John-Jose Nunez
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
John-Jose Nunez has a total of 149 co-authors. Here are the top co-authors by number of shared publications:
- Alan T. Bates
BC Cancer, Vancouver, BC, Canada
7 shared publications - Raymond W. Lam
University of British Columbia
7 shared publications Claudio N. Soares
Department of Psychiatry, Providence Care, Queen's University, Kingston, ON, Canada
4 shared publications- Daniel Hilbers
Faculty of Medicine, University of British Columbia, Vancouver, BC, Canada
4 shared publications - Kamyar Keramatian
University of British Columbia
4 shared publications Raymond T. Ng
University of British Columbia, Faculty of Science, Department of Computer Science, Vancouver, Canada
4 shared publications
Latest Publications
- 18 November 2026
- Published by American Society of Clinical Oncology (ASCO) in JCO Oncology Practice
Impact of Social Isolation and Psychological Distress on Overall Survival in Cancer: A Population-Based Cohort Study
Purpose: Social isolation is associated with increased mortality in cancer, but prior studies often examined single cancer types, approximated social isolation using composite indices, and rarely accounted for co-occurring psychological distress. We evaluated the independent associations of five social isolation domains and psychological distress with overall survival (OS) in a large population-based cancer cohort. Methods: We included 45,651 patients who completed the Psychosocial Screen for Cancer-Revised (PSSCAN-R) at their initial BC Cancer visit between April 18, 2011, and December 30, 2016. Five social isolation domains, depression symptoms, and anxiety symptoms were derived from the PSSCAN-R. OS was estimated using Kaplan-Meier methods, and independent associations were evaluated using multivariable Cox regression adjusting for age, sex, and metastatic status. Results: Among the cohort (mean age 64.9 years, 55.2% female), 32.5% screened positive for at least 1 social isolation domain, most commonly living alone (23.0%) or recent partner loss (10.9%). In unadjusted analyses, all five isolation domains, depression, and anxiety were associated with reduced OS (all P < .05). A composite social isolation score showed a graded, dose-response association, with the hazard of death increasing approximately 9% per additional domain. In adjusted models, lack of regular social contact (hazard ratio [HR], 1.29 [95% CI, 1.17 to 1.41]), living alone (HR, 1.18 [95% CI, 1.13 to 1.22]), depression (HR, 1.43 [95% CI, 1.37 to 1.48]), and anxiety (HR, 1.23 [95% CI, 1.19 to 1.28]) were each independently associated with mortality. Conclusion: Specific, individually screenable domains of social isolation and psychological distress were independently associated with mortality in cancer. Because these domains can be captured with brief self-report items within psychosocial screening, which is increasingly standard in oncology, social isolation screening represents a low-burden addition to risk stratification and supportive care prioritization.
- 18 November 2026
- Published by American Society of Clinical Oncology (ASCO) in JCO Oncology Practice
Impact of Social Isolation and Psychological Distress on Overall Survival in Cancer: A Population-Based Cohort Study
Purpose: Social isolation is associated with increased mortality in cancer, but prior studies often examined single cancer types, approximated social isolation using composite indices, and rarely accounted for co-occurring psychological distress. We evaluated the independent associations of five social isolation domains and psychological distress with overall survival (OS) in a large population-based cancer cohort. Methods: We included 45,651 patients who completed the Psychosocial Screen for Cancer-Revised (PSSCAN-R) at their initial BC Cancer visit between April 18, 2011, and December 30, 2016. Five social isolation domains, depression symptoms, and anxiety symptoms were derived from the PSSCAN-R. OS was estimated using Kaplan-Meier methods, and independent associations were evaluated using multivariable Cox regression adjusting for age, sex, and metastatic status. Results: Among the cohort (mean age 64.9 years, 55.2% female), 32.5% screened positive for at least 1 social isolation domain, most commonly living alone (23.0%) or recent partner loss (10.9%). In unadjusted analyses, all five isolation domains, depression, and anxiety were associated with reduced OS (all P < .05). A composite social isolation score showed a graded, dose-response association, with the hazard of death increasing approximately 9% per additional domain. In adjusted models, lack of regular social contact (hazard ratio [HR], 1.29 [95% CI, 1.17 to 1.41]), living alone (HR, 1.18 [95% CI, 1.13 to 1.22]), depression (HR, 1.43 [95% CI, 1.37 to 1.48]), and anxiety (HR, 1.23 [95% CI, 1.19 to 1.28]) were each independently associated with mortality. Conclusion: Specific, individually screenable domains of social isolation and psychological distress were independently associated with mortality in cancer. Because these domains can be captured with brief self-report items within psychosocial screening, which is increasingly standard in oncology, social isolation screening represents a low-burden addition to risk stratification and supportive care prioritization.
