Frontiers in Digital Health

Introduction Large language models are increasingly being used in healthcare. In interventional pain medicine, clinical reasoning is essential for procedural planning. Prior studies show that simplified prompts reduce clinical detail in AI-generated responses. It remains unclear whether this reflects knowledge loss or simply prompt-driven suppression of information. Methods We performed a controlled comparative study using 15 standardized low back pain questions representing common interventional pain questions. Each question was submitted to ChatGPT under three conditions, professional-level prompt (DP), fourth-grade reading-level prompt (D4), and clinician-directed rewriting of the D4 response to a medical level (U4[->]MD). No follow-up prompting was allowed. Three physicians independently rated responses for accuracy using a 0-2 ordinal scale. Clinical completeness was determined by consensus. Word count and Flesch-Kincaid Grade Level (FKGL) were also measured. Paired t-tests compared conditions. Results Accuracy was highest with professional prompting (1.76). Accuracy declined with the fourth-grade prompt (1.33; p = 0.00086). When simplified responses were rewritten for clinicians, accuracy returned to baseline (1.76; p {approx} 1.00 vs DP). Clinical completeness followed the same pattern showing DP 80.0%, D4 6.7%, U4[->]MD 73.3%. Fourth-grade responses were shorter and less complex. Upscaled responses were more complex and similar in length to professional responses. Inter-rater reliability was low (Fleiss {kappa} = 0.17), but trends were consistent across conditions. Conclusions Reduced clinical detail under simplified prompts appears to reflect constrained output rather than loss of knowledge. Clinician-directed reframing restores omitted content. LLM performance in interventional pain depends strongly on prompt design and intended audience.

A Wearable Monitoring System (WMS), comprising a chest patch, wrist-worn pulse oximeter, and arm-worn blood pressure device, was developed in preparation for a pilot Randomised Controlled Trial (RCT) on a UK surgical ward. The system was designed to support continuous physiological monitoring and early detection of deterioration. An initial prototype user interface was developed by the research team based on prior clinical experience and engineering knowledge. To ensure suitability for clinical practice, iterative user-centred refinement was undertaken through a series of clinician focus groups and wearability assessments. Six focus groups were conducted between November 2019 and May 2021 involving multidisciplinary healthcare professionals. Feedback from these sessions informed successive interface and system modifications. System development spanned the COVID-19 pandemic, during which the WMS was rapidly adapted and deployed to support clinical care on isolation wards. Feedback obtained during this period was incorporated into later versions of the system and provided a unique opportunity to examine changes in clinician priorities under pandemic conditions. Clinicians consistently prioritised alert visibility, alarm fatigue mitigation, parameter flexibility, and centralised monitoring. Notably, preferences regarding alert modality and access mechanisms evolved over time: early enthusiasm for mobile or smartphone-type devices shifted towards a preference for fixed, ward-based displays and audible alerts at the nurses station following pandemic deployment. Building on previous wearability testing in healthy volunteers, wearability testing using a validated questionnaire was completed by 169 patient participants during the RCT. The chest patch and pulse oximeter demonstrated high tolerability, whereas the blood pressure cuff showed poor wearability and was removed from the final system. These findings demonstrate the importance of iterative, clinician-led design for wearable WMS and highlight how extreme clinical contexts such as the COVID-19 pandemic can significantly reshape perceived requirements for safety-critical monitoring technologies.

Objectives Publicly available datasets dedicated to clinical speech deidentification tasks remain scarce due to privacy constraints and the complexity of speech-level annotation. To address this gap, we compiled the SREDH-AICup sensitive health information (SHI) speech corpus, a time-aligned clinical speech dataset annotated across 38 SHI categories. Methods Two publicly available English medical-domain datasets were adapted to support speech-level de-identification, including script reformulation and controlled re-recorded by 25 participants. Additional Mandarin Chinese clinical-style materials were incorporated to extend linguistic coverage. All audio data were annotated with million-level, time-aligned SHI spans using Label Studio. Inter-annotator agreement was evaluated using Cohen's kappa, following iterative calibration rounds. The resulting corpus supports both automatic speech recognition (ASR) and speech-level recognition of SHIs. Results The final dataset comprises 20 hours of annotated audio, divided into training (10 hours, 1,539 files), validation (5 hours, 775 files), and test (5 hours, 710 files) subsets, totalling 7,830 SHI entities. The language distribution reflects the composition of the selected source materials, with 19.36 hours of English and 0.89 hours of Mandarin Chinese speech. Discussion The corpus exhibits a long-tail distribution consistent with clinical documentation patterns and highlights the limited availability of Chinese medical speech resources. These characteristics underscore both the realism of the dataset and structural challenges associated with multilingual speech de-identification. Conclusion The SREDH-AICup SHI speech corpus provides a clinically grounded, time-aligned speech dataset supporting automated medical speech de-identification research and facilitating future development of multilingual speech-based privacy protection systems.

