JMIR AI
An open access, peer-reviewed journal focused on research and applications for the health artificial intelligence (AI) community.
Editor-in-Chief:
Khaled El Emam, PhD, Canada Research Chair in Medical AI, University of Ottawa; Senior Scientist, Children’s Hospital of Eastern Ontario Research Institute: Professor, School of Epidemiology and Public Health, University of Ottawa, Canada Bradley Malin, PhD, Accenture Professor of Biomedical Informatics, Biostatistics, and Computer Science; Vice Chair for Research Affairs, Department of Biomedical Informatics: Affiliated Faculty, Center for Biomedical Ethics & Society, Vanderbilt University Medical Center, Nashville, Tennessee, USA
Impact Factor 6.1 More information about Impact Factor CiteScore 5 More information about CiteScore
Recent Articles

Continuous, noninvasive blood pressure (BP) monitoring is clinically important for managing hemodynamic instability in the intensive care unit (ICU). However, current ICU practice primarily relies on invasive arterial catheterization. While this is the clinical gold standard for continuous measurement, it carries risks such as infection and thrombosis and is typically limited to select critical care settings.

Health systems are increasingly deploying patient-facing conversational AI platforms, yet validated methodologies for safety evaluation under realistic high-risk conditions are lacking. Prior work has documented safety failures in clinical AI but not shown whether iterative, health system–embedded testing can produce measurable safety improvement.

People share health experiences online as stories, not isolated facts. Most computational systems reduce these narratives to keywords or claims, overlooking the agency, causality, and interpretation that determine whether health information is safe or potentially harmful to act on. Content can use medically accurate terminology while describing unsafe behavior, rendering it invisible to fact-checking and keyword approaches. This limitation, which we term , constrains the ability of computational systems to extract, represent, and reason over narratives as coherent wholes.

Pediatric influenza is an important public health concern. Chinese patent medicines are used in its treatment, but relevant clinical evidence is dispersed across guidelines, expert consensus statements, and clinical studies. The multidimensional knowledge required for syndrome-based prescribing complicates information retrieval and standardized clinical use.

Retrieval-augmented generation over clinical text depends on the embedding model chosen—a choice that interacts with documentation genre. Cross-site stability of embedding-model rankings has not been evaluated on real electronic health record (EHR) narratives with genre held constant; public benchmarks are predominantly single-source, and multi-institution evaluations compare models and query formulations within tasks rather than ranking stability across institutions.

Race, ethnicity, online health information sharing, and AI-mediated environments can intersect in ways that may importantly shape individuals’ willingness to share health data with researchers. In this Viewpoint, we closely examine this intersection by drawing parallels between historical and contemporary public health controversies and discussing recent empirical work centered on privacy-related experiences, data donation, and emerging evidence on AI-mediated digital disenfranchisement. We describe how AI-driven online engagement centered on race, ethnicity, and health can facilitate mutual aid and peer support while also posing privacy, reputational, algorithmic, and psychosocial risks when improperly mediated. These effects, alone or combined, can erode trust in digital health research and other engagement initiatives. Therefore, we propose 2 complementary sets of evaluation metrics in response: one focused on how race-related online experiences influence trust, consent behavior, and participation in data sharing and the other focused on race-conscious AI performance within the scope of fairness, representation, and downstream health outcomes. Together, these metrics position race and ethnicity as a key analytic dimension, rather than an optional covariate, to guide teams focused on understanding race as a key analytic dimension in AI-enabled, health-related social engagement online and its full role in and impact on health research.

Natural language processing and AI models are increasingly used to extract emotional signals from textual data; however, systematic validation of different computational paradigms in emotionally complex relational contexts remains limited. Such approaches may contribute to digital health research, particularly in the fields of digital mental health and couples therapy, where understanding emotional expressions within romantic relationships is especially relevant.

Clinical diagnosis of tic disorders (TDs) is often delayed by fluctuating symptoms and the subjective nature of assessments. While automated methods have been proposed to bridge this gap, traditional machine learning models frequently lack robustness in unconstrained environments and fail to capture the complex, multisite tics characteristic of severe TDs.

AI has the potential to improve clinical decision-making, efficiency, and patient outcomes, yet adoption in routine clinical practice has been slower than anticipated. While prior studies have identified factors that influence AI implementation, a comprehensive synthesis grounded in implementation science frameworks is lacking.

Repurposing approved drugs offers a time- and cost-effective alternative to traditional drug development. However, in silico prediction of repurposing candidates is challenging and requires the effective collaboration of specialists in various fields, including pharmacology, medicine, biology, and bioinformatics. Fragmented, specialized algorithms and tools often address only narrow aspects of the overall problem. Heterogeneous, unstructured data landscapes require the expertise of specialized users. Hence, diverse data sources might not integrate smoothly across workflows.








