Medical Decision Making

Background: Cost-effectiveness analyses (CEA) built on simulation models are usually delivered through project-specific scripts. Each new model, and often each new question, rebuilds the same scaffolding: parameter management, calibration, base-case analysis, sensitivity analysis and export. This is costly, and is where traceability between an estimate and the model version producing it is lost. We present THALASSA, a web application providing this infrastructure once, for any model satisfying a small programming contract. Methods: A simulation model is connected by one file implementing a fixed contract of R functions that determine its behaviour. THALASSA exposes the complete CEA workflow in a browser: a generated overview of the connected model, parameter editing, calibration, base-case comparison with the efficiency curve, and deterministic and probabilistic sensitivity analysis at a willingness-to-pay threshold, each with timestamped file export carrying the parameter values it was run with. Every session is bound to a repository version, so results carry their generating model version. Results: We demonstrate THALASSA on two public models of contrasting difficulty: a simple tutorial model and a generic cancer screening model with two competing natural-history pathways and a 21-dimensional calibration deliberately built to be more complex. Both ran on the same unmodified setup, implementing the same contract, with no code changes. Conclusions: Separating analysis infrastructure from model logic makes the CEA workflow reusable across projects and reproducible by construction. A modelling team can move from a versioned repository to audit-ready, exportable CEA outputs without writing analysis code. The integration effort does not grow with model size and one installation serves unrelated projects.

Personalized medicine aims to tailor treatments to individual patients, but its implementation is limited by scarce clinical resources and heterogeneous treatment effects. Electronic health records now offer longitudinal data that can support data-driven therapy selection. We focus on SGLT2 inhibitors, medications with broad metabolic and cardiovascular benefits, whose clinical impact varies significantly based on individual patient profiles. We propose a counterfactual Structural Causal Model (SCM) to estimate individual treatment effects (ITE) of SGLT2 inhibitors from observational data. The model assumes counterfactual stability and uses a Gumbel-Max sampling mechanism to capture stochastic treatment assignment while enabling differentiable inference. Optimal treatments are defined by maximizing the counterfactual value of future diagnostic outcomes. We evaluate the Counterfactual Gumbel-Max SCM using a doubly robust value estimator and show that, across diagnostic categories, it outperforms real-world prescriptions in predicted treatment value, highlighting subpopulations where SGLT2 inhibitors yield superior outcomes. By integrating causal inference with electronic health data, this framework provides individualized counterfactual reasoning about SGLT2 treatment effects and can support clinicians in identifying optimal therapies for type 2 diabetes and comorbid conditions.

Background: While health system costs of COVID-19 are well documented, evidence on out-of-pocket costs to patients and families, including non-medical and time burden, is limited. This study aimed to estimate out-of-pocket medical, non-medical, and time burden of COVID-19 illness directly from patients and caregivers. Methods: Two cross-sectional surveys were fielded to national nonprobability samples of U.S. adults in Fall 2023: a personal burden survey (adult sample) and a caregiver burden survey (child sample). Surveys collected data on care setting, visit duration including travel and wait time, and medical and non-medical out-of-pocket costs. After applying quality-based exclusion criteria, 409 adult sample and 52 child sample responses were analyzed. Results: Over 87% of respondents were insured. Median out-of-pocket medical costs per visit were low, ranging from $5 (emergency department) to $10 (urgent care) in the adult sample, and from $0 (emergency department) to $75 (hospitalization) in the child sample. Median non-medical costs were $28 or less across care settings. Median time spent receiving care ranged from 90 minutes (primary care, child sample) to 300 minutes (emergency department, adult sample). Despite the relatively low reported out-of-pocket costs, COVID-related financial challenges were reported by 34% of the adult sample and 29% of the child sample, with the most common difficulty being paying non-medical bills (11% and 15%, respectively). Twelve percent in both groups reported problems paying COVID-19-related medical bills. Conclusions: In this post-pandemic non-probability sample, individual out-of-pocket costs for COVID-19 illness were relatively modest. However, time burden was substantial and one-third of respondents reported financial hardship. These estimates, including non-medical and time burden not captured in administrative data, can support more robust economic evaluations of COVID-19 prevention and treatment interventions.