Background: The urgent care departments in Europe face a structural paradox: accelerating digitalisation is accompanied by a patient population that is disproportionately unable to engage with standard digital tools. An internal analysis at the Emergency Department (Akutafdelingen) of Nordsjaellands Hospital in Hilleroed, Denmark found that 43% of emergency patients struggle with digital solutions - a figure that reflects the predictable composition of acute care populations rather than any individual failing. Objective: This paper presents the design, iterative development, and secondary validation of the ED Adaptive Interface (v5): a prototype adaptive patient terminal developed in response to this challenge. The system operationalises what the author terms impairment-first design - a methodology that treats the most constrained patient experience as the primary design problem and derives the standard experience as a subset. The interface configures itself in under ten seconds via nurse-led setup, adapting across four axes of impairment: visual, motor, speech, and cognitive. System: Version 4 supports five accessibility modes, a heatmap pain assessment grid, a Privacy and Dignity panel, a live workflow tracker with care notifications, structured dual-category help requests, and plain-language medical term definitions across four languages. Version 5, reported here for the first time, introduces a Condition Worsening Escalation button, a Referral Pathway Display, a "Why Am I Waiting?" triage explainer, a Symptom Progression Log, MinSP/Yellow Card Scan simulation, expanded language support (seven languages: English, Danish, Arabic with full RTL layout, Turkish, Romanian, Polish, and Somali), and an expanded ten-item Communication Board. The entire system runs as a single 79-kilobyte HTML file with zero infrastructure requirements. Methods: To base the design on patient-generated evidence, two independent social media threads were subjected to an inductive thematic analysis (Braun and Clarke, 2006): a primary corpus of 83 entries in the Facebook group Foreigners in Denmark (collected March 2026) and a corroborating corpus in an international community group in the Aarhus region (collected April 2026). All identifiers in both datasets were fully anonymised under GDPR Article 89 research provisions prior to analysis. No participants were contacted. Generative AI tools were used to assist with drafting, writing, and prototype code development; all scientific content, data collection, analysis, and conclusions are the sole responsibility of the authors. Results: The first discourse corpus produced five major themes corresponding to the five problem areas the prototype was designed to address: system navigation and triage literacy gaps (31 entries); language and cultural barriers (6 entries); communication failures during care (5 entries); staff overload and capacity constraints (8 entries); and pain and severity assessment failures (14 entries). The corroborating dataset supported all five themes and introduced two additional themes: differential treatment of international patients and medical gaslighting as a long-term pattern of patient advocacy failure. One structural finding - the five most-liked comments incorrectly criticised the original poster for self-referring when she had received explicit 1813 telephone triage approval - directly inspired the Referral Pathway Display and "Why Am I Waiting?" features in v5. Conclusions: The convergence of design rationale and independent social evidence across all five problem categories suggests that impairment-first design is not a niche accessibility concern but a structural approach to healthcare interface quality. The prototype is ready for a structured clinical pilot using the System Usability Scale (SUS) and semi-structured staff interviews. The long-term roadmap includes full MinSP integration, hospital PMS connectivity, and clinical validation.