- Open Access
- 6 November 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Evaluating the effectiveness of the PolarUs app on quality of life outcomes in people with bipolar disorder: Randomized controlled trial protocol with a qualitative component (Preprint)
BACKGROUND Bipolar disorder (BD) is characterized by recurrent episodes of depression and mania or hypomania. Although BD can be associated with significant disability, mortality, and healthcare costs, people with BD can experience good health and quality of life (QoL). QoL outcomes are a priority for people with BD, and treatment guidelines recommend psychoeducation and self-management support. Despite this, access remains limited, particularly for diverse and equity-denied populations. Digital mental health interventions (DMHIs), including smartphone apps, offer a promising means to improve access to self-management support and QoL monitoring. However, most apps lack BD-specific tailoring, co-design with people with BD, and cultural and linguistic adaptations that support equitable access. The PolarUs mobile app is an evidence-informed DMHI co-designed with people with BD, grounded in the 17 domains of the Quality of Life in Bipolar Disorder (QoL.BD) scale, and supporting self-management through personalized strategies and resources, QoL assessments, and daily mood and sleep tracking. OBJECTIVE This randomized controlled trial with an embedded qualitative component aims to evaluate the efficacy of PolarUs in improving QoL among diverse people with BD in North America. We hypothesize that participants randomized to 12 weeks of PolarUs access (available in English, French, Spanish, and Mandarin Chinese) will show significant improvements in QoL at week 12 compared to an assessment-only waitlist control condition. The qualitative component will explore participants' experiences with and perceptions of PolarUs. METHODS Eligible participants with BD type I, II, cyclothymic disorder, other specified, or unspecified BD residing in North America will be randomized to either: (1) a 12-week period of access to the PolarUs app; or (2) a 12-week assessment-only waitlist period. The primary outcome measure is QoL, measured at week 12 using the Brief QoL.BD scale. Secondary outcomes include self-reported mood and psychological outcomes and exploratory healthcare resource use. Based on previous pilot data, a power analysis (80% power, two-sided α = 0.05) indicated a required sample size of 364, increased to 504 to account for 72% completion, to detect statistical superiority. Semi-structured interviews will be conducted with a purposively selected subsample of intervention group participants (n=30) to explore the acceptability and feasibility of the app, use of self-management strategies, and perceived impact. RESULTS This study was funded by the Canadian Institutes of Health Research in October 2024. Study recruitment is anticipated to launch in November 2026. CONCLUSIONS This study will test PolarUs, the first QoL-focused, BD-specific, and culturally adapted DMHI developed through co-design involving people with BD. Findings will provide evidence on the effect of offering PolarUs compared with delayed access as a low-cost, scalable DMHI for improving QoL in BD. If effective, PolarUs could offer an accessible, scalable tool to enhance self-management and improve outcomes for diverse populations with BD. CLINICALTRIAL The trial has been registered with ClinicalTrials.gov (registration number: NCT07890805).
- 6 November 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessEvaluating the effectiveness of the PolarUs app on quality of life outcomes in people with bipolar disorder: Randomized controlled trial protocol with a qualitative component (Preprint)
BACKGROUND Bipolar disorder (BD) is characterized by recurrent episodes of depression and mania or hypomania. Although BD can be associated with significant disability, mortality, and healthcare costs, people with BD can experience good health and quality of life (QoL). QoL outcomes are a priority for people with BD, and treatment guidelines recommend psychoeducation and self-management support. Despite this, access remains limited, particularly for diverse and equity-denied populations. Digital mental health interventions (DMHIs), including smartphone apps, offer a promising means to improve access to self-management support and QoL monitoring. However, most apps lack BD-specific tailoring, co-design with people with BD, and cultural and linguistic adaptations that support equitable access. The PolarUs mobile app is an evidence-informed DMHI co-designed with people with BD, grounded in the 17 domains of the Quality of Life in Bipolar Disorder (QoL.BD) scale, and supporting self-management through personalized strategies and resources, QoL assessments, and daily mood and sleep tracking. OBJECTIVE This randomized controlled trial with an embedded qualitative component aims to evaluate the efficacy of PolarUs in improving QoL among diverse people with BD in North America. We hypothesize that participants randomized to 12 weeks of PolarUs access (available in English, French, Spanish, and Mandarin Chinese) will show significant improvements in QoL at week 12 compared to an assessment-only waitlist control condition. The qualitative component will explore participants' experiences with and perceptions of PolarUs. METHODS Eligible participants with BD type I, II, cyclothymic disorder, other specified, or unspecified BD residing in North America will be randomized to either: (1) a 12-week period of access to the PolarUs app; or (2) a 12-week assessment-only waitlist period. The primary outcome measure is QoL, measured at week 12 using the Brief QoL.BD scale. Secondary outcomes include self-reported mood and psychological outcomes and exploratory healthcare resource use. Based on previous pilot data, a power analysis (80% power, two-sided α = 0.05) indicated a required sample size of 364, increased to 504 to account for 72% completion, to detect statistical superiority. Semi-structured interviews will be conducted with a purposively selected subsample of intervention group participants (n=30) to explore the acceptability and feasibility of the app, use of self-management strategies, and perceived impact. RESULTS This study was funded by the Canadian Institutes of Health Research in October 2024. Study recruitment is anticipated to launch in November 2026. CONCLUSIONS This study will test PolarUs, the first QoL-focused, BD-specific, and culturally adapted DMHI developed through co-design involving people with BD. Findings will provide evidence on the effect of offering PolarUs compared with delayed access as a low-cost, scalable DMHI for improving QoL in BD. If effective, PolarUs could offer an accessible, scalable tool to enhance self-management and improve outcomes for diverse populations with BD. CLINICALTRIAL The trial has been registered with ClinicalTrials.gov (registration number: NCT07890805).
