Runze Yan

Objective: Complex predictive models are increasingly proposed as substitutes for laboratory measurements, yet they are typically assessed with aggregate accuracy metrics that reveal little about downstream clinical impact. We developed an evaluation framework that assesses a prediction model by its utility for guiding decisions under realistic resource constraints, demonstrated through non-invasive lactate estimation in the intensive care unit (ICU).Methods: The framework assumes the model is used to derive a resource allocation policy that selects which ICU patients to re-measure lactate and which to rely on the model's estimate, under a pre-specified budget for invasive blood draws. Policies are evaluated by average clinical cost of errors, area under the receiver operating characteristic curve (AUROC) for hyperlactatemia classification at the clinical threshold, and regression mean squared error (MSE). Beyond random allocation and prediction-ranked or threshold-distance heuristics, we propose Predictive Risk-Informed Sampling (PRIS), which estimates each patient's false-negative and false-positive risk with separate classifiers and ranks by a cost-weighted score, prioritizing draws for cases most vulnerable to harmful misclassification. On MIMIC-III (788 ICU stays) with a pretrained lactate predictor, we compared PRIS against heuristic baselines across budget levels.Results: PRIS significantly reduced average clinical cost versus random allocation at low-to-moderate budgets. All rules improve as measured values replace predictions, but PRIS reaches consequential errors sooner, resolving 22 of 61 misclassifications by 50 draws versus 14 for the prediction-ranked baseline, and achieving the highest classification performance at every budget. Because it ranks by error probability rather than predicted magnitude, PRIS led in clinical cost and classification but was least effective for regression error: the preferred rule depended on the evaluation metric, a distinction that aggregate accuracy would not reveal. At the budget of the routine draw-all-abnormal rule, prediction-ranked identified none of the 34 missed hyperlactatemia cases, whereas PRIS caught 14.Conclusion: Aggregate accuracy does not capture a model's clinical usefulness; evaluating decision utility under a measurement budget is a better alternative. As a triage layer on an existing lactate predictor, PRIS directs lab resources toward cases where inaccurate predictions carry the highest risk of harmful clinical consequences.

Impaired driving kills over 10,000 people in the US each year. Whether educational campaigns cause behavior change remains unclear because many evaluations rely on correlational rather than causal evidence. This study addresses a fundamental question in transportation safety: Do educational campaigns truly cause changes in impaired driving behavior? We applied advanced causal inference methods, Linear Non-Gaussian Acyclic Model (LiNGAM), and the DoWhy framework, to survey data from 876 respondents to identify genuine causal drivers of behavior change intentions. Our approach combined LiNGAM-based causal discovery to learn the complete directed acyclic graph among variables, followed by DoWhy’s systematic causal effect estimation with backdoor adjustment to estimate average treatment effects (ATE). Multiple refutation tests validated the robustness of our findings. The results showed that perceived campaign effectiveness (ATE = 0.264), concern about neighborhood driving under the influence incidents (0.254), and belief in sobriety measure effectiveness (0.195) emerged as significant positive causal drivers of behavior change likelihood. In contrast, traditional awareness-based interventions, including campaign exposure (−0.125) and enforcement knowledge (−0.101), demonstrated negative causal effects, challenging conventional assumptions about information-based prevention strategies. These findings suggest that successful impaired driving prevention requires shifting from information dissemination to strategies that actively cultivate risk perceptions and intervention efficacy beliefs. This study’s findings show how modern causal discovery and validation tools can provide actionable evidence for policymakers, demonstrating that campaigns perceived as credible and locally relevant drive meaningful behavioral shifts, while mere exposure to educational content may be counterproductive.

