Evaluation framework for conversational agents with artificial intelligence in health interventions: a systematic scoping review

Large language models for the mental health community: framework for translating code to care

Large language models (LLMs) offer promising applications in mental health care to address gaps in treatment and research. By leveraging clinical notes and transcripts as data, LLMs could improve diagnostics, monitoring, prevention, and treatment of mental health conditions. However, several challenges persist, including technical costs, literacy gaps, risk of biases, and inequalities in data representation. In this Viewpoint, we propose a sociocultural-technical approach to address these challenges. We highlight five key areas for development: (1) building a global clinical repository to support LLMs training and testing, (2) designing ethical usage settings, (3) refining diagnostic categories, (4) integrating cultural considerations during development and deployment, and (5) promoting digital inclusivity to ensure equitable access. We emphasise the need for developing representative datasets, interpretable clinical decision support systems, and new roles such as digital navigators. Only through collaborative efforts across all stakeholders, unified by a sociocultural-technical framework, can we clinically deploy LLMs while ensuring equitable access and mitigating risks.

AI for IMPACTS Framework for Evaluating the Long-Term Real-World Impacts of AI-Powered Clinician Tools: Systematic Review and Narrative Synthesis

BACKGROUND

Artificial intelligence (AI) has the potential to revolutionize health care by enhancing both clinical outcomes and operational efficiency. However, its clinical adoption has been slower than anticipated, largely due to the absence of comprehensive evaluation frameworks. Existing frameworks remain insufficient and tend to emphasize technical metrics such as accuracy and validation, while overlooking critical real-world factors such as clinical impact, integration, and economic sustainability. This narrow focus prevents AI tools from being effectively implemented, limiting their broader impact and long-term viability in clinical practice.

OBJECTIVE

This study aimed to create a framework for assessing AI in health care, extending beyond technical metrics to incorporate social and organizational dimensions. The framework was developed by systematically reviewing, analyzing, and synthesizing the evaluation criteria necessary for successful implementation, focusing on the long-term real-world impact of AI in clinical practice.

METHODS

A search was performed in July 2024 across the PubMed, Cochrane, Scopus, and IEEE Xplore databases to identify relevant studies published in English between January 2019 and mid-July 2024, yielding 3528 results, among which 44 studies met the inclusion criteria. The systematic review followed PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) guidelines and the Cochrane Handbook for Systematic Reviews. Data were analyzed using NVivo through thematic analysis and narrative synthesis to identify key emergent themes in the studies.

RESULTS

By synthesizing the included studies, we developed a framework that goes beyond the traditional focus on technical metrics or study-level methodologies. It integrates clinical context and real-world implementation factors, offering a more comprehensive approach to evaluating AI tools. With our focus on assessing the long-term real-world impact of AI technologies in health care, we named the framework AI for IMPACTS. The criteria are organized into seven key clusters, each corresponding to a letter in the acronym: (1) I-integration, interoperability, and workflow; (2) M-monitoring, governance, and accountability; (3) P-performance and quality metrics; (4) A-acceptability, trust, and training; (5) C-cost and economic evaluation; (6) T-technological safety and transparency; and (7) S-scalability and impact. These are further broken down into 28 specific subcriteria.

CONCLUSIONS

The AI for IMPACTS framework offers a holistic approach to evaluate the long-term real-world impact of AI tools in the heterogeneous and challenging health care context and lays the groundwork for further validation through expert consensus and testing of the framework in real-world health care settings. It is important to emphasize that multidisciplinary expertise is essential for assessment, yet many assessors lack the necessary training. In addition, traditional evaluation methods struggle to keep pace with AI's rapid development. To ensure successful AI integration, flexible, fast-tracked assessment processes and proper assessor training are needed to maintain rigorous standards while adapting to AI's dynamic evolution.

TRIAL REGISTRATION

reviewregistry1859; https://tinyurl.com/ysn2d7sh.

Encouraging arm use in stroke survivors: the impact of smart reminders during a home-based intervention

BACKGROUND

Upper limb impairment post-stroke often leads to a predominant use of the less affected arm and consequent learned disuse of the affected side, hindering upper limb outcome. Wearable sensors such as accelerometers, combined with smart reminders (i.e., based on the amount of arm activity), offer a potential approach to promote increased use of the affected arm to improve upper limb use during daily life. This study aimed to evaluate the efficacy of wrist vibratory reminders during a six-week home-based intervention in chronic stroke survivors.