- 27 October 2026
- Published by SAGE Publications in The Canadian Journal of Psychiatry
From Target to Trial: Artificial Intelligence in Psychiatric Drug Development
Plain Language Summary: How Artificial Intelligence Could Improve Psychiatric Drug Development Developing new medicines for mental health conditions is difficult, expensive, and often unsuccessful. One reason is that people with the same diagnosis can differ greatly in their underlying biology and in how they respond to treatment. Clinical trials also rely heavily on symptom ratings and can be affected by placebo responses. This editorial explains how artificial intelligence (AI) could help at several stages of psychiatric drug development. AI can help researchers identify promising drug targets, design new treatment molecules, and use biological, brain-imaging, and clinical data to predict treatment response and improve clinical trials. Emerging approaches include digital twins, synthetic control groups, and digital measures of symptoms and behaviour. However, these technologies are still developing and should not be viewed as a solution to all the challenges of psychiatric drug development. Wider use will require representative datasets, strong data governance, transparent and interpretable models, rigorous clinical validation, and appropriate regulation.
- 27 October 2026
- Published by SAGE Publications in The Canadian Journal of Psychiatry
From Target to Trial: Artificial Intelligence in Psychiatric Drug Development
Plain Language Summary: How Artificial Intelligence Could Improve Psychiatric Drug Development Developing new medicines for mental health conditions is difficult, expensive, and often unsuccessful. One reason is that people with the same diagnosis can differ greatly in their underlying biology and in how they respond to treatment. Clinical trials also rely heavily on symptom ratings and can be affected by placebo responses. This editorial explains how artificial intelligence (AI) could help at several stages of psychiatric drug development. AI can help researchers identify promising drug targets, design new treatment molecules, and use biological, brain-imaging, and clinical data to predict treatment response and improve clinical trials. Emerging approaches include digital twins, synthetic control groups, and digital measures of symptoms and behaviour. However, these technologies are still developing and should not be viewed as a solution to all the challenges of psychiatric drug development. Wider use will require representative datasets, strong data governance, transparent and interpretable models, rigorous clinical validation, and appropriate regulation.
- Open Access
- 20 September 2026
- Published by MDPI AG in Preprints.org
Reproducibility Challenges, Solutions, and Future Directions in Recommender Systems: A Survey
Reproducibility underpins scientific progress, yet in recommendation systems (RS) research, it remains one of the most persistent and underappreciated problems. Algorithms have advanced rapidly, but reported performance gains are often fragile and difficult to verify independently. To address this gap, using the PRISMA framework, we systematically review peer-reviewed studies published between 2013, when reproducibility began receiving attention in the RS community, and 2026. We analyze all identified papers and introduce a six-category hierarchical taxonomy of reproducibility challenges in RS. Building on this, we synthesize the solutions proposed in prior work into a structured framework that maps each category of solutions to the challenges it addresses. Our review shows that reproducibility in RS is a property of the entire experimental workflow, not simply the availability of code and data. To this end, we propose an end-to-end Reproducibility Pipeline, spanning data construction, splitting, candidate generation, configuration, training or execution, evaluation, and reporting. Recent advances in LLM-based and agentic recommendation systems introduce new reproducibility risks, including prompt sensitivity, nondeterministic inference, proprietary APIs, multi-step reasoning, tool use, and external dependencies. Finally, we outline future directions for reproducibility in RS. Relevant materials are available at Github.
