 2024,  https://www.healthaffairs.org/content/forefront/biden-administration-proposes-rule-ban-medical-debt-credit-reporting . Copyright © 2024 Health Affairs by Project HOPE – The People-to-People Health Foundation, Inc.

Subscribe to Updates from our Team

Future of medicaid likely hinges on election but voters are not yet hearing about candidates’ views, back-to-school season is key to reconnecting kids to health coverage.

CCF.GEORGETOWN.EDU

Search Icon

Events See all →

Driskell and friends.

The Arthur Ross Gallery presents the work of artist, scholar, and curator David Driskell and explores his relationships with other artists. Friends include: Romare Bearden, Elizabeth Catlett, Jacob Lawrence, Keith Morrison, James Porter, and Hale Woodruff.

6:00 a.m. - 1:00 p.m.

Arthur Ross Gallery, 220 S. 34th St.

Sense of Place: Japan

’chocolate milk’.

Fagin Hall, 418 Curie Blvd.

Present Futures

The Forum at the Annenberg School for Communication, 3620 Walnut St.

Health Sciences

Questioning restrictions on physical activity for those at risk of preterm birth

Two undergraduates conducting research this summer gathering data from patients for a study by penn medicine physician beth leong pineles..

Restrictions on physical activity and even complete bedrest are frequently prescribed for pregnant patients by doctors or suggested by others to prevent preterm birth, even though both practices have been proven ineffective and in some cases even harmful.

Beth Leong Pineles , a maternal-medicine doctor at Pennsylvania Hospital and an assistant professor of obstetrics and gynecology at the Perelman School of Medicine , is conducting a study of activity restriction during pregnancy with the goal of “deimplementation” in the Penn Medicine system and ultimately nationwide. The long-term goal is to use exercise as a tool to reduce preterm birth and improve the overall health of pregnant women.

Penn undergraduates Ellie Mayers and Gladys Smith were involved in the first stage of the study's research with Pineles during the summer. The goal is to determine the prevalence of bedrest and activity restriction, starting with the patient population at Penn Med, through an in-person survey of pregnant patients.

Smith, a second-year from Williamsburg, Virginia, is a student in the College of Arts and Sciences who plans to go to medical school. Mayers, a third-year from Beachwood, Ohio, is in the School of Nursing with plans to become a midwife and is also pursuing a minor in the history of health and humanities, a collaboration between Nursing and the College.

The research opportunity, which provides a $5,000 award for each student, is through the Penn Undergraduate Research Mentoring Program , a 10-week summer program supported by the Center for Undergraduate Research and Fellowships (CURF).

In-person research

Mayers and Smith were assigned to work at Penn Maternal Fetal Medicine Washington Square and the Hospital of the University of Pennsylvania in Philadelphia. They read patient charts to determine who was considered at high risk for preterm labor, including those who previously delivered preterm, are pregnant with more than one baby, have certain conditions affecting their cervix, or who have had preterm contractions.

They then asked those patients to complete the survey. For the best response rate, face-to-face is better than phone calls, texts, or emails, Pineles says, noting that her goal is to get at least 300 survey responses. “It’s a better experience for the students. It’s more engaging. They like to talk to the patients and be involved in the clinic, and they get some exposure to medicine,” she says. “And they’re helping. They’re doing a great job getting a lot of survey responses.”

Mayers and Smith are planning to continue the research next semester and into the future, invited by Pineles to stay with the project. Pineles says she expects the entire deimplementation project to take about three more years.

“There’s a wealth of data that we can look at in a multitude of different ways,” Pineles says. “So, if Ellie and Gladys want to continue to be involved, that would be great. We can continue to work together, answer new questions, and publish more papers.”

Mayers says she’s seen a “different view of health care” doing research versus clinical work as a nursing student. “It’s been super interesting to see a snapshot of time when the ultimate goal of the research is to solve these problems,” informing actions “that nurses and doctors are going to implement,” she says. “And I love science. I love talking to people. I'm very social, and I just want to help people. I feel like that all culminated in my interest in nursing.”

Smith says asking the research questions and being in the clinic has steered her towards pursuing a “more social science-based way of thinking regarding medicine.” She says working with Pineles “has cemented my desire to go to medical school. Being in the clinic with her has made it seem more immediate and achievable.”

Understanding activity restrictions

The survey asks patients to describe their high-risk conditions, what advice they were given about physical activity during their pregnancy, and who gave that advice, among other questions about their health. “We don't know what women are being told. We don't know who's telling them. And I think it’s important to know that if we're going tackle the problem and understand it better,” Pineles says.

Physical activity restriction can include limits on walking, working, sexual activity, exercising. Bedrest can mean the patient doesn’t even get up to go to the bathroom. Some negative complications can include muscle atrophy, deep vein thrombosis, mental and emotional stress, boredom and loss of social connections, and potential financial losses.

