Postoperative intensive care unit (ICU) admission affects 15% to 20% of surgical patients and represents a major source of morbidity and health care costs. Current anesthetic dosing relies on empirical guidelines rather than individualized risk assessment. We developed a counterfactual dose-response model to identify optimal fentanyl-propofol combinations. This study aimed to develop and evaluate a stratified, causal machine learning framework using electronic health record data to identify optimal fentanyl-propofol dose combinations and predict postoperative ICU admission risk, enabling precision anesthesia and individualized clinical decision support. We analyzed perioperative electronic health records of 67,134 surgical procedures from UC Irvine Medical Center (2017-2022). A hierarchical learning framework was used to estimate causal effects while controlling for confounding variables. A total of 6 dose-sensitive subgroups were identified through stratified analysis. The primary end point was postoperative ICU admission. High-risk combinations (fentanyl >5 mcg/kg with propofol <1 mg/kg) increased ICU admissions' absolute risk difference by 36% (absolute risk increase; 95% CI 0.351-0.509; P<.001). A total of 6 patient subgroups demonstrated distinct dose-response patterns, with populations considered vulnerable (high glucose, elevated creatinine) showing elevated risk even at standard doses. The optimal dose range for decision-making was determined to be 1.25 to 4.25 mg/kg for propofol and 3.5 to 4.0 mcg/kg for fentanyl. Fentanyl-propofol combinations exhibit complex, nonlinear dose-response relationships with ICU admission risk. High-dose combinations markedly increase risk through synergistic effects, while specific patient subgroups require enhanced monitoring even at standard doses. These findings support the development of individualized dosing algorithms and risk assessment tools that could inform future decision support tools aimed at reducing postoperative ICU use, although their predictive performance and clinical impact would require external validation.

Multiple-choice examinations (MCQs) are widely used in medical education to ensure standardized and objective assessment. Developing high-quality items requires both subject expertise and methodological rigor. Large language models (LLMs) offer new opportunities for automated item generation. However, most evaluations rely on general-purpose prompting, and psychometric comparisons with faculty-written items remain scarce. This study aimed to evaluate whether a fine-tuned LLM can generate MCQs (Type A) in anesthesiology with psychometric properties comparable to those written by expert faculty. The study was embedded in the regular written anesthesiology examination of the eighth-semester medical curriculum with 157 students. The examination comprised 30 single best-answer MCQs, of which 15 were generated by senior faculty and 15 by a fine-tuned GPT-based model. A custom GPT-based (GPT-4) model was adapted with anesthesiology lecture slides, the National Competence-Based Learning Objectives Catalogue (NKLM 2.0), past examination questions, and faculty publications using supervised instruction-tuning with standardized prompt-response pairs. Item analysis followed established psychometric standards. In total, 29 items (14 expert, 15 LLM-generated) were analyzed. Expert-generated questions had a mean difficulty of 0.81 (SD 0.19), point-biserial correlation of 0.19 (SD 0.07), and discrimination index of 0.09 (SD 0.08). LLM-generated items had a mean difficulty of 0.79 (SD 0.18), point-biserial correlation of 0.17 (SD 0.04), and discrimination index of 0.08 (SD 0.11). Mann-Whitney U tests revealed no significant differences between expert- and LLM-generated items for difficulty (P=.38), point-biserial correlation coefficient (P=.96), or discrimination index (P=.59). Categorical analyses confirmed no significant group differences. Both sets, however, showed only modest psychometric quality. Supervised fine-tuned LLMs are capable of generating MCQs with psychometric properties comparable to those written by experienced faculty. Given the limitations and cohort-dependency of psychometric indices, automated item generation should be considered a complement rather than a replacement for manual item writing. Further research with larger item sets and multi-institutional validation is needed to confirm generalizability and optimize integration of LLM-based tools into assessment development.

BackgroundDespite a 40-year prevalence of HIV, the AIDS epidemic prevails. Effective AIDS treatment requires specialist care and high adherence often hindered by structural issues in care access. Innovative eHealth solutions like treatment-supportive mobile apps can help address these issues. Successful implementation depends on user acceptance. Acceptance can be operationalized as behavioral intention and measured through the Unified Theory of Acceptance and Use of Technology (UTAUT).ObjectiveThis study examines the acceptance and its influencing factors of treatment-supportive mobile apps among men living with HIV.MethodsA cross-sectional study was conducted among 172 men living with HIV between September 2021 and April 2024. In addition to the collection of sociodemographic, medical, and eHealth-related data, acceptance and its influencing factors were assessed by applying the UTAUT model. A multiple hierarchical regression analysis was conducted.ResultsHigh acceptance of treatment-supportive mobile apps in men living with HIV was reported by 45.3% (n=78) of the participants. Significant predictors of acceptance were age (β=−0.27; P<.001); health literacy regarding disease (β=0.11; P<.001); eHealth literacy (β=0.10; P=.001); internet anxiety (β=−0.18; P=.04); and the UTAUT predictors: effort expectancy (β=0.38; P<.001), performance expectancy (β=0.24; P<.001), and social influence (β=0.40; P<.001). The UTAUT model explained 72% of the variance in acceptance.ConclusionsSince the acceptance of eHealth services is a reliable indicator of the actual usage behavior, the results of this study are a promising basis for the successful implementation of eHealth offerings in the group of men living with HIV.