Objective: To estimate the impact of removing race-adjusted antihypertensive treatment recommendations for Black adults on health outcomes, healthcare utilization, and racial/ethnic disparities. Materials and methods: A discrete-time state-transition microsimulation model of hypertension natural history and health system utilization was developed to simulate lifetime trajectories among 100,000 non-Hispanic Black and non-Hispanic white 40-year-old adults. We evaluated the effect of removing race-adjusted antihypertensive treatment recommendations under two scenarios: one assuming racial heterogeneity in treatment effectiveness, reflecting evidence behind prior race-specific guidelines, and the other assuming no racial heterogeneity. Our main outcomes included life expectancy, cumulative incidence and duration of hypertension, and blood pressure control with antihypertensive treatment. Results: Under assumptions of heterogeneous treatment effectiveness by race, guidelines without race adjustment decreased life expectancy among non-Hispanic Black adults by 0.4 years, increasing the life expectancy gap with non-Hispanic white adults from 4.1 to 4.5 years. Hypertension control among non-Hispanic Black adults also worsened. Years with uncontrolled hypertension increased by 0.8 years, widening the disparity with non-Hispanic white adults from 1.7 to 2.5 years. Assuming no racial heterogeneity, these effects were diminished: life expectancy decreased by 0.1 years and years with uncontrolled hypertension increased by 0.2 years, less than one-half and one-quarter of the effects observed under assumptions of heterogeneous treatment effectiveness by race. Discussion: The potential effects of removing race-adjusted antihypertensive treatment guidelines depended on assumptions of racial heterogeneity in treatment effectiveness. Conclusion: Our findings highlight the importance of understanding drivers of previously observed racial differences when considering modifying antihypertensive treatment guidelines.

Importance: The Medicare Physician Fee Schedule values ambulatory services using survey-based estimates of physician time, a costly process vulnerable to inaccuracy. EHR audit logs are passively and universally collected and likely meet the American Medical Association Relative Value Scale Update Committee (AMA RUC) Extant Database Standards for alternative data to inform service valuation estimates. However, audit logs' validity for estimating face-to-face time, a core component of many of the most used procedural codes, is unknown. Objective: To assess whether EHR audit logs can facilitate accurate estimation of physician-patient face-to-face time during ambulatory visits. Design: Single-site validation study pairing EHR audit logs with direct audio/video recordings of ambulatory primary care visits (reference standard). We assessed agreement between audit-log estimates and recorded visit duration with Bland-Altman analysis and mean absolute error. Setting: University of Pennsylvania Observer data repository, a multi-modal database of ambulatory visits to Penn Medicine primary care clinics including video and audio recordings, transcripts, clinical EHR data, and EHR audit-log data. Participants: 75 visits from 12 physicians with data collected between August 2023 and February 2025. Exposures: A rule-based algorithm estimated face-to-face time via clusters of physician EHR activity relative to patient check-in time. Inactivity thresholds were applied to differentiate intraservice time with no EHR activity (e.g., a physical exam) from the end of a visit, with longer thresholds used for longer visits. Main Outcomes and Measures: Duration of recorded face-to-face visit time. Results: Recorded face-to-face time averaged 31.6 minutes (SD, 13.9). The audit-log estimator was unbiased (mean bias: -0.63 minutes; 95% CI: -6.0 to 2.6; mean absolute error: 9.0 minutes) and estimated face-to-face time within approximately 2% (38 seconds) of the recorded visit time with no proportional bias (slope: 0.13; P = .25). The 95% limits of agreement were -25.7 to 24.4 minutes, and 37% of estimates fell within 10% of recorded duration. Conclusions and Relevance: In this validation study, EHR audit logs recovered unbiased estimates of ambulatory face-to-face time. Larger-scale studies are required to verify generalizability, but EHR audit logs offer a scalable means to improve the rigor of retrospective physician work relative value unit estimation. This study represents a first step in a potential paradigm shift for measuring components of the fee schedule across procedures, settings, and provider types.