BACKGROUND Chronic disease management increasingly depends on monitoring symptoms and health behaviors outside clinical settings, yet most mobile health (mHealth) systems treat behavioral data capture and intervention delivery as separate functions. Platforms that collect high-resolution sensor or ecological momentary assessment (EMA) data often provide limited tools for clinician-authored intervention content, whereas intervention applications commonly embed content and decision logic directly in application code. This coupling slows iteration, limits personalization, and creates dependence on software developers. OBJECTIVE This paper presents CanBeWell, a disease-agnostic telehealth platform that integrates multimodal behavioral data collection with a clinician-operated, code-free intervention authoring and delivery workflow. CanBeWell consists of a patient-facing mobile application, a backend application programming interface (API) and data layer, and a clinician-facing dashboard. We describe the platform's system architecture, design rationale, implementation for cancer survivorship care, and formative stakeholder evaluation. METHODS CanBeWell was developed as a platform-oriented system informed by prior work in remote patient monitoring, mHealth sensing, and adaptive intervention theory. The platform was evaluated through formative semi-structured interviews with cancer patients (n=12) and their caregivers (n=9). [CX3.1]All study participants reviewed the mobile application prototype and provided feedback during in-person interviews on its usability, perceived value, and suggested improvements. Participants also rated the platform's overall usability on a 10-point scale. Additionally, clinicians (n = 6) were provided access to both the patient-facing mobile application and the clinician dashboard to evaluate the platform's usability, functionality, clinical relevance, and intervention workflow. Qualitative data were analyzed using thematic analysis by two independent coders RESULTS Twenty-one participants (patients and caregivers) completed the demographic survey (mean age 49.4 years; 61.9% female; 57.1% non-Hispanic White). Four themes emerged based on patient and caregiver interviews: ease of use, interest in the application, valued features, and overall user-friendliness[CX4.1]. The mean usability rating across patients and caregivers was 9 out of 10. Patients and caregivers valued the phased intervention structure, baseline activity tracking, and incremental goal progression. Clinicians responded positively to the no-code authoring workflow, noting that the ability to review behavioral data, draft content, and deploy approved phases without developer involvement could reduce technical barriers and accelerate intervention refinement[CX5.1]. The most common requested improvement by clinicians was stronger notification support to sustain engagement over longer deployments. CONCLUSIONS CanBeWell extends remote patient monitoring infrastructure by adding a clinician-governed intervention authoring and delivery layer that closes the loop between behavioral sensing and patient-facing support. Formative validation in a cancer survivorship context suggests that the platform is usable and clinically meaningful to patients, caregivers, and clinicians. By separating intervention content from application code, the platform enables clinicians to create and update personalized interventions without extensive programming support, reducing engineering bottlenecks and facilitating scalable implementation across chronic disease settings.

Objective: Accurate classification of physiological signals in real-world deployments is challenged by sensor noise, motion artifacts, and distribution shifts between training and deployment data. Inference-time augmentation (ITA), which applies augmentations during inference rather than retraining, offers a simple, model-agnostic mechanism to improve robustness. However, ITA application to physiological signals has remained narrow in scope, relying on limited augmentation methods with fixed, unoptimized parameters. This work proposes a unified ITA framework to address that gap. Approach: The framework incorporates 13 augmentation methods spanning time-domain, amplitude-domain, frequency-domain, and artifact-injection transformations, with hyperparameters systematically optimized via Bayesian optimization. We evaluate the framework on atrial fibrillation (AF) detection from 30-second photoplethysmography (PPG) signals using two deep learning architectures, GPT-PPG (in two sizes) and ResNet, across five datasets comprising more than 400 patients and ~9,800 hours of PPG recording. Two evaluation strategies are assessed: standard ITA applied to all inputs, and selective ITA applied to initially positive predictions. Main results: Standard ITA consistently improved the area under the receiver operating characteristic curve (AUROC, up to 8.5% for GPT-PPG and 0.7% for ResNet) and the area under the precision-recall curve (AUPRC, up to 10.6% for GPT-PPG and 0.8% for ResNet) across all model–dataset combinations. Selective ITA further reduced average FPR by up to 4.4% (GPT-PPG) and 1.3% (ResNet) on non-AF PPG datasets. Significance: These findings establish ITA as a practical, model-agnostic approach for improving the reliability of PPG-based AF classification in deployment settings where retraining is not feasible, with broader applicability to physiological signal analysis.

Background: Epigenetic age acceleration in head and neck cancer (HNC) patients undergoing radiotherapy has been linked to adverse treatment outcomes. Understanding a patient's likely epigenetic aging response prior to radiotherapy has the potential to inform healthcare planning and clinical decision‐making; however, current methods cannot effectively predict these changes throughout treatment, particularly without expensive assays of epigenetic alterations. Methods: This study developed and validated a machine learning framework to predict EAA following radiotherapy, utilizing pre‐radiotherapy (Time 1) sociodemographic information, symptom reports, clinical measurements, and immune biomarkers. Various machine learning models were explored to forecast EAA across three post‐treatment clinic stages: Immediately post‐radiotherapy (Time 2), six months (Time 3), and 12 months post‐radiotherapy (Time 4). Results: Our results demonstrate that: (1) deep learning methods, particularly TabNet, outperform conventional algorithms with an average RMSE of 4.08 (SD = 0.32); (2) predictions are most accurate immediately post‐treatment (Time 2 RMSE: 4.87); (3) baseline immune markers, including absolute eosinophil count and hemoglobin levels, are consistent predictors across timepoints; and (4) specific patient subgroups show distinct prediction accuracies (RMSE range: 1.70–4.34). Conclusion: These results suggest that pre‐treatment demographic and clinical data effectively predict post‐treatment EAA trajectories without expensive epigenetic assays, enabling cost‐effective early identification of high‐risk patients for potential targeted interventions before adverse effects manifest.