- 20 September 2026
- Published by MDPI AG in Preprints.org
Open AccessReproducibility Challenges, Solutions, and Future Directions in Recommender Systems: A Survey
Reproducibility underpins scientific progress, yet in recommendation systems (RS) research, it remains one of the most persistent and underappreciated problems. Algorithms have advanced rapidly, but reported performance gains are often fragile and difficult to verify independently. To address this gap, using the PRISMA framework, we systematically review peer-reviewed studies published between 2013, when reproducibility began receiving attention in the RS community, and 2026. We analyze all identified papers and introduce a six-category hierarchical taxonomy of reproducibility challenges in RS. Building on this, we synthesize the solutions proposed in prior work into a structured framework that maps each category of solutions to the challenges it addresses. Our review shows that reproducibility in RS is a property of the entire experimental workflow, not simply the availability of code and data. To this end, we propose an end-to-end Reproducibility Pipeline, spanning data construction, splitting, candidate generation, configuration, training or execution, evaluation, and reporting. Recent advances in LLM-based and agentic recommendation systems introduce new reproducibility risks, including prompt sensitivity, nondeterministic inference, proprietary APIs, multi-step reasoning, tool use, and external dependencies. Finally, we outline future directions for reproducibility in RS. Relevant materials are available at Github.
- Open Access
- 16 September 2026
- Published by JMIR Publications Inc. in JMIR Medical Education
- Vol. 12, e108766
The AWARE (AI Use, Why, Attachment, Reality and Risk, and Effect on Functioning) Framework for Assessing Patients’ Use of AI in Mental Health Care
Conversational AI has rapidly become integrated into everyday life, with many individuals using large language model–based systems, such as ChatGPT, Gemini, and Claude, for education, productivity, health information, decision-making, emotional support, and companionship. As patients increasingly incorporate conversational AI into their cognitive, emotional, and social lives, these interactions may influence coping strategies, treatment engagement, clinical decision-making, and mental health outcomes. However, although medical education has increasingly emphasized AI literacy and the responsible use of generative AI by clinicians, comparatively little attention has been given to preparing health care professionals to assess patients’ use of conversational AI during routine clinical encounters. This Viewpoint introduces the AWARE (AI use, why, attachment, reality and risk, and effect on functioning) framework, a practical educational framework designed to help mental health professionals systematically assess patients’ use of conversational AI. Rather than functioning as a diagnostic instrument or psychometric scale, AWARE provides a structured approach to psychiatric interviewing across 5 clinically relevant domains: AI use, why, attachment, reality and risk, and effect on functioning. Together, these domains guide clinicians in exploring patterns of AI use, the motivations underlying engagement, the emotional significance of patient-AI interactions, potential influences on reality testing and clinical risk, and the overall impact of AI on psychological well-being, daily functioning, relationships, and recovery. The framework is intended to complement existing psychiatric interviewing practices by supporting comprehensive history taking, clinical reasoning, risk assessment, documentation, and learner education. We discuss the rationale for routinely asking patients about AI use, review emerging evidence regarding both the potential benefits and risks of conversational AI, and describe how AWARE may be incorporated into undergraduate, postgraduate, and continuing professional education through simulation, objective structured clinical examinations, workplace-based assessment, and clinical supervision. We also outline priorities for future research, including validation, implementation, educational evaluation, and cross-cultural adaptation. As conversational AI becomes increasingly embedded within patients’ daily lives, clinicians require practical approaches to understanding its role in mental health and health care. The AWARE framework offers a structured educational starting point for integrating assessment of patient-AI interactions into routine psychiatric interviewing while supporting patient-centered, evidence-informed clinical practice and health professions education.