Background: Secure text messages (TMs) exchanged among interdisciplinary care teams in nursing homes (NHs) contain clinical information that aligns with the Age-Friendly Health Systems 4Ms: What Matters, Medication, Mentation, and Mobility, yet, this information is not captured in any structured form, making it unavailable for systematic monitoring or quality reporting. Automatically extracting 4M information accurately and efficiently from these messages could enable several downstream applications within long term care settings. This task, however, is challenging because of the fragmented syntax, brevity, abbreviations, and informality of TMs. Objective: This study aimed to develop and evaluate a multi-stage 4M Entity Recognition (4M-ER) pipeline that combines a fine-tuned token classifier with large language model (LLM) revision, using only locally deployed open-source models, to improve 4M information extraction from clinical TMs. Methods: We used an expert-annotated dataset of 1,169 TMs collected from interdisciplinary teams across 16 Midwest NHs. The pipeline first identifies candidate text spans using a fine-tuned Bio-ClinicalBERT token classifier. A semantic similarity retriever then selects in-context exemplars to guide an LLM revision in which the LLM (Gemma, Phi, Qwen, or Mistral) performs boundary correction, label evaluation, and selective acceptance or rejection of candidate spans. Baselines for comparison included single-stage zero-shot LLMs, single-stage fine-tuned Bio-ClinicalBERT, and a fine-tuned LLM (Gemma) from a prior study. Ablation studies assessed the contribution of each pipeline stage and the effect of message filtering. Robustness was evaluated across 5 repeated runs. Results: The 4M-ER pipeline outperformed the previously fine-tuned Gemma LLM across all 4M domains, achieving F1 (entity type) improvements of +2 to +11 percentage points without any additional fine-tuning and at roughly half the GPU memory (12 vs 24 GB). It also improved upon single-stage fine-tuned Bio-ClinicalBERT in Mobility, Mentation, and What Matters (+0.02 to +0.05 F1). Error analysis showed that LLM revision reduced false positives by 25% to 35% by correcting misclassifications caused by conversational ambiguity, while the fine-tuned Bio-ClinicalBERT's high recall captured subtle entities that the fine-tuned Gemma missed. Silver data augmentation further improved the hardest domains, raising What Matters F1 from 0.59 to 0.67 and Mobility from 0.64 to 0.67. Ablation studies confirmed that restricting LLMs to revision only yielded optimal accuracy and efficiency. Conclusions: The 4M-ER pipeline enables accurate and scalable extraction of 4M entities from clinical TMs by combining fine-tuned Bio-ClinicalBERT with LLM revision using only locally deployed open-source models. The structured 4M data produced by the pipeline can support 4M taxonomy and ontology construction, as demonstrated in the prior work, and provides a foundation for downstream applications including real-time clinical surveillance, compliance with emerging age-friendly quality measures, and predictive modeling in long-term care settings.

Accurate recognition and deidentification of sensitive health information (SHI) in spoken dialogues requires multimodal algorithms that can understand medical language and contextual nuance. However, the recognition and deidentification risks expose sensitive health information (SHI). Additionally, the variability and complexity of medical terminology, along with the inherent biases in medical datasets, further complicate this task. This study introduces the SREDH/AI-Cup 2025 Medical Speech Sensitive Information Recognition Challenge, which focuses on two tasks: Task-1: Speech transcription systems must accurately transcribe speech into text; and Task-2: Medical speech de-identification to detect and appropriately classify mentions of SHI. The competition attracted 246 teams; top-performing systems achieved a mixed error rate (MER) of 0.1147 and a macro F1-score of 0.7103, with average MER and macro F1-score of 0.3539 and 0.2696, respectively. Results were presented at the IW-DMRN workshop in 2025. Notably, the results reveal that LLMs were prevalent across both tasks: 97.5% of teams adopted LLMs for Task 1 and 100% for Task 2. Highlighting their growing role in healthcare. Furthermore, we finetuned six models, demonstrating strong precision ([~]0.885-0.889) with slightly lower recall ([~]0.830-0.847), resulting in F1-scores of 0.857-0.867.

Social determinants of health (SDoH) are important for clinical care, but it remains unclear how much AI-captured social context is preserved after clinician editing in ambient documentation workflows. We retrospectively analyzed 75,133 paired ambient AI-drafted and clinician-finalized note sections from ambulatory care at a large academic health system. Using a rule-based NLP pipeline, we extracted 21 SDoH categories and quantified retention, deletion, and addition. SDoH appeared in 25.2% of AI drafts versus 17.2% of final notes. At the mention level, AI captured 29,991 SDoH mentions, of which 45.1% were deleted, 54.9% were retained with clinicians adding 3,583 new mentions. Insurance and marital status were most often deleted, whereas substance use and physical activity were more often retained. Deletion patterns also varied by specialty, supporting the need for specialty-aware ambient AI systems.