METHODS

We evaluated the impact of the home-based intervention on the primary outcome, the Motor Activity Log-14 Item Version scores Amount of Use (MAL-14 AOU), and the secondary outcomes MAL-14 Quality of Movement (QOM) and sensor-derived activity metrics from the affected arm. A randomized controlled trial design was used for the study: the intervention group received personalized reminders based on individualized arm activity goals, while the control group did not receive any feedback. Mixed linear models assessed the influence of the group, week of the intervention period, and initial impairment level on MAL-14 and arm activity metrics.

RESULTS

Forty-two participants were enrolled in the study. Overall, participants exhibited modest but not clinically relevant increases in MAL-14 AOU (+ 0.2 points) and QOM (+ 0.2 points) after the intervention period, with no statistically significant differences between the intervention and control group. Feasibility challenges were noted, such as adherence to wearing the trackers and sensor data quality. However, in participants with sufficiently available sensor data (n = 23), the affected arm use extracted from the sensor data was significantly higher in the intervention group (p < 0.05). The initial impairment level strongly influenced affected arm use and both MAL-14 AOU and QOM (p < 0.01).

CONCLUSIONS

The study investigated the effectiveness of incorporating activity trackers with smart reminders to increase affected arm activity among stroke survivors during daily life. While the results regarding the increased arm use at home are promising, patient-reported outcomes remained below clinically meaningful thresholds and showed no group differences. Further, it is essential to acknowledge feasibility issues such as adherence to wearing the trackers during the intervention and missing sensor data.

TRIAL REGISTRATION

NCT03294187.

Nurses' perspectives on privacy and ethical concerns regarding artificial intelligence adoption in healthcare

Background

With the increasing integration of artificial intelligence (AI) technologies into healthcare systems, there is a growing emphasis on privacy and ethical considerations. Nurses, as frontline healthcare professionals, are pivotal in-patient care and offer valuable insights into the ethical implications of AI adoption.

Objectives

This study aimed to explore nurses' perspectives on privacy and ethical concerns associated with the implementation of AI in healthcare settings.

Methods

We employed Van Manen's hermeneutic phenomenology as the qualitative research approach. Data were collected through purposive sampling from the December 7, 2023 to the January 15, 2024, with interviews conducted in Bengali. Thematic analysis was utilized following member checking and an audit trail.

Results

Six themes emerged from the research findings: Ethical dimensions of AI integration, highlighting complexities in incorporating AI ethically; Privacy challenges in healthcare AI, revealing concerns about data security and confidentiality; Balancing innovation and ethical practice, indicating a need to reconcile technological advancements with ethical considerations; Human touch vs. technological progress, underscoring tensions between automation and personalized care; Patient-centered care in the AI era, emphasizing the importance of maintaining focus on patients amidst technological advancements; and Ethical preparedness and education, suggesting a need for enhanced training and education on ethical AI use in healthcare.

Conclusions

The findings underscore the importance of addressing privacy and ethical concerns in AI healthcare development. Nurses advocate for patient-centered approaches and collaborate with policymakers and tech developers to ensure responsible AI adoption. Further research is imperative for mitigating ethical challenges and promoting ethical AI in healthcare practice.

Redefining Virtual Assistants in Health Care: The Future With Large Language Models

This editorial explores the evolving and transformative role of large language models (LLMs) in enhancing the capabilities of virtual assistants (VAs) in the health care domain, highlighting recent research on the performance of VAs and LLMs in health care information sharing. Focusing on recent research, this editorial unveils the marked improvement in the accuracy and clinical relevance of responses from LLMs, such as GPT-4, compared to current VAs, especially in addressing complex health care inquiries, like those related to postpartum depression. The improved accuracy and clinical relevance with LLMs mark a paradigm shift in digital health tools and VAs. Furthermore, such LLM applications have the potential to dynamically adapt and be integrated into existing VA platforms, offering cost-effective, scalable, and inclusive solutions. These suggest a significant increase in the applicable range of VA applications, as well as the increased value, risk, and impact in health care, moving toward more personalized digital health ecosystems. However, alongside these advancements, it is necessary to develop and adhere to ethical guidelines, regulatory frameworks, governance principles, and privacy and safety measures. We need a robust interdisciplinary collaboration to navigate the complexities of safely and effectively integrating LLMs into health care applications, ensuring that these emerging technologies align with the diverse needs and ethical considerations of the health care domain.