The ability to continuously and passively monitor human behavior using data from mobile and wearable devices in everyday environments presents new opportunities for tracking behavioral patterns that require detailed and long-term multimodal data. This paper introduces a dataset and corresponding methods for capturing and analyzing cyclical patterns of varying lengths from biobehavioral data collected passively through smartphones and wearable devices. The dataset includes up to 16 months of continuous records from smartphones and Fitbit devices, along with daily surveys from 166 university students. In addition to evaluating existing methods for modeling cyclical behavior, we also develop and present a new approach that facilitate multidimensional modeling and comparison of biobehavioral cycles within a population and across different time periods. We evaluate our methods using collected data to identify differences in cyclical behavioral patterns among various groups of students over different periods of the study.

Background Accurate dietary assessment is essential for health research. While smartphone-based food image recognition offers a convenient alternative to traditional methods, existing computer vision approaches struggle with real-world food images and analyze only basic macronutrients, limiting their utility for comprehensive nutritional research. Methods We developed DietAI24, a framework for automated nutrition estimation from food images that combines multimodal large language models (MLLMs) with Retrieval-Augmented Generation (RAG) technology to ground the MLLM’s visual recognition in authoritative nutrition databases rather than relying on the model’s internal knowledge. In our work, we used the Food and Nutrient Database for Dietary Studies (FNDDS) as the authoritative nutrition database. Through this approach, DietAI24 enables accurate nutrient estimation without extensive data collection or model training. Results DietAI24 significantly outperforms existing methods when evaluated against commercial platforms and computer vision baselines using the ASA24 and Nutrition5k datasets. Performance is measured through mean absolute error (MAE). DietAI24 achieves a 63% reduction in MAE for food weight estimation and four key nutrients and food components compared to existing methods when tested on real-world mixed dishes (p < 0.05). Notably, DietAI24 estimates 65 distinct nutrients and food components, far exceeding the basic macronutrient profiles of existing solutions. Conclusions DietAI24 demonstrates that integrating MLLMs with RAG and standardized nutrition databases can substantially improve dietary assessment accuracy while enabling comprehensive nutrient analysis. This framework offers a scalable solution for nutrition research and clinical applications, potentially transforming large-scale epidemiological studies and personalized dietary interventions through more accurate and less burdensome dietary data collection.

Digital health interventions offer promise for scalable and accessible health care, but access is still limited by some participatory challenges, especially for disadvantaged families facing limited health literacy, language barriers, low income, or living in marginalized areas. These issues are particularly pronounced for patients with colorectal cancer (CRC), who often experience distressing symptoms and struggle with educational materials due to complex jargon, fatigue, or reading level mismatches. To address these issues, we developed and assessed the feasibility of a digital health platform, CRCWeb, to improve the accessibility of educational resources on symptom management for disadvantaged patients with CRC and their caregivers facing limited health literacy or low income. CRCWeb was developed through a stakeholder-centered participatory design approach. Two-phase semistructured interviews with patients, caregivers, and oncology experts informed the iterative design process. From the interviews, we developed the following 5 key design principles: user-friendly navigation, multimedia integration, concise and clear content, enhanced accessibility for individuals with vision and reading disabilities, and scalability for future content expansion. Initial feedback from iterative stakeholder engagements confirmed high user satisfaction, with participants rating CRCWeb an average of 3.98 out of 5 on the postintervention survey. Additionally, using generative artificial intelligence tools, including large language models like ChatGPT and multimedia generation tools such as Pictory, complex health care guidelines were transformed into concise, easily comprehensible multimedia content, and made accessible through CRCWeb. User engagement was notably higher among disadvantaged participants with limited health literacy or low income, who logged into the platform 2.52 times more frequently than nondisadvantaged participants. The structured development approach of CRCWeb demonstrates that generative artificial intelligence–powered multimedia interventions can effectively address health care accessibility barriers faced by disadvantaged patients with CRC and caregivers with limited health literacy or low income. This structured approach highlights how digital innovations can enhance health care.

Objective. This study aims to introduce a novel generative pre-trained transformer (GPT)-based foundation model specifically tailored to photoplethysmography (PPG) signals, enabling effective adaptation to various downstream biomedical tasks. Approach. We adapted the standard GPT architecture to handle the continuous characteristics of PPG signals, leveraging extensive pre-training on a large dataset comprising over 200 million 30 s PPG samples, followed by supervised fine-tuning strategies for task-specific optimization. Main results. Our approach achieves performance comparable to or exceeding current state-of-the-art methods on various downstream tasks, notably atrial fibrillation detection, and demonstrates a unique generative capability, such as effective signal denoising, inherently available without additional fine-tuning. Significance. The successful adaptation of GPT to PPG signals underscores the potential of generative transformer frameworks as versatile foundation models in biomedical signal processing, highlighting their dual role in predictive and generative tasks.