In this study, we investigate gender bias in a Retrieval-Augmented Generation (RAG) based AI assistant developed for Finnish wellbeing services counties. We tested the system using 36 clinically relevant queries, each rendered in three gendered variants (male, female, gender-neutral), and evaluated responses using both an LLM-as-a-judge approach and a human expert panel consisting of a physician and a sociologist specializing in ethics. We observed substantial and clinically significant differences across gendered variants, including differential treatment urgency, inappropriate symptom associations, and misidentification of clinical context. Female variants disproportionately framed responses around childcare and reproductive health regardless of clinical relevance, reflecting societal stereotypes rather than medical reasoning. Bias manifested both at the LLM generation stage and the RAG retrieval stage, in several cases causing the model to hallucinate responses entirely. Some bias patterns were persistent across repeated runs, while others appeared inconsistently, highlighting the challenge of distinguishing systematic bias from stochastic variation.

ObjectiveLong term monitoring of obstructive sleep apnea (OSA) severity may be relevant for several clinical applications. We developed a method for estimating the apnea-hypopnea index (AHI) using wrist-worn, reflective photoplethysmography (PPG). ApproachA neural network was developed to detect respiratory events using PPG and PPG-derived sleep stages as input. The development database encompassed retrospective data from three polysomnographic datasets (N=3111), including a dataset with concurrent reflective PPG recordings from a wrist-worn device (N=969). The model was pre-trained with (transmissive) finger-PPG signals from all overnight recordings and then fine-tuned to wrist-PPG characteristics using transfer learning. Validation was performed on the test portion of the development set and on a fourth, external hold-out dataset containing both wrist-PPG and PSG data (N=171). Performance was evaluated in terms of AHI estimation accuracy and OSA severity classification. Main ResultsThe fine-tuned wrist-PPG model demonstrated strong agreement with the PSG-derived gold-standard AHI, achieving intra-class correlation coefficients of 0.87 in the test portion of the development set and 0.91 in the external hold-out validation set. Diagnostic performance was high, with accuracies above 80% for all severity thresholds. SignificanceThe study highlights the potential of reflective PPG-based AHI estimation, achieving high estimation performance in comparison with PSG. These measurements can be performed with relatively comfortable sensors integrated in convenient wrist-worn wearables, enabling long-term assessment of sleep disordered breathing, both in a diagnostic phase, and during therapy follow-up.

BackgroundRecruitment remains a major barrier to timely clinical trial completion. Trialshub is an LLM-powered, chat-based platform intended to help users identify relevant trials and connect with coordinators to streamline recruitment workflows. ObjectiveTo evaluate the perceived usability and operational value of Trialshub, and identify implementation considerations for real-world deployment. MethodsA usability test was conducted at Morehouse School of Medicine for the Trialshub application. Purposively selected participants included clinical research coordinators and individuals with and without clinical trial search experience. Participants completed a pre-test survey assessing demographics, digital health information behaviors, and familiarity with AI tools, followed by a moderated usability session using a Trialshub prototype. Users completed scenario-based tasks (locating a breast cancer trial, reviewing results, and initiating coordinator contact) using a think-aloud protocol. Task ratings, screen recordings, and transcribed feedback were analyzed descriptively and thematically, and reported. ResultsParticipants reported high comfort with using digital tools and moderate-to-high familiarity with AI. Trialshubs chat-first design, guided prompts, and checklist-style eligibility display were perceived as intuitive and reduced cognitive load. Fast access to trials and the coordinator-contact workflow were viewed positively. Key usability issues included uncertainty at step transitions, insufficient cues for selecting results and next actions, and inconsistent system reliability (loading delays, errors, and broken trial detail pages). Participants also noted redundant questioning due to limited conversational memory, requested improved filtering/sorting, and clearer calls-to-action. All participants indicated that Trialshub has strong potential to meaningfully improve clinical trial processes. ConclusionsTrialshub shows promise for improving trial discovery and recruitment workflows, with identified design implications for real-world deployment.