As health AI tools enter clinical spaces, how they are perceived by providers in safety-net environments shapes whether their benefits reach underserved populations. We examined the degree and drivers of healthcare providers' trust in health AI, whether that trust moderates perceived benefits for safety-net populations, and how providers balance benefit against concern across four specific hypothetical AI tools. We used a two-study mixed-methods design: 18 semi-structured interviews with healthcare practitioners and research leaders, followed by a survey of 229 Texas healthcare providers incorporating four vignettes describing hypothetical AI tools, each rated for appeal and concern and followed by open-ended explanation. Responses were thematically coded and stratified by self-reported trust in health AI (47.6% high trust, 52.4% low trust). Low-trust respondents most often cited building trust (50.5%) and data bias and accuracy (33.0%) as drivers of their current trust level, while high-trust respondents more often cited efficient and effective care (23.5% vs. 3.7% low-trust, p<0.001) and accuracy of data (11.8% vs. 1.8% low-trust, p=0.004); concerns about data security, governance, and the need for human oversight were raised at similar rates by both groups. When asked how AI could specifically benefit safety-net populations, both groups converged on expanding access to care (45.9% each), and none of 25 benefit-related codes differed significantly by trust group. Across all four vignettes, mean appeal ratings exceeded mean concern ratings, and high-trust providers rated every tool as more appealing and, for three of four tools, less concerning, without changing which tool ranked where relative to others. Providers exhibit conditional optimism: support is contingent on governance, human oversight, and context-sensitive implementation rather than on the technology itself. Realizing AI's promise for safety-net populations will require human-in-the-loop models, localized implementation strategies, community-informed design, and engaging providers directly in policy design and tool development.

Dynamic systems often evolve faster than reliable measurements can be collected. More frequent indirect observations may fill these gaps, but their noise and differing scales complicate estimation of the underlying state. We developed Bayesian Anchored Latent Learning (BALL) to combine sparse reference measurements with denser records on an interpretable scale. A bidirectional transformer teacher reconstructs trajectories from complete training histories, and a forward-only transformer student learns from those reconstructions and observed measurements. Across 30 simulations, BALL reduced composite trajectory error by 7% relative to the same transformer trained directly from measurements, achieving lower error across all datasets. In an applied example using psychiatric records from 1,609 patients, BALL improved the prediction of withheld questionnaire scores compared with direct transformer training at 21- and 28-day measurement schedules. Requests guided by estimated uncertainty further improved prediction. BALL connects retrospective learning with prospective state estimation and provides a framework for directing additional measurement.

We develop an age- and immunity-structured COVID-19 model with setting-specific contacts, calibrated to Ontario incidence and hospital admissions before Omicron. We simulate counterfactual vaccination strategies and contact restrictions in schools, workplaces, and other community settings, evaluating projected severe-case demand against an illustrative healthcare-capacity benchmark. Vaccination combined with workplace and community contact restrictions keeps simulated demand within capacity. In contrast, relaxing restrictions in either setting produces large surges and repeated capacity threats, even when school restrictions remain. School-only changes produce smaller effects on severe disease. These findings highlight the value of vaccination and setting-specific contact restrictions that prioritize workplace and community settings for pathogens with adult-concentrated severe outcomes.

People experiencing homelessness (PEH) have substantially lower quality-adjusted life expectancy and substantially higher healthcare spending compared to housed populations. California has the largest population of PEH in the United States, and an estimated ten percent of PEH have co-occurring opioid use disorder (OUD). We developed a lifetime, individual-level microsimulation of the 18,928 PEH with OUD in California to evaluate the health and economic effects of expanding access to housing through a Housing First approach, which does not require OUD treatment or abstinence. Increases in permanent housing and shelter access were based on Los Angeles County's Measure A goals, and we accounted for both housing and healthcare costs. Expanding housing access increased mean age at death by 0.34 years, reduced fatal overdoses for the cohort by 39, and produced an incremental cost-effectiveness ratio of $31,083 per quality-adjusted life-year gained, well below conventional willingness-to-pay thresholds. We provide evidence that housing is a high-value health intervention for PEH with OUD in California. Policymakers should support Housing First programs as part of integrated responses to the intertwined crises of homelessness and opioid use.