- 16 September 2026
- Published by JMIR Publications Inc. in JMIR Medical Education
- Vol. 12, e108766
Open AccessThe AWARE (AI Use, Why, Attachment, Reality and Risk, and Effect on Functioning) Framework for Assessing Patients’ Use of AI in Mental Health Care
Conversational AI has rapidly become integrated into everyday life, with many individuals using large language model–based systems, such as ChatGPT, Gemini, and Claude, for education, productivity, health information, decision-making, emotional support, and companionship. As patients increasingly incorporate conversational AI into their cognitive, emotional, and social lives, these interactions may influence coping strategies, treatment engagement, clinical decision-making, and mental health outcomes. However, although medical education has increasingly emphasized AI literacy and the responsible use of generative AI by clinicians, comparatively little attention has been given to preparing health care professionals to assess patients’ use of conversational AI during routine clinical encounters. This Viewpoint introduces the AWARE (AI use, why, attachment, reality and risk, and effect on functioning) framework, a practical educational framework designed to help mental health professionals systematically assess patients’ use of conversational AI. Rather than functioning as a diagnostic instrument or psychometric scale, AWARE provides a structured approach to psychiatric interviewing across 5 clinically relevant domains: AI use, why, attachment, reality and risk, and effect on functioning. Together, these domains guide clinicians in exploring patterns of AI use, the motivations underlying engagement, the emotional significance of patient-AI interactions, potential influences on reality testing and clinical risk, and the overall impact of AI on psychological well-being, daily functioning, relationships, and recovery. The framework is intended to complement existing psychiatric interviewing practices by supporting comprehensive history taking, clinical reasoning, risk assessment, documentation, and learner education. We discuss the rationale for routinely asking patients about AI use, review emerging evidence regarding both the potential benefits and risks of conversational AI, and describe how AWARE may be incorporated into undergraduate, postgraduate, and continuing professional education through simulation, objective structured clinical examinations, workplace-based assessment, and clinical supervision. We also outline priorities for future research, including validation, implementation, educational evaluation, and cross-cultural adaptation. As conversational AI becomes increasingly embedded within patients’ daily lives, clinicians require practical approaches to understanding its role in mental health and health care. The AWARE framework offers a structured educational starting point for integrating assessment of patient-AI interactions into routine psychiatric interviewing while supporting patient-centered, evidence-informed clinical practice and health professions education.
- Open Access
Limits of Single-Pass Retrieval-Augmented Generation for AI-Powered Cancer Care Navigation: A Comparison of Retrieval Strategies
Background: People affected by cancer often face difficulty finding relevant clinical, psychological, and practical support services. AI-powered navigation assistants may improve access to these resources, but their retrieval performance must be reliable. Objective: To develop a single-pass retrieval-augmented generation assistant for cancer-care navigation and compare the retrieval strategies, including their robustness to reworded questions. Methods: We created a database of 853 cancer-support resources reviewed by librarians, clinicians, researchers, and patient partners. We evaluated the system using 100 questions derived from questions submitted by patients. We compared keyword-based, semantic, and hybrid retrieval using Precision@K, Hit@K, and nDCG@K. The best-performing configuration was then tested using semantically equivalent rewordings of the original questions. Results: Keyword-based retrieval performed poorly, achieving a P@1 of 25.0% and Hit@5 of 43.0%. Semantic retrieval improved these results to 58.0% and 86.0%, respectively. The best hybrid configuration achieved a P@1 of 64.0%, Hit@5 of 90.0%, and nDCG@5 of 51.0%. Performance remained relatively stable when the questions were reworded, with a P@1 of 61.0%, Hit@5 of 88.0%, and nDCG@5 of 46.1%. Conclusions: Hybrid retrieval performed best and remained relatively stable when questions were reworded. However, its limited ability to rank a relevant resource first highlights the limitations of single-pass retrieval for patient-facing cancer navigation. Future work will explore metadata filtering and a multi-agent architecture to improve retrieval reliability.
Open AccessLimits of Single-Pass Retrieval-Augmented Generation for AI-Powered Cancer Care Navigation: A Comparison of Retrieval Strategies
Background: People affected by cancer often face difficulty finding relevant clinical, psychological, and practical support services. AI-powered navigation assistants may improve access to these resources, but their retrieval performance must be reliable. Objective: To develop a single-pass retrieval-augmented generation assistant for cancer-care navigation and compare the retrieval strategies, including their robustness to reworded questions. Methods: We created a database of 853 cancer-support resources reviewed by librarians, clinicians, researchers, and patient partners. We evaluated the system using 100 questions derived from questions submitted by patients. We compared keyword-based, semantic, and hybrid retrieval using Precision@K, Hit@K, and nDCG@K. The best-performing configuration was then tested using semantically equivalent rewordings of the original questions. Results: Keyword-based retrieval performed poorly, achieving a P@1 of 25.0% and Hit@5 of 43.0%. Semantic retrieval improved these results to 58.0% and 86.0%, respectively. The best hybrid configuration achieved a P@1 of 64.0%, Hit@5 of 90.0%, and nDCG@5 of 51.0%. Performance remained relatively stable when the questions were reworded, with a P@1 of 61.0%, Hit@5 of 88.0%, and nDCG@5 of 46.1%. Conclusions: Hybrid retrieval performed best and remained relatively stable when questions were reworded. However, its limited ability to rank a relevant resource first highlights the limitations of single-pass retrieval for patient-facing cancer navigation. Future work will explore metadata filtering and a multi-agent architecture to improve retrieval reliability.