Chronic low back pain (CLBP) is persistent and refractory, affecting 20-30% of population worldwide. Neurofeedback has been explored as a potential non-pharmacological intervention for chronic pain, although evidence in CLBP remains limited. This study evaluated PainWaive, a consumer-grade digitally-delivered neurofeedback intervention targeting multiple pain-related frequency bands recorded over the sensorimotor cortex in individuals with CLBP. In a multiple-baseline experimental design, four participants completed daily assessments of pain severity and pain interference during randomly-assigned baseline phases of 7, 10, 14, and 20 days, followed by 20 sessions of the PainWaive intervention over four weeks. Daily pain assessments continued during the post-intervention and follow-up phases. Participants rated PainWaive's usability and acceptability at post-intervention. Anxiety, depression, wellbeing, and sleep disturbance were assessed at three timepoints. Aggregated Tau-U analyses indicated a large effect (-0.67) on pain severity from baseline to intervention and very large from baseline to post-intervention (-0.92) and follow-up (-0.92) phases. Large effects (-0.63, -0.62, and -0.70) were also observed for pain interference. Individual-level analyses showed significant reductions across all participants, with visual inspection confirming progressive decreases over time. The intervention was rated usable and acceptable by all participants, while psychological outcomes were mixed and varied across participants. The findings provide promising evidence that the PainWaive neurofeedback intervention may reduce pain severity and pain interference in some individuals with CLBP. By prioritising accessibility, usability, and self-administration, PainWaive supports a foundation for more patient-centred, technology-enabled approaches to chronic pain management. Further evaluation of this approach in randomised trials is required to establish efficacy.

Inadequate discharge communication is a well-documented contributor to medication non-adherence, missed follow-ups, and preventable readmissions across healthcare systems worldwide. In resource-limited oncology settings, where patients are often low-literate, speak non-dominant languages, and manage complex multi-drug regimens, this problem is acute and largely unaddressed. We present Aakhyan, a vernacular patient communication platform that addresses the full post-discharge arc: from converting English-language discharge summaries into structured, voice-based vernacular explanations, through medication adherence support, to proactive follow-up management - all delivered via WhatsApp. The architecture is novel in its strict separation of concerns: a vision-language model performs structured JSON extraction from discharge images; all patient-facing content is generated deterministically from clinician-approved templates with community-sensitive vocabulary registers. This design eliminates the hallucination risk inherent in generative AI patient communication (documented at 18-82% in prior studies) while preserving the extraction capability of large language models. The platform supports four language registers, Bengali, Hindi, simplified English for tribal populations, and Assamese, with text-to-speech synthesis across all registers, including a custom grapheme-to-phoneme engine developed for Assamese phonology. Beyond discharge communication, the platform includes scheduled medication adherence nudges, interactive follow-up reminders, and a Daily Availability and Patient Notification System (DAPNS) that notifies patients the evening before their follow-up whether their doctor and required investigations are available, preventing wasted trips by rural patients who travel 2-6 hours to reach the centre. A 100-patient stratified randomised controlled study is planned at Silchar Cancer Centre, with structured teach-back assessment at 48-72 hours post-discharge as the primary comprehension outcome and preliminary clinical efficacy as a secondary objective. This paper describes the clinical rationale, technical architecture, safety framework, and positioning of Aakhyan within the existing literature on mHealth patient communication interventions.