BackgroundEffective diabetes management requires individualized treatment strategies tailored to patients’ clinical characteristics. With recent advances in artificial intelligence, large language models (LLMs) offer new opportunities to enhance clinical decision support, particularly in generating personalized recommendations.ObjectiveThis study aimed to develop and evaluate an LLM-based outpatient treatment support system for diabetes and examine its potential value in routine clinical decision-making.MethodsThree compact LLMs (Llama 3.1-8B, Qwen3-8B, and GLM4-9B) were fine-tuned on deidentified outpatient electronic health records using a parameter-efficient low-rank adaptation approach. The optimized models were embedded into a prototype hospital information system via a retrieval-augmented generation framework to generate individualized treatment recommendations, laboratory test suggestions, and medication prompts based on demographic and clinical data.ResultsAmong the models evaluated, the fine-tuned GLM4-9B demonstrated the strongest performance, producing clinically reasonable treatment plans and appropriate laboratory test recommendations and medication suggestions. It achieved a mean Bilingual Evaluation Understudy for 4-grams score of 67.93 (SD 2.74) and mean scores of 44.30 (SD 3.91) for Recall-Oriented Understudy for Gisting Evaluation for overlap of unigrams, 27.34 (SD 1.85) for Recall-Oriented Understudy for Gisting Evaluation for overlap of bigrams, and 37.67 (SD 2.88) for Recall-Oriented Understudy for Gisting Evaluation for Longest Common Subsequence.ConclusionsThe fine-tuned GLM4-9B shows strong potential as a clinical decision support tool for personalized diabetes care. It can provide reference recommendations that may improve clinician efficiency and support decision quality. Future work should focus on enhancing medication guidance, expanding data sources, and improving adaptability in cases involving complex comorbidities.

In the United Kingdom, there was an increased demand for young people's mental health helpline services during COVID-19 pandemic, when face-to-face services were often inaccessible. Despite this, there is scant research examining young people's experiences with these helplines during the pandemic and postpandemic periods. Using a cross-sectional survey, this mixed methods study aims to examine young people's (aged 16-25+ years) experiences with the multichannel helpline provided by The Mix (Mental Health Innovations), the United Kingdom's leading online mental health support service for young people during and after the pandemic. From February 2020 to October 2023, approximately 16,000 users aged 16-25+ years contacted The Mix's helpline. All users were sent an email by The Mix following helpline contact to answer their user survey. Of these, 796 participants aged 16-25+ years answered the survey, representing a response rate of 5%, with a survey completion rate of 65.3%. To address potential nonresponse bias and missing data concerns, a multiple imputation procedure using the Multiple Imputation by Chained Equations (MICE) package in R (R Core Team) provided a final imputed sample for both the pandemic (n=295) and postpandemic (n=501) periods. Open-ended survey responses from users were also explored. Of the 796 participants who responded to the survey, there were 1183 open-ended responses from 486 respondents. Of these, a total of 731 open-ended responses (approximately 60% of the total responses) were coded. The criteria for inclusion were applied by 2 independent coders. Excluded responses focused on single words (eg, "thanks"), irrelevant text, or duplicated entries, ensuring only responses containing substantive feedback were analyzed. During the pandemic, young people who contacted the helpline reported greater satisfaction after the first lockdown and a stronger perceived impact on their well‑being after the first lockdown and during the second and third lockdowns, compared with those who contacted the helpline during the gradual easing period; phone users reported higher satisfaction than those using the contact form. Postpandemic, helpline users who identified as "other" in terms of their gender reported less satisfaction, while male users reported a greater impact on their well-being compared to female users. Qualitative analysis revealed how the participants felt supported by the helpline, such as "feeling heard" and "being empowered," and areas for improvement across service delivery, protocol, and technicalities. The findings highlight the important role that helplines play in supporting young people's mental health, particularly in crises like the pandemic. This study underscores the need for service improvements to ensure young people continue to feel supported by helplines, highlighting key areas for improvement in service delivery, protocols, and technical infrastructure. Future research should explore channel preferences and minority experiences, often underrepresented in helpline studies.

