Browse research papers in Health Informatics - Medicine Discover 12 academic papers with AI-powered analysis

The WHOOP Healthspan Feature

This whitepaper describes WHOOP's new Healthspan feature, which calculates "functional age" based on lifestyle data like sleep, activity, and fitness. While it aims to provide actionable health insights grounded in science, the study relies on observational data and makes some assumptions where scientific consensus is lacking. Furthermore, the data comparing WHOOP users to national averages might be biased, as WHOOP users tend to be more health-conscious.

★

☆

WHOOP_2025_White_Paper_Healthspan.pdf

Aug 30, 08:55 AM

Beyond the Leaderboard: Rethinking Medical Benchmarks for Large Language Models

This study introduces MedCheck, a framework with 46 criteria to assess the quality of medical benchmarks for large language models (LLMs). Analysis of 53 existing benchmarks revealed systemic issues including a disconnect from clinical practice, poor data quality control, and a lack of safety and fairness evaluations. The paper proposes MedCheck as a tool to guide the creation of more robust and clinically relevant benchmarks.

Multimodal Al predicts clinical outcomes of drug combinations from preclinical data

MADRIGAL uses multiple data sources to predict how well and how safely different drugs will work together. It predicts the effects before the drug combinations are tested on people and can be used to screen for adverse drug interactions and personalize cancer treatments. However, the predictions rely on existing data, which may be biased, and clinical trials are still necessary to confirm the results.

Towards physician-centered oversight of conversational diagnostic AI

In a simulated clinical setting with human oversight, a diagnostic AI (g-AMIE) demonstrated high performance in taking patient histories, generating SOAP notes, and proposing diagnoses and management plans. While the study setting was artificial and didn't fully replicate real-world clinical practice, g-AMIE performed favorably compared to human clinicians (nurse practitioners, physician assistants, and early-career physicians) under a similar constrained workflow. Notably, the AI consistently observed safety guardrails by deferring individualized medical advice to overseeing physicians.

★

☆

Towards physician-centered oversight of conversational diagnostic AI.pdf

Jul 31, 02:29 AM

AI-based Clinical Decision Support for Primary Care: A Real-World Study

This real-world study found that an AI-powered clinical decision support tool, AI Consult, led to a significant reduction in diagnostic and treatment errors made by clinicians in primary care clinics in Nairobi, Kenya. The tool improved clinician performance across various areas, including history taking, investigations, diagnosis, and treatment, with particularly strong effects observed after the introduction of active deployment strategies. While there was no statistically significant difference found in patient-reported outcomes, the study suggests the potential of LLM-based tools to improve the quality of care in real-world settings.

Performance of GPT-3.5 and GPT-4 on the Japanese Medical Licensing Examination: Comparison Study

GPT-4 achieved a passing score on the Japanese Medical Licensing Examination (JMLE), while GPT-3.5 did not. This highlights the significant improvement in GPT-4's ability to process complex medical information in a non-English language, surpassing GPT-3.5 in various question types and difficulty levels.

Utility of ChatGPT in Clinical Practice

This viewpoint explores the potential applications of ChatGPT in clinical practice, including clinical decision support, medical documentation, and question answering. It identifies several benefits such as improved efficiency and diagnostic accuracy, but also acknowledges potential risks related to data privacy, bias, and the need for expert oversight.

ChatGPT: Bullshit spewer or the end of traditional assessments in higher education?

This paper explores the implications of ChatGPT for higher education, focusing on its potential to disrupt traditional assessments. It discusses the challenges and opportunities presented by the technology, along with recommendations for students, teachers, and institutions. The authors advocate for a balanced approach, emphasizing the need for educators to adapt and innovate rather than simply banning the use of AI tools.

The state of artificial intelligence-based FDA-approved medical devices and algorithms: an online database

The authors identified 64 FDA-approved medical devices and algorithms using AI/ML, with only 29 explicitly mentioning AI/ML in their official announcements. Radiology and Cardiology are the leading specialties for these technologies, and the authors created an open-access database to track and categorize these devices.

★

☆

s41746-020-00324-0.pdf

Jul 14, 11:25 AM

Explainability for artificial intelligence in healthcare: a multidisciplinary perspective

Explainability in clinical decision support systems (CDSS) is crucial for maintaining patient autonomy, trust, and ethical medical practice. Opaque AI algorithms pose challenges to informed consent, shared decision-making, and the equitable distribution of healthcare resources, potentially hindering the responsible integration of AI into medicine.

★

☆

s12911-020-01332-6.pdf

Jul 14, 11:25 AM

How Does ChatGPT Perform on the United States Medical Licensing Examination (USMLE)? The Implications of Large Language Models for Medical Education and Knowledge Assessment

ChatGPT demonstrated performance equivalent to a passing score for a third-year medical student on USMLE Step 1 and Step 2 practice questions, exceeding the accuracy of other large language models like GPT-3 and InstructGPT. While demonstrating logical reasoning in all responses and using internal information effectively, the model's reliance on external information was stronger for correct answers, highlighting a potential link between knowledge access and performance.

Performance of ChatGPT on USMLE: Potential for Al-assisted medical education using large language models

ChatGPT performed at or near the passing threshold for all three USMLE exams without specialized training. It demonstrated high concordance in its explanations and offered insights potentially valuable for medical education, though the study acknowledges limitations and potential biases.

Health Informatics

Paperzilla Papers in Health Informatics

The WHOOP Healthspan Feature

Beyond the Leaderboard: Rethinking Medical Benchmarks for Large Language Models

Multimodal Al predicts clinical outcomes of drug combinations from preclinical data

Towards physician-centered oversight of conversational diagnostic AI

AI-based Clinical Decision Support for Primary Care: A Real-World Study

Performance of GPT-3.5 and GPT-4 on the Japanese Medical Licensing Examination: Comparison Study

Utility of ChatGPT in Clinical Practice

ChatGPT: Bullshit spewer or the end of traditional assessments in higher education?

The state of artificial intelligence-based FDA-approved medical devices and algorithms: an online database

Explainability for artificial intelligence in healthcare: a multidisciplinary perspective

How Does ChatGPT Perform on the United States Medical Licensing Examination (USMLE)? The Implications of Large Language Models for Medical Education and Knowledge Assessment

Performance of ChatGPT on USMLE: Potential for Al-assisted medical education using large language models