Patient-generated streaming data from wearable and digital technologies is increasingly promoted as a means of supporting mental health monitoring and clinical decision-making. While patient acceptance of these technologies has been reported, clinician perspectives remain underexplored despite their central role in determining whether streaming data are meaningfully integrated into routine care. This study explored clinicians experiences, as well as perceived facilitators and barriers, related to integrating patient-generated streaming data into routine mental health practice. A qualitative, exploratory interview study was conducted to examine clinicians experiences and perspectives on integrating patient-generated streaming data into mental health care. Semi-structured interviews were conducted with 33 clinicians, including family physicians (n=11), psychiatrists (n=12), and psychologists (n=10). Data were analyzed using reflexive thematic analysis guided by Braun and Clarkes six-step approach. Six themes were identified. Clinicians described variable use of digital and streaming technologies, ranging from routine engagement to deliberate non-use. Streaming data were viewed as clinically valuable when they provided longitudinal and objective insights, identified physiological and behavioural pattern changes, and supported patient engagement. However, clinicians emphasized that clinical usefulness was contingent on interpretability, contextual information, and relevance to decision-making. Major barriers included poor integration with electronic medical records, time constraints, data volume, limited organizational support, and uncertainty regarding data reliability and validity. Clinicians also expressed persistent concerns about privacy, governance, and regulatory oversight, highlighting the need for clear safeguards and accountability structures. Clinicians view patient-generated streaming data as a promising adjunct to mental health care, particularly for capturing longitudinal change between visits. However, meaningful clinical integration remains constrained by usability, workflow, organizational, and regulatory challenges, as well as limited confidence in data interpretation. Addressing these barriers through improved system integration, interpretive support, validation, and governance will be essential for translating the potential of streaming data into routine clinical practice.

The goal of this work was to leverage a large corpus of text based psychotherapy data to create novel machine learning algorithms that can identify suicide risk in asynchronous text therapy. Advances in the field of natural language processing and machine learning have allowed us to include novel data sources as well as use encoding models that can represent context. Our models utilize advanced natural language processing techniques, including fine-tuned transformer models like RoBERTa, to classify risk. Subsequent model versions incorporated non-text data such as demographic features and census-derived social determinants of health to improve equitable and culturally responsive risk assessment, as well as multiclass models that can identify tiered levels of risk. All new models demonstrated significant improvements over our previous model. Our final version, a multiclass model, provides a tiered system that classifies risk as "no risk," "moderate," or "severe" (weighted F1 of 0.85). This tiered approach enhances clinical utility by allowing providers to quickly prioritize the most urgent cases, ensuring a more accurate and timely intervention for clients in need.

Objectives: Large language models (LLMs) have shown promise in creating clinical codelists for research purposes, a time-consuming task requiring expert domain knowledge. Here, we evaluate the performance and assess failure modes of a retrieval augmented generation (RAG) approach to creating clinical codelists for the large and complex medical terminology used by the Clinical Practice Research Datalink (CPRD). Materials & Methods: We set up a RAG system using a database of word embeddings of the medical terminology that we created using a general-purpose word embedding model (gemini-embedding). We developed 7 reference codelists presenting different challenges and tagged required and optional codes. We ran 168 evaluations (7 codelists, 2 different database subsets, 4 models, 3 epochs each). Scoring was based on the omission of required codes, and inclusion of irrelevant codes. We used model-grading (i.e., grading by another LLM with the reference codelists provided as context) to evaluate the output codelists (a score of 0% being all incorrect and 100% being all correct). Results: We saw varying accuracy across models and codelists, with Gemini 3 Pro (Score 43%) generally performing better than Claude Sonnet 4.6 (36%), Gemini 3 Flash, and OpenAI GPT 5.2 performing worst (14%). Models performed better with shorter target codelists (e.g., Eosinophilic esophagitis with four codes, and Hidradenitis suppurativa with 14 codes). For example, all models consistently failed to produce a complete Wrist fracture codelist (with 214 required codes). We further present evaluation summaries, and failure mode evaluations produced by parsing LLM chat logs. Discussion: Besides demonstrating that a single-shot RAG approach is currently not suitable for codelist generation, we demonstrate failure modes including hallucinations, retrieval failures and generation failures where retrieved codes are not used. Conclusions: Our findings suggest that while RAG systems using current frontier LLMs may create correct clinical codelists in some cases, they still struggle with large and complex terminologies and codelists with a large number of codes. The failure mode we highlight can inform the creation of future workflows to avoid failures.