BackgroundDespite digital mental health services growing at a rapid pace to address global mental health needs, there exist challenges of low engagement and attrition. Ensuring continuity of care in the digital context can positively impact mental health care delivery and adherence to treatment, helping to establish digital mental health interventions (DMHIs) as a viable option for mental health support.ObjectiveThis study aimed to examine the impact of adjunct app features of the mental health app Wysa and their ability to promote engagement and adherence to the text-based coaching sessions.MethodsThis retrospective mixed methods observational study was based on real-world app data from users (n=1213) who subscribed to text-based sessions with mental health coaches (MHCs) between February 1 and July 31, 2022. Their engagement with the adjunct app features, such as brief interventions with the conversational agent, self-management tools, and journaling, was analyzed quantitatively using descriptive statistics. Acceptability of the app features was also assessed using qualitative feedback data. Adherence to sessions with MHCs was compared between app feature users (n=1042, 85.9%) and nonfeature users (n=171, 14.1%) using inferential statistics. Subgroup analysis was not feasible in the absence of demographic and clinical user data, potentially limiting the generalizability of the findings.ResultsFindings demonstrated high use of the adjunct app features, which allowed communication with the MHCs in between sessions. The thematic analysis captures user experiences of helpfulness within the app and with the MHCs. The Mann-Whitney U test indicated that users who accessed one or more features completed significantly more sessions compared with users who did not use any feature (Mann-Whitney U=154,085.0; P<.001; rB=0.73) with a large effect size. The odds ratio analysis indicated that users were almost thrice as likely to complete sessions after using the adjunct app features (odds ratio 2.91, 95% CI 2.24-3.38; P<.001).ConclusionsInclusion of adjunct app features enhances continuity in care delivery between sessions with MHCs and is associated with improved engagement with DMHIs. Further efforts are needed to assess the impact of this approach in DMHIs on clinical mental health outcomes.

Caregivers' self-perceived preparedness for caregiving influences care recipients' and caregivers' emotional health, and care recipients' aging in place. Dementia's unique, long, and progressive nature compared to other age-related illnesses, along with associated behavioral symptoms and personality changes, may cause caregivers' preparedness to vary significantly from that of those caring for patients with other chronic conditions. This study aimed to describe and compare specific domains and tasks in which family caregivers of veterans with and without dementia reported wanting to be better prepared. Using the Veterans Affairs' HERO CARE (Home Excellence Resource Outcome Center to Advance, Redefine, and Evaluate Non-Institutional Care) Survey data, we analyzed caregivers' responses to one open-ended question: "Out of all the tasks that you help the veteran with, is there anything specific you would like to be better prepared for?" Response themes were deductively coded into 9 domains, and differences in reported domains between caregivers of care recipients with and without dementia were compared. A total of 732 caregivers were included: 301 (41.1%) caregivers of veterans with dementia and 431 (58.9%) without. Caregivers of veterans with and without dementia, respectively, were similar except in age, being spousal caregivers, working at least part-time, hours of care provision per week, and proportion with a high burden. Veterans with dementia, versus without, were older and had higher frailty and risk scores. Preparedness concerns among caregivers (N=732) included care coordination (n=164, 22.4%), emotional and social support (n=145, 19.8%), advance planning (n=116, 15.8%), nursing and health monitoring (n=94, 12.8%), personal care (n=65, 8.9%), mobility (n=79, 10.8%), household (n=58, 7.9%), caregiver self-care (n=36, 4.9%), and emergent situations (n=28, 3.8%). The commonest tasks caregivers expressed needs for included managing emotional and behavioral symptoms (n=74, 10.1%), recognizing and responding to significant changes in the veterans' condition (n=66, 9.0%), seeking medical information relevant to the veterans' needs (n=54, 7.4%), handling financial and legal matters (n=52, 7.1%), and advocating for services (n=49, 6.7%). Similar proportions of caregivers of veterans with and without dementia reported preparedness needs in all domains and tasks. The preparedness needs of caregivers of veterans with and without dementia were mostly similar in most domains and tasks. The commonest preparedness gaps were in the domains of care coordination, emotional and social support, and advance planning. The findings can inform interventions to prepare all caregivers to support aging in place.