- Open Access
- 14 August 2026
- Published by Springer Nature in Implementation Science Communications
- Vol. 7 (1)
Towards clinical implementation of artificial intelligence in cancer care: concept mapping analysis of provincial workshop findings
Background: Artificial intelligence (AI) has rapidly garnered interest in healthcare. Cancer care’s multidisciplinary nature and high coordination demands are well positioned to benefit from AI. While attitudes toward implementation of AI in medicine have been explored generally, literature remains scarce with specific regards to AI in cancer care. This study sought to understand perspectives of both patients and professionals in guiding responsible, effective implementation of evidence-based AI in cancer care. Methods: We conducted a workshop at a provincial Cancer Summit (Vancouver, Canada). Discussions addressed concerns, benefits, and priorities for AI in cancer care. Responses from 48 workshop participants underwent structured conceptualization by concept mapping. Sorting and rating of the resulting statements were performed by a purposively sampled 13-member expert panel. Multidimensional scaling, hierarchical cluster and subcluster analysis produced visual and quantitative maps of findings. Results: A total of 265 statements on perceived benefits, concerns, and priorities related to the implementation of AI in cancer care were generated; a deduplicated and consolidated final set of 100 statements underwent concept mapping. Two main clusters identified pertained to “Challenges and Safeguards for AI Implementation " (Cluster 1) and “Clinical Benefits and Efficiency Gains” (Cluster 2). Subcluster analysis distinguished 8 thematic subclusters (4 per cluster). Mean importance and feasibility ratings were higher for Cluster 2, with large effect sizes for both importance (Cohen’s d = 0.94) and feasibility (Cohen’s d = 1.59). Ratings by clinical and nonclinical professionals were similar across comparisons except for Cluster 2 feasibility, rated higher by clinical participants ( P = 0.029, Hedges’ g = 0.465). Further go-zone analysis classified statements according to their relative superiority/inferiority in importance and feasibility from overall average. Conclusions: Expert panel ratings were higher for statements describing clinical benefits and efficiency gains than for those describing challenges and safeguards for AI implementation in cancer care. Concept mapping analysis distinguished between workflow-aligned AI applications, perceived as ready for implementation, and system-level governance requirements requiring longer-term investment. Present findings offer initial, context-specific structuring of stakeholder perspectives that may inform the prioritization and sequencing of AI implementation efforts in cancer care, providing a foundation for validation in other settings.
- 14 August 2026
- Published by Springer Nature in Implementation Science Communications
- Vol. 7 (1)
Open AccessTowards clinical implementation of artificial intelligence in cancer care: concept mapping analysis of provincial workshop findings
Background: Artificial intelligence (AI) has rapidly garnered interest in healthcare. Cancer care’s multidisciplinary nature and high coordination demands are well positioned to benefit from AI. While attitudes toward implementation of AI in medicine have been explored generally, literature remains scarce with specific regards to AI in cancer care. This study sought to understand perspectives of both patients and professionals in guiding responsible, effective implementation of evidence-based AI in cancer care. Methods: We conducted a workshop at a provincial Cancer Summit (Vancouver, Canada). Discussions addressed concerns, benefits, and priorities for AI in cancer care. Responses from 48 workshop participants underwent structured conceptualization by concept mapping. Sorting and rating of the resulting statements were performed by a purposively sampled 13-member expert panel. Multidimensional scaling, hierarchical cluster and subcluster analysis produced visual and quantitative maps of findings. Results: A total of 265 statements on perceived benefits, concerns, and priorities related to the implementation of AI in cancer care were generated; a deduplicated and consolidated final set of 100 statements underwent concept mapping. Two main clusters identified pertained to “Challenges and Safeguards for AI Implementation " (Cluster 1) and “Clinical Benefits and Efficiency Gains” (Cluster 2). Subcluster analysis distinguished 8 thematic subclusters (4 per cluster). Mean importance and feasibility ratings were higher for Cluster 2, with large effect sizes for both importance (Cohen’s d = 0.94) and feasibility (Cohen’s d = 1.59). Ratings by clinical and nonclinical professionals were similar across comparisons except for Cluster 2 feasibility, rated higher by clinical participants ( P = 0.029, Hedges’ g = 0.465). Further go-zone analysis classified statements according to their relative superiority/inferiority in importance and feasibility from overall average. Conclusions: Expert panel ratings were higher for statements describing clinical benefits and efficiency gains than for those describing challenges and safeguards for AI implementation in cancer care. Concept mapping analysis distinguished between workflow-aligned AI applications, perceived as ready for implementation, and system-level governance requirements requiring longer-term investment. Present findings offer initial, context-specific structuring of stakeholder perspectives that may inform the prioritization and sequencing of AI implementation efforts in cancer care, providing a foundation for validation in other settings.