Background: Continuous adhesive patch electrocardiographic (ECG) wearables are increasingly prescribed. Patient experience with these devices can influence adherence, but research in this area is limited. This study aimed to explore the perceptions and experiences of patients receiving wearable cardiac monitoring technology as part of their routine care through the lens of treatment burden. Methods: This was a qualitative study with semi-structured phone interviews conducted between February and May 2024. We recruited participants from primary care and outpatient clinics using maximum variation sampling to ensure diversity in sex, ethnicity, and education levels. Interviews were audio-recorded, transcribed, and analysed using reflexive thematic analysis. Results: Sixteen participants (mean age 51 years, 63% female) were interviewed (average duration: 33 minutes). Three themes were developed: 1) ?Experience using the device: Burden vs Ease of Use?, which captured participants? perceptions of how easily they could integrate the device in their daily lives; 2) ?Individual variability in responses to ECG self-monitoring? covered participants? emotional and cognitive response to knowing their heart rhythm was monitored; and 3) ?The care process shapes patient experiences? reflected support preferences during the set-up and monitoring period and the uncertainty regarding timely clinical and device feedback. Conclusions: Patients valued cardiac wearables for facilitating diagnosis and felt reassured knowing they were clinically monitored. However, gaps in information provided to patients seemed to cause anxiety for some participants. These concerns could be mitigated through clearer clinician communication and patient education at the time of prescription.

Abstract Background: Cardiovascular disease is the leading cause of mortality in Nigeria and across sub-Saharan Africa, with rising incidence attributable to urbanisation, sedentary lifestyles, and limited access to early detection tools. Concurrently, patient dropout from rehabilitation programs remains a critical operational challenge for Nigerian clinics, with many patients failing to return after their initial consultation. Methods: We developed CardioAI, an Explainable Artificial Intelligence system comprising two predictive modules. The cardiovascular risk module trained four machine learning models - Logistic Regression, Random Forest, Gradient Boosting (XGBoost), and a Multilayer Perceptron - on a combined UCI Heart Disease dataset of 1,025 patient records. A novel Lifestyle Risk Index was engineered from five modifiable clinical markers. SHAP (SHapley Additive exPlanations) was applied for per-prediction feature attribution. The patient retention module trained three classifiers on a synthetic dataset of 800 records, modelling 10 operational and behavioural dropout factors. An NLP and OCR pipeline using Tesseract v5.5 and spaCy was implemented for clinical document processing. Results: The cardiovascular risk module achieved an AUC-ROC of 0.999 (XGBoost), 0.998 (Random Forest), 0.994 (MLP), and 0.927 (Logistic Regression) on the held-out test set. Cross-validated AUC with constrained tree depth was 0.97, confirming generalisation. SHAP analysis identified the Lifestyle Risk Index, ST depression, resting blood pressure, exercise-induced angina, and cholesterol as the five most influential predictors. The retention module achieved AUC-ROC of 0.66 (Logistic Regression), demonstrating the difficulty of dropout prediction with synthetic data. Conclusions: CardioAI demonstrates that explainable machine learning can provide clinically actionable cardiovascular risk assessment and patient retention intelligence in a low-resource Nigerian healthcare context. The system is freely deployable, open-source, and designed for pilot validation in teaching hospitals across Lagos and Port Harcourt. Keywords: cardiovascular risk prediction, machine learning, explainable AI, SHAP, patient retention, clinical decision support, Nigeria, sub-Saharan Africa, XGBoost, random forest, digital health

Purpose: High clinical workload is associated with worse patient and hospital outcomes and is a well-established driver of clinician burnout. Trainees may be particularly exposed, shouldering both clinical and educational responsibilities. Evidence-based work design offers a data-driven approach to healthcare work but relies on robust workload measurements. Trainee workload remains poorly characterized, as commonly used metrics (e.g., duty hours, patient census) overlook cognitive and contextual dimensions. This pilot evaluated the feasibility of combining survey-based and electronic health record (EHR) data to characterize internal medicine (IM) trainee workload. Methods: A pilot study was conducted including IM and Medicine-Pediatrics residents (postgraduate years 1-4) between March 31 and June 22, 2025. Participants completed daily surveys during a seven-day inpatient schedule assessing workload and work experience domains, including environment, professional fulfillment, psychological safety, autonomy, and rounding experience, using validated instruments where available. Concurrently, EHR data captured chart review, documentation, orders, and secure messaging activity. Associations between survey and EHR data were assessed. Results: Among 37 eligible residents, 28 (76%) participated in the pilot capturing 166 shifts. Trainees spent 4.4 +/- 1.6 (mean +/- SD) minutes completing daily surveys and 8.6 +/- 2.3 minutes completing the final survey. Trainees reported working 11.6 +/- 1.0 hours/day and a median census of 9.0 (IQR 6.0-11.0). NASA-TLX score was 50.8 +/- 12.6. Positive shift ratings were associated with lower NASA-TLX scores and perceived rounding length. First-to-last EHR login duration was 15 +/- 2 hours/day, and EHR data showed 204 +/- 46 active minutes/day. Login duration correlated with self-reported hours (r=0.43, p<0.0001), and notes signed correlated with self-reported team (r=0.19, p=0.013) and personal census (r=0.34, p<0.0001). Conclusions: Integrating survey-based and EHR-derived workload measures provides multidimensional insight into trainee work. This novel approach supports scalable measurement and evidence-based work design interventions to improve trainee well-being, education, and clinical efficiency.