Studies have shown that pregnant women who were given activity restrictions “actually had worse outcomes in their pregnancy and were more likely to deliver prematurely,” Pineles says. “We really don't see any benefit of activity restriction. We have several national guidelines about exercise and physical activity and pregnancy, and all recommend against bedrest and recommend against routine activity restriction.”

Pineles is deploying her study survey nationally “to get a national prevalence of activity restriction during pregnancy, which has not been done,” she says. They received the expected 1,500 responses on the national survey, distributed by a marketing service through social media, she says.

Mayers and Smith used the national survey data for the research posters they will present at the CURF Research Expo on Sept. 16, and for abstracts they each submitted to the Society of Maternal-Fetal Medicine national conference.

Interest in medical care

Mayers says she wanted to be a nurse as long as she can remember, coming from a long line of nurses, including her grandmother and great-grandmothers. Also, Mayers spent much of her early years in hospitals surrounded by medical professionals because her younger sister needed treatment as an infant. “Doctors are amazing, but it was always nurses that were with my family constantly and are the ones that I really have memories of,” Mayers says, adding that her sister will be a first-year at Penn this fall.

Mayers says she previously wasn’t interested in research, but at Penn she learned about various possibilities, including a Ph.D. in nursing. “I've been lucky in that the Nursing School has prepared me really well to do a lot of aspects of this research. I am very comfortable interacting with patients in a variety of ways,” Mayers says, noting that she was on an OB/GYN rotation last semester. “I was quick in understanding different conditions and in understanding how to read the chart and look at the overall patient profile.”

Specifically, Mayers says she is interested in reproductive health, both the clinical and political aspects. She’s the co-president of Penn’s chapter of Nurses for Sexual and Reproductive Health and also is a member of Penn Hillel and the Kite and Key Society . Ultimately Mayers says she wants a career as a midwife or a women's health nurse practitioner.

The research is interesting to Smith because she may want to pursue OB/GYN in her medical career. She once visited her mother’s friend when she was pregnant on bed rest, who lost her baby shortly thereafter. “That specific memory was jarring to me and is what made me particularly interested in this research opportunity,” Smith says, “because every mom is trying their hardest to maximize the chances that their baby will be healthy.”

She also wanted a summer experience that would allow her to directly interact with patients, seeking a position in which she could “understand what providers are doing, actually get to see the influence of the research being done.”

Smith also works at the Center for Autism Research at the Roberts Center for Pediatric Research at the Children’s Hospital of Philadelphia .

Smith says that conducting the patient survey was “extremely daunting” initially; she worried that the way she was speaking with the patients was inhibiting them from participating, and so she adjusted her script and her approach throughout the summer.

“Ellie is outspoken and very comfortable working with patients. Using her as a role model during the first few weeks was really helpful because I could model my attitude and my body language to be more like hers,” Smith says.

Pineles started her own research when she was an undergraduate at University of California, Irvine, so she wanted to provide the opportunity for Penn undergraduates. Mayers and Smith are the first undergraduate research assistants she has recruited since joining Penn Medicine two years ago from University of Texas Health Science Center in Houston. Pineles was at the University of Maryland for residency and the University of Southern California for her M.D. and Ph.D.

“The young students are so eager and interested, and it's a good way to get them involved and get started on something of their own,” Pineles says. “I always ask them to come up with their own project. What question can you ask with the data that we have that seems interesting? What do you want to know?”

Personal research interests

Mayers’ project is focused on who is giving the advice and how likely it is that patients will follow the advice depending on who is giving it. “Are we seeing a statistical difference in the wellbeing of the patient if it's a health care professional who prescribed it or if it's someone else?” Mayers says.

Smith’s project is looking into health equity, specifically regarding variance between different racial and ethnic groups’ agreement with and adherence to their prescribed activity recommendations. “My question is also social science-based,” Smith says. “I’m curious as to why these disparities might exist and how we might go about eliminating them.”

Both Mayers and Smith also have shadowed Pineles during her workday in the hospital, which has had an impact on them both. “I can’t imagine a more formative experience after my freshman year of college,” Smith says. “I've learned so much, especially shadowing and actually seeing patients in Labor and Delivery,” Smith says.

Mayers says that Pineles has been a “really excellent mentor” as an attending physician at Penn Med. “That's a connection that I hope I’m going to have for a really long time, and I hope to continue to build on,” Mayers says, “She’s someone that when I enter the workforce will look out for me and already is providing me with professional development opportunities.”

Move-In coordinators help ease transition to college

The power of protons

Penn Medicine has treated more than 10,000 cancer patients at three proton therapy centers across the region, including the largest and busiest center in the world—while also leading the way in research to expand the healing potential of these positive particles.

Campus & Community

To Penn’s Class of 2024: ‘The world needs you’

The University celebrated graduating students on Monday during the 268th Commencement.

Class of 2025 relishes time together at Hey Day

An iconic tradition at Penn, third-year students were promoted to senior status.