- Open Access
- 4 August 2026
- Published by JMIR Publications Inc. in JMIR Preprints
The AWARE (AI Use, Why, Attachment, Reality and Risk, and Effect on Functioning) Framework for Assessing Patients’ Use of AI in Mental Health Care (Preprint)
UNSTRUCTURED Conversational AI has rapidly become integrated into everyday life, with many individuals using large language model–based systems, such as ChatGPT, Gemini, and Claude, for education, productivity, health information, decision-making, emotional support, and companionship. As patients increasingly incorporate conversational AI into their cognitive, emotional, and social lives, these interactions may influence coping strategies, treatment engagement, clinical decision-making, and mental health outcomes. However, although medical education has increasingly emphasized AI literacy and the responsible use of generative AI by clinicians, comparatively little attention has been given to preparing health care professionals to assess patients’ use of conversational AI during routine clinical encounters. This Viewpoint introduces the AWARE (AI use, why, attachment, reality and risk, and effect on functioning) framework, a practical educational framework designed to help mental health professionals systematically assess patients’ use of conversational AI. Rather than functioning as a diagnostic instrument or psychometric scale, AWARE provides a structured approach to psychiatric interviewing across 5 clinically relevant domains: AI use, why, attachment, reality and risk, and effect on functioning. Together, these domains guide clinicians in exploring patterns of AI use, the motivations underlying engagement, the emotional significance of patient-AI interactions, potential influences on reality testing and clinical risk, and the overall impact of AI on psychological well-being, daily functioning, relationships, and recovery. The framework is intended to complement existing psychiatric interviewing practices by supporting comprehensive history taking, clinical reasoning, risk assessment, documentation, and learner education. We discuss the rationale for routinely asking patients about AI use, review emerging evidence regarding both the potential benefits and risks of conversational AI, and describe how AWARE may be incorporated into undergraduate, postgraduate, and continuing professional education through simulation, objective structured clinical examinations, workplace-based assessment, and clinical supervision. We also outline priorities for future research, including validation, implementation, educational evaluation, and cross-cultural adaptation. As conversational AI becomes increasingly embedded within patients’ daily lives, clinicians require practical approaches to understanding its role in mental health and health care. The AWARE framework offers a structured educational starting point for integrating assessment of patient-AI interactions into routine psychiatric interviewing while supporting patient-centered, evidence-informed clinical practice and health professions education.
- 4 August 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessThe AWARE (AI Use, Why, Attachment, Reality and Risk, and Effect on Functioning) Framework for Assessing Patients’ Use of AI in Mental Health Care (Preprint)
UNSTRUCTURED Conversational AI has rapidly become integrated into everyday life, with many individuals using large language model–based systems, such as ChatGPT, Gemini, and Claude, for education, productivity, health information, decision-making, emotional support, and companionship. As patients increasingly incorporate conversational AI into their cognitive, emotional, and social lives, these interactions may influence coping strategies, treatment engagement, clinical decision-making, and mental health outcomes. However, although medical education has increasingly emphasized AI literacy and the responsible use of generative AI by clinicians, comparatively little attention has been given to preparing health care professionals to assess patients’ use of conversational AI during routine clinical encounters. This Viewpoint introduces the AWARE (AI use, why, attachment, reality and risk, and effect on functioning) framework, a practical educational framework designed to help mental health professionals systematically assess patients’ use of conversational AI. Rather than functioning as a diagnostic instrument or psychometric scale, AWARE provides a structured approach to psychiatric interviewing across 5 clinically relevant domains: AI use, why, attachment, reality and risk, and effect on functioning. Together, these domains guide clinicians in exploring patterns of AI use, the motivations underlying engagement, the emotional significance of patient-AI interactions, potential influences on reality testing and clinical risk, and the overall impact of AI on psychological well-being, daily functioning, relationships, and recovery. The framework is intended to complement existing psychiatric interviewing practices by supporting comprehensive history taking, clinical reasoning, risk assessment, documentation, and learner education. We discuss the rationale for routinely asking patients about AI use, review emerging evidence regarding both the potential benefits and risks of conversational AI, and describe how AWARE may be incorporated into undergraduate, postgraduate, and continuing professional education through simulation, objective structured clinical examinations, workplace-based assessment, and clinical supervision. We also outline priorities for future research, including validation, implementation, educational evaluation, and cross-cultural adaptation. As conversational AI becomes increasingly embedded within patients’ daily lives, clinicians require practical approaches to understanding its role in mental health and health care. The AWARE framework offers a structured educational starting point for integrating assessment of patient-AI interactions into routine psychiatric interviewing while supporting patient-centered, evidence-informed clinical practice and health professions education.