Biomedical Machine Learning (ML) is increasingly evaluated through Independent and Identically Distributed (IID) benchmarks, even when deployment involves irregular observation, missingness, distribution shift, and high failure costs. This position paper argues for a stability-oriented reporting standard for safety-relevant biomedical ML. The central requirement is not a specific model class or optimizer, but auditable evidence: performance should be reported under clinically plausible perturbations such as thinning, timestamp jitter, bursty missingness, channel dropout, and Signal-to-Noise Ratio (SNR) shift. When authors invoke training-stability diagnostics, those diagnostics should be hazard-linked: their assumptions, operational meaning, and relationship to stress-test degradation should be stated explicitly. We use controlled mechanism probes to illustrate two hazards that IID scores can hide: observation-process dependence under irregular sampling, and perturbation sensitivity associated with high-curvature training regimes such as Edge of Stability (EoS). We do not propose Neural ODEs, continuous-time models, Sharpness-Aware Minimization (SAM), or EoS metrics as default prescriptions. Instead, we propose a minimum reporting standard that makes robustness claims testable and supports escalation to more complex modelling or optimization only when stress tests justify the cost.

Background Pediatric neurocritical care (PNCC) has emerged as a distinct field, and many hospitals have moved to establishing dedicated PNCC programs. How practice change is implemented within these programs is not well understood. This study had two objectives. First, we sought to understand the barriers and facilitators of practice change in PNCC from the perspectives of the multiprofessional clinicians who deliver it. Second, we sought to characterize how well the integrated Promoting Action on Research Implementation in Health Services (i-PARIHS) implementation determinant framework captures them. Methods We conducted semi-structured interviews with a purposive sample of clinicians involved in PNCC at three U.S. academic medical centers. The interview guide was developed using i-PARIHS to organize barriers and facilitators of practice change across the innovation, recipients, context, and facilitation constructs. We applied framework analysis, combining deductive i-PARIHS, study-specific, and inductively derived categories. Two coders independently indexed the transcripts and reconciled them to consensus, with first-author adjudication. Preliminary site-level findings were validated with participants. Results We interviewed 33 clinicians: 24 physicians, 7 nurses, and 2 advanced practice providers. Most determinants of practice change mapped to existing i-PARIHS constructs, but three were not adequately captured by the framework as currently specified. The most consequential was an interprofessional authority gradient that determined whose preferences prevailed when intensive care, neurology, and neurosurgery disagreed. The other two were clinician autonomy, which served as the lens through which the acceptability of a change was judged, and the affective state clinicians bring to practice change. Receptivity depended more on recipients' values, prior experience, and capacity to absorb change into a demanding workflow. Facilitation was variable, frequently under-resourced, and dependent on credibility across professional groups. We translated these findings into three propositions linked to implementation strategies and outcomes. Conclusions Participants attributed practice change in PNCC more to the recipients of change, the multiprofessional context, and the availability of facilitation than to the strength of evidence for a given practice. i-PARIHS captured most barriers and facilitators well but may be adapted in this setting. Credible, resourced, authority-sensitive facilitation may improve the prospects for sustained practice change as institutions develop PNCC programs.