Arts, Humanities, & Social Sciences

Picturing artistic pursuits

Hundreds of undergraduates take classes in the fine arts each semester, among them painting and drawing, ceramics and sculpture, printmaking and animation, photography and videography. The courses, through the School of Arts & Sciences and the Stuart Weitzman School of Design, give students the opportunity to immerse themselves in an art form in a collaborative way.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings
• My Bibliography
• Collections
• Citation manager

Save citation to file

Email citation, add to collections.

• Create a new collection
• Add to an existing collection

Add to My Bibliography

Your saved search, create a file for external citation management software, your rss feed.

• Search in PubMed
• Search in NLM Catalog
• Add to Search

Individual patient data meta-analysis is needed in Chinese medical research

Affiliation.

• 1 Institute of Basic Research in Clinical Medicine, China Academy of Chinese Medical Sciences, Beijing, 100700, China.
• PMID: 25411018
• DOI: 10.1007/s11655-014-1852-0

Publication biases and collection limitations are the main disadvantages of a traditional meta-analysis based on aggregate patient data (APD) from published articles. Individual patient data (IPD) meta-analysis, as the gold standard of systematic review, is a possible alternative in this context. However, the publications relative to IPD meta-analyses are still rare compared with the traditional ones, especially in the research oriented to Chinese medicine (CM). In this article, the strengths and detailed functioning of IPD meta-analysis are described. Furthermore, the need for IPD meta-analysis to assess the treatments based on CM was also discussed. Compared with the traditional APD meta-analysis, the IPD meta-analysis might give a more accurate and unbiased assessment and is worth to be recommended to CM researchers.

PubMed Disclaimer

Similar articles

• The strengths and limitations of meta-analyses based on aggregate data. Lyman GH, Kuderer NM. Lyman GH, et al. BMC Med Res Methodol. 2005 Apr 25;5:14. doi: 10.1186/1471-2288-5-14. BMC Med Res Methodol. 2005. PMID: 15850485 Free PMC article.
• Evidence synthesis combining individual patient data and aggregate data: a systematic review identified current practice and possible methods. Riley RD, Simmonds MC, Look MP. Riley RD, et al. J Clin Epidemiol. 2007 May;60(5):431-9. doi: 10.1016/j.jclinepi.2006.09.009. Epub 2007 Feb 5. J Clin Epidemiol. 2007. PMID: 17419953 Review.
• Recommendations for individual participant data meta-analyses on work stressors and health outcomes: comments on IPD-Work Consortium papers. Choi BK, Schnall P, Landsbergis P, Dobson M, Ko S, Gómez-Ortiz V, Juárez-Garcia A, Baker D. Choi BK, et al. Scand J Work Environ Health. 2015 May 1;41(3):299-311. doi: 10.5271/sjweh.3484. Epub 2015 Feb 5. Scand J Work Environ Health. 2015. PMID: 25719554
• Meta-analysis using individual participant data from randomised trials: opportunities and limitations created by access to raw data. Rogozińska E, Marlin N, Thangaratinam S, Khan KS, Zamora J. Rogozińska E, et al. Evid Based Med. 2017 Oct;22(5):157-162. doi: 10.1136/ebmed-2017-110775. Epub 2017 Aug 17. Evid Based Med. 2017. PMID: 28818966
• Translational Metabolomics of Head Injury: Exploring Dysfunctional Cerebral Metabolism with Ex Vivo NMR Spectroscopy-Based Metabolite Quantification. Wolahan SM, Hirt D, Glenn TC. Wolahan SM, et al. In: Kobeissy FH, editor. Brain Neurotrauma: Molecular, Neuropsychological, and Rehabilitation Aspects. Boca Raton (FL): CRC Press/Taylor & Francis; 2015. Chapter 25. In: Kobeissy FH, editor. Brain Neurotrauma: Molecular, Neuropsychological, and Rehabilitation Aspects. Boca Raton (FL): CRC Press/Taylor & Francis; 2015. Chapter 25. PMID: 26269925 Free Books & Documents. Review.
• J Infect Dis. 2004 Jun 15;189(12):2154-66 - PubMed
• Gastroenterology. 2005 Jun;128(7):1838-44 - PubMed
• J Acquir Immune Defic Syndr. 2005 Aug 1;39(4):430-8 - PubMed
• Trials. 2011 May 13;12:122 - PubMed
• Arch Intern Med. 2012 Oct 22;172(19):1444-53 - PubMed

Publication types

• Search in MeSH

LinkOut - more resources

Full text sources, other literature sources.

• scite Smart Citations
• Citation Manager

NCBI Literature Resources

MeSH PMC Bookshelf Disclaimer

The PubMed wordmark and PubMed logo are registered trademarks of the U.S. Department of Health and Human Services (HHS). Unauthorized use of these marks is strictly prohibited.