BackgroundEating difficulties are increasingly prevalent among young people, yet service capacity remains limited. Digital interventions may provide accessible, scalable support, particularly for those with mild, subthreshold, or early-stage symptoms who do not meet criteria for specialist care. Low self-esteem is widely recognized as a key psychological risk factor in the onset and persistence of eating disorders, and negative self-evaluation, particularly around body image and social acceptance, can heighten vulnerability to the maladaptive thoughts and behaviors seen in conditions such as anorexia nervosa, bulimia nervosa, and binge-eating disorder. Clarifying this relationship is essential for developing effective prevention and early intervention strategies.ObjectiveThis pilot case series reports on 5 individuals aged 19-25 (mean 22, SD 2.19) years with mild eating disorders who used the Worth Warrior app, a mobile intervention incorporating principles of enhanced cognitive behavioral therapy strategies targeting low self-esteem, body image concerns, and disordered eating behaviors.MethodsAn uncontrolled 3-phase design (baseline, post–app familiarization, and follow-up) was used. Participants completed standardized self-report tools, including the Eating Disorder Examination Questionnaire and Rosenberg Self-Esteem Scale. Feedback on usability, acceptability, and safety was also collected via online questionnaires.ResultsOutcome measures at follow-up showed improvements in eating disorder symptomatology in 3/5 cases, and in self-esteem in 4/5 cases; those with milder symptomatology indicated the most benefit. Reductions in eating concerns, weight concerns, and related behaviors were observed in most, though not all, cases. Participants valued interactive enhanced cognitive behavioral therapy features and journaling functions, while noting areas for improvement such as reminders and incentives for use and preventions for maladaptive use of the free-text facilities.ConclusionsFindings suggest the Worth Warrior app may be suited as an acceptable and effective standalone tool for individuals with mild eating disorder symptoms, and used as an adjunct to traditional treatment alongside clinician supervision for those with more severe presentations to promote the greatest patient safety. These exploratory case study findings suggest that the app has the potential to support improvements in self-esteem and mild eating disorder symptomatology; however, as a preliminary case series, these results are not generalizable but provide a foundation for larger, controlled studies of digital early intervention.

BackgroundAccess to care that affirms one’s entire self is essential, especially for gender-diverse individuals. Gender-affirming care includes medical, social, and nonmedical supports to affirm gender identity.ObjectiveThis study qualitatively examined the importance of social and nonmedical gender-affirming services as described by gender-diverse community members.MethodsThematic analysis was conducted on qualitative data from 5 participants (3 rural and 2 urban; 2 with doctoral-level education and 3 health professionals) with experiences accessing gender-affirming care in Nova Scotia, Canada, between October 2023 and November 2023.ResultsParticipants included transgender and/or nonbinary individuals who highlighted the significance of social and nonmedical gender-affirming care over traditional medical interventions. Themes included the centrality of belonging, the use of online spaces such as TikTok for gender affirmation, and the emotional impact of barriers such as cost and safety concerns. Four of 5 participants emphasized the importance of social and nonmedical gender-affirming care over medical interventions. Participants stressed the importance of fostering a sense of belonging and accessing supportive communities, which is crucial in navigating transphobic environments without support. Many felt abandoned by public systems and resorted to passing as cisgender due to barriers such as cost in accessing gender-affirming resources. Internet platforms such as TikTok provided valuable guidance, supplementing limited access to medical gender-affirming care. Participants emphasized a crucial need for health care providers to understand basic gender-affirming care, including respect for preferred pronouns and gender identities.ConclusionsThis study found that members of the gender-diverse community significantly value social and nonmedical gender-affirming care services with respect to their well-being. The findings underscore the complex interplay among social support, health care access, and resilience in transgender and/or nonbinary individuals’ lives. This work can aid in exploring how best to educate health care providers in gender-inclusive care and enable increased access to all forms of care that can help affirm an individual’s gender.

BackgroundA growing volume of mental health research is conducted with participants recruited and responding online. However, to date, few psychometric scales have been specifically validated for online research.ObjectiveWe aimed to devise a brief, 12-item version of the Center for Epidemiological Studies Depression Scale (CES-D) in which first order factors are sufficiently measured.MethodsWe recruited 218 adults with depression and 226 comparison participants with no mental health history. Both groups completed the original 20-item CES-D and measures of social support, psychological distress, and sociodemographic information (eg, age, gender, and household income). Measurement of social support included online support, and psychological distress included symptoms of social media use disorder along with loneliness and life dissatisfaction.ResultsThis brief, 12-item version of the CES-D was devised with persons with depression and replicated with comparison participants. For both, core sadness, somatic symptoms, interpersonal detachment, and absence of well-being each significantly contributed to measurement of a higher-order depression latent construct (P<.01). Structural equation modeling was performed to establish the construct validity of this 4-factor model in which depression is predicted by socioeconomic factors and depression predicts lower social support as well as greater psychological distress.ConclusionsResponses to this 12-item, online version of the CES-D demonstrate factorial and construct validity. Clinical research is required in future to ascertain whether scores greater than 11 (of 36) are suggestive of elevated depressive symptomology.