Study ObjectivesTo evaluate wearable sleep staging across sleep apnea severity, including very severe sleep apnea defined as an apnea-hypopnea index (AHI)[&ge;] 50 events/h, and to assess how training-set composition affects performance in this subgroup. MethodsWe analyzed 552 overnight recordings, 318 from the Sleep Lab Dataset and 234 from the Hospital Dataset. In the Hospital Dataset, 26.5% had very severe sleep apnea. We developed a deep learning model for sleep staging using RR intervals from wrist-worn photoplethysmography and three-axis accelerometry. Baseline performance was assessed by cross-validation under 5-stage and 4-stage staging. We examined night-level associations with AHI severity. We also compared the baseline model with an ablation model trained on the same number of recordings but with more Sleep Lab Dataset and lower-AHI Hospital Dataset recordings, evaluating both models in the very severe subgroup. ResultsIn 5-stage classification, Cohens kappa was 0.586 in the Sleep Lab Dataset and 0.446 in the Hospital Dataset. Under 4-stage staging, the gap narrowed, with kappa values of 0.632 and 0.525, respectively. In the Hospital Dataset, performance declined with increasing AHI severity. Among 62 recordings with very severe sleep apnea, reducing high-AHI representation in training lowered kappa from 0.365 to 0.303. ConclusionsWearable sleep staging performance declined across greater sleep apnea severity in this clinical cohort. Clinical utility may benefit from training data that better represent the target severity spectrum and from selecting staging granularity to match the intended use case. Statement of SignificanceRepeated laboratory polysomnography is impractical for long-term sleep apnea management. Wearable sleep staging could support scalable monitoring, yet its reliability in clinically severe sleep apnea has remained unclear. This study developed and evaluated a wearable sleep staging approach in both sleep-laboratory and hospital cohorts. The hospital cohort included many severe and very severe cases. Performance was lower in the hospital cohort and declined with greater sleep apnea severity. A coarser staging scheme reduced the gap between cohorts, and models trained without representative very severe cases performed worse in this target population. These findings highlight the value of severity-aware model development and motivate future multi-night home validation with reliability cues.

Objective: To develop the first inventory to measure psychosocial concerns about use of the non-preferred hand, toward the long-term goal of identifying the casual factors of left-right hand choices ("hand usage"). Design: Cross-sectional Setting: Online question battery Participants: 181 healthy adults Interventions; Not applicable Main Outcome Measures: Self-reported concerns about emotional and physical consequences of using the non-preferred hand. Results: Emotional and physical consequences reflected internally consistent categories (Cronbach's > 0.9) that were moderately correlated with each other ({rho} = 0.783 p = 0.002). Concerns were activity-dependent in each category ({rho} < 1x10-100). Reliability analysis and principal components analysis were used to reduce the battery to the 51-item Changed Hand Usage Concerns inventory, which encompasses everyday tasks and concerns about physical and emotional consequences of using the non-preferred hand in those tasks. Conclusions: Concerns about emotional vs. physical consequences of non-preferred hand use reflect coherent and internally consistent categories The Changed Hand Usage Concerns inventory allows assessment of psychosocial concerns about usage of the non-preferred hand for future attempts to manipulate hand usage via rehabilitation in patients with unilateral or asymmetric impairment.