- Open Access
- 11 June 2026
- Published by SAGE Publications in The Canadian Journal of Psychiatry
A Chatbot for the Management of Bipolar Disorder: Using Retrieval-Augmented Generation With an Open-Weight Large Language Model to Answer Clinical Questions Based on the CANMAT and ISBD 2018 Guidelines for Bipolar Disorder: Un dialogueur pour la prise en charge du trouble bipolaire : utiliser la génération augmentée par récupération avec un grand modèle de langage (GML) à poids ouverts pour répondre aux questions cliniques fondées sur les lignes directrices de 2018 de CANMAT et de l'ISBD relatives au trouble bipolaire
Objective: Clinical practice guidelines support evidence-based care but are often underused due to complexity, time constraints, and navigation challenges. We investigated whether a conversational agent (chatbot) using an open-weight large language model (LLM) with retrieval-augmented generation (RAG) could provide guideline-consistent answers for bipolar disorder management based on the full 2018 Canadian Network for Mood and Anxiety Treatments (CANMAT) and ISBD guidelines, comparing against a system using only the base LLM. Method: We developed a multi-step RAG-based chatbot that retrieves relevant guideline sections and generates responses using Llama 3.3 70B. Twenty-one clinical vignettes spanning all guideline sections were created. Six expert psychiatrists generated queries and were presented with paired responses without labels from 2 systems: one using the base Llama 3.3 70B model, the other RAG-enhanced. Responses were rated for guideline consistency on a 3-point scale, and were analyzed using mixed-effects ordinal logistic regression. Results: Experts evaluated 126 responses, of which 110 (87.3%) were rated as correct as or more correct than the baseline system. The RAG system produced 80 answers (63.5%) rated fully consistent with the guidelines versus 24 (19.0%) for baseline, and only 10 answers with major deviation (7.9%) versus 48 (38.1%) for baseline. Ordinal regression showed RAG responses were significantly more likely to be more correct (OR = 9.1, 95% CI [5.3-16.3], P < 0.001), which was consistent across all raters. Preference ratings favoured RAG answers in 78.7% of cases. Performance varied by vignette, with some errors in both retrieval and reasoning. Conclusion: The use of RAG with an open-weight model helped produce answers consistent with the CANMAT guidelines across vignettes that required adapting or combining guideline text, suggesting a proof-of-concept of a bipolar guideline chatbot. We identified areas to improve results and evaluation. Future work should explore additional retrieval strategies and LLMs, and test in more naturalistic settings.
- 11 June 2026
- Published by SAGE Publications in The Canadian Journal of Psychiatry
Open AccessA Chatbot for the Management of Bipolar Disorder: Using Retrieval-Augmented Generation With an Open-Weight Large Language Model to Answer Clinical Questions Based on the CANMAT and ISBD 2018 Guidelines for Bipolar Disorder: Un dialogueur pour la prise en charge du trouble bipolaire : utiliser la génération augmentée par récupération avec un grand modèle de langage (GML) à poids ouverts pour répondre aux questions cliniques fondées sur les lignes directrices de 2018 de CANMAT et de l'ISBD relatives au trouble bipolaire
Objective: Clinical practice guidelines support evidence-based care but are often underused due to complexity, time constraints, and navigation challenges. We investigated whether a conversational agent (chatbot) using an open-weight large language model (LLM) with retrieval-augmented generation (RAG) could provide guideline-consistent answers for bipolar disorder management based on the full 2018 Canadian Network for Mood and Anxiety Treatments (CANMAT) and ISBD guidelines, comparing against a system using only the base LLM. Method: We developed a multi-step RAG-based chatbot that retrieves relevant guideline sections and generates responses using Llama 3.3 70B. Twenty-one clinical vignettes spanning all guideline sections were created. Six expert psychiatrists generated queries and were presented with paired responses without labels from 2 systems: one using the base Llama 3.3 70B model, the other RAG-enhanced. Responses were rated for guideline consistency on a 3-point scale, and were analyzed using mixed-effects ordinal logistic regression. Results: Experts evaluated 126 responses, of which 110 (87.3%) were rated as correct as or more correct than the baseline system. The RAG system produced 80 answers (63.5%) rated fully consistent with the guidelines versus 24 (19.0%) for baseline, and only 10 answers with major deviation (7.9%) versus 48 (38.1%) for baseline. Ordinal regression showed RAG responses were significantly more likely to be more correct (OR = 9.1, 95% CI [5.3-16.3], P < 0.001), which was consistent across all raters. Preference ratings favoured RAG answers in 78.7% of cases. Performance varied by vignette, with some errors in both retrieval and reasoning. Conclusion: The use of RAG with an open-weight model helped produce answers consistent with the CANMAT guidelines across vignettes that required adapting or combining guideline text, suggesting a proof-of-concept of a bipolar guideline chatbot. We identified areas to improve results and evaluation. Future work should explore additional retrieval strategies and LLMs, and test in more naturalistic settings.