JMIR AI

Preventable adverse drug reactions in geriatric patients are caused by overdosing, especially in cases of impaired renal function. Artificial intelligence (AI) chatbots are being discussed as tools to generate drug information, which can adjust drug dosing and prevent subsequent adverse drug reactions based on individualized patient data. However, the question arises as to the extent to which such AI chatbots can withstand scientific evaluation in this task.

Medical documentation imposes a significant administrative burden on physicians and reduces time for direct patient care. Artificial intelligence (AI)-assisted tools such as automatic speech recognition and large language models (LLMs) promise to reduce this burden, but their performance in multilingual environments has not been explored. Switzerland is highly multilingual, and non-native German-speaking physicians may find documentation particularly challenging.

Artificial intelligence (AI) is rapidly transforming health care and health research, offering new opportunities for improving efficiency, accessibility, and equity. However, the ethical, societal, and regulatory challenges of AI development and deployment are particularly pronounced in low- and middle-income countries (LMICs). While existing literature often emphasizes high-level ethical principles or technical frameworks, there is a notable gap in empirical, qualitative research that centers on human involvement and sociocultural dynamics throughout the AI lifecycle in LMIC contexts.

Suicide is a critical global public health issue, with millions experiencing suicidal ideation (SI) each year. Global estimates suggest that the lifetime prevalence of SI ranges between 9% and 12% worldwide, underscoring the scale of this public health concern. Online platforms, such as Reddit, provide spaces where individuals express suicidal thoughts and seek peer support. While prior computational research has leveraged machine learning and natural language analysis to detect SI, much of it lacks grounding in psychological theory, limiting interpretability and intervention design.

Artificial intelligence (AI) has emerged as a powerful tool for fostering positive behavior change and enhancing mental health support. However, the abrupt discontinuation or functional degradation of AI-driven interventions, particularly those featuring conversational agents, may trigger unintended psychological consequences. Therefore, we introduce and examine the concept of Artificial Intelligence Discontinuation Effects (AI-DICE), drawing parallels from abandonment-like experiences observed from problematic termination experiences with therapists. We propose a conceptual framework for AI-DICE mitigation that draws on evidence-based behavior change principles and explores clinical modalities that may inform mitigation toolkits (eg, Acceptance and Commitment Therapy, Cognitive Behavioral Therapy, Dialectical Behavior Therapy, and Motivational Interviewing). We also ground our approach in user experience research and community-engaged research. AI-DICE raises critical ethical challenges, including transparency, the ability to withdraw or adapt participation as the users’ knowledge of the intervention grows, and access to support postintervention. Prioritizing long-term continuation, or at least some form of ongoing access, over the best-planned complete discontinuation strategy may help ensure that AI-driven mental health solutions deliver lasting benefits rather than unintended harm. Finally, although not yet empirically established, incorporating discontinuation planning and AI-DICE mitigation from the outset may also improve intervention effectiveness by strengthening user autonomy, supporting skills transfer, and reducing dependence-related vulnerabilities.

Vancomycin is a widely used antibiotic that requires therapeutic drug monitoring (TDM) for optimized individual dosage. A deep learning–based model, pharmacokinetic recurrent neural network–1 compartment model (PKRNN-1CM), has shown the advantage of leveraging time-series electronic health record data for individualized estimation of vancomycin pharmacokinetic (PK) parameters. While 1-compartment PK models are commonly used because of their simplicity and previous trough-based clinical practices for dose adjustment, the pre–deep learning literature suggests the superiority of 2-compartment models.

Sex and age have long been known to affect lung function. Several biological variables and anatomical factors may contribute to sex- and age-related differences in pulmonary metrics.

The issue of population aging has emerged as a critical global challenge, driving the imperative for effective self-care and scalable health management solutions for older adults. Against the backdrop of the accelerating application of generative artificial intelligence (GenAI) in health care, a systematic evaluation is necessary to investigate how multimodal GenAI can support older adults in maintaining health and managing well-being.

In the emergency department, rapid prognostic assessment of patients with intracerebral hemorrhage (ICH) is essential for guiding early management decisions, particularly when stroke specialists are not immediately available. Recent advances in large language models have generated interest in their potential utility as clinical decision-support tools.

Large language models (LLMs) are being increasingly incorporated into clinical workflows due to their ability to synthesize medical knowledge and support diagnosis and treatment planning. However, their opaque internal decision-making processes limit trust, reliability, and safe clinical adoption. Mechanistic interpretability seeks to address this challenge by revealing how LLMs transform inputs into outputs. This paper explores the use of sparse autoencoders (SAEs) as a promising approach to improving mechanistic interpretability of LLMs in medicine. We discuss how SAE-based analyses can illuminate model reasoning, detect potential failure modes, and complement existing interpretability frameworks. Improving mechanistic interpretability through SAEs may be essential for safely deploying LLMs as trustworthy cognitive aids in clinical medicine.

Artificial intelligence (AI) technologies are increasingly being integrated into mental health settings to support tasks such as clinical documentation and decision-making. In parallel, AI-enabled deception detection, which leverages multimodal behavioral cues like facial expressions, vocal tone, and body movements, is an emerging research area. These technologies may hold relevance in mental health contexts, where deception can compromise treatment outcomes and therapeutic trust. However, most research on AI-based deception detection has focused on law enforcement domains, resulting in a limited understanding of its applicability to mental health. The ethical, relational, and practical implications of using such technologies in clinical settings remain underexplored.

Previous studies have highlighted the benefits of using artificial intelligence–powered remote patient monitoring (AI RPM) in detecting health changes across various disease cohorts. However, the use of AI RPM for identifying health deteriorations in patients following major surgical procedures remains underexplored.