Background: A new class of AI "decision models" returns a probability for every option in a single pass and is marketed as calibrated, which invites use in clinical triage. Whether those probabilities hold up on clinical diagnosis has not been independently tested. Objective: To test the diagnostic accuracy, calibration, and probabilistic coherence of a decision model (Jev 1.13, TypeSafe AI) against frontier language models, with and without reasoning. Methods: In a pre-registered study (OSF, registered 26 September 2026), we compared Jev with Claude Opus 5.5 (adaptive reasoning, always on) and Claude Opus 5 (reasoning disabled) on three case sets: the 45 Semigran symptom-checker vignettes, 1,000 stratified synthetic patients from DDXPlus with reference differentials, and 80 simulated ophthalmic complaints (40 diagnoses at two levels of detail). Each model returned a probability for every diagnosis on a fixed list. Outcomes were top-1 and top-3 accuracy, expected calibration error (ECE), Brier score, the rate of exactly zero probability on the true diagnosis, sensitivity to option order, behavior on uninformative inputs, and behavior when the true diagnosis was absent. Results: On DDXPlus, top-1 accuracy was 72.0% for Jev, 79.5% for Opus 5 without reasoning, and 87.4% for Opus 5.5 (both comparisons with Jev P<.001). Jev assigned exactly zero probability to the true diagnosis in 62 of 1,000 cases (6.2%); neither comparator did so in any case. Calibration error did not differ between Jev and the registered comparator (ECE difference 0.007, 95% CI -0.026 to 0.038). Offered an explicit "none of these" option, Jev declined to diagnose all 30 uninformative inputs; without that option, it assigned a mean 58% to a single diagnosis. Order sensitivity was small in all models. Given the option list with no patient information, Jev placed up to 96% on a single diagnosis, while the comparators spread probability almost evenly. On the ophthalmic complaints, accuracy did not differ (Jev 88.8%, Opus 5 91.2%, Opus 5.5 88.8%), and Jev was better calibrated than the registered comparator (ECE difference -0.081, 95% CI -0.124 to -0.018). Conclusions: The decision model matched frontier models on short, distinctive ophthalmic complaints but fell well behind on multi-symptom synthetic patients, including behind a comparator without reasoning, and there it excluded the correct diagnosis outright in about 1 in 16 cases. Its probabilities signaled uncertainty only when the question offered a way to abstain. Calibration claims for clinical decision models should be tested on clinical tasks, with explicit abstention options and zero-probability rates reported.

Voxel-level prediction of treatment response from longitudinal fluorodeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT) can enable spatially adaptive dose planning in advanced non-small cell lung cancer (NSCLC), but point predictions alone lack the uncertainty quantification needed for safe clinical decision-making. Standard conformal prediction (CP) provides distribution-free coverage but yields spatially uniform intervals and assumes exchangeability, an assumption violated by intra-tumoral spatial correlation. We propose a multiscale CP framework built on a re-optimized VoxelForecast generalized least squares model that explicitly captures voxel-level covariance. Variogram sensitivity analysis identified the Stable model as significantly superior to the previously used Matern family. The framework combines hierarchical nested leave-one-patient-out calibration, which restores exchangeability for valid voxel-level coverage, with residual-variance CP (VarCP), which learns a heteroscedastic, location-dependent scale to produce adaptive intervals while preserving finite-sample marginal coverage guarantees. On two prospective NSCLC trial cohorts, FLARE-RT (N=25 locally advanced, chemoradiotherapy; 11,100 voxels) and BRIGHT (N=19 metastatic, chemoimmunotherapy; 26,980 voxels), multiscale feature integration reduced voxel-level prediction error by 35% and 31%, respectively. VarCP maintained nominal coverage while producing 7-18% (FLARE-RT) and 4-8% (BRIGHT) narrower voxel-level intervals than a hierarchically calibrated standard CP baseline (paired Wilcoxon signed-rank, significant at most miscoverage levels), with comparable lesion-level gains (7-18% and 6-10%) via median-based aggregation of voxel residuals. VarCP delivers tighter yet mathematically guaranteed spatial uncertainty maps, supporting both voxel-level dose adjustment and whole-lesion treatment planning in advanced NSCLC.

Mechanistic models of infectious disease transmission provide a framework for understanding and predicting disease dynamics, but statistical inference is often computationally prohibitive. This challenge is especially pronounced for models with high-dimensional structures and many unknown parameters, where trajectory matching may require thousands of evaluations of an expensive likelihood function. We developed a surrogate-assisted optimization framework that uses Bayesian optimization to reduce the number of evaluations of the underlying mechanistic model required for parameter inference. We evaluated the approach using simulated fitting problems based on age-structured susceptible-exposed-infected-recovered models with 4, 8, and 16 age classes and a 210-compartment age-structured rotavirus transmission model. Applying surrogate-assisted optimization substantially reduced the number of likelihood evaluations required to identify good parameter estimates, with larger improvements as model dimensionality increased. In the rotavirus model, the surrogate Bayesian optimization approach reached the benchmark after approximately 0.7 hours, compared with approximately 9.4 hours using traditional Nelder-Mead optimization. These results demonstrate that surrogate-assisted optimization can substantially reduce the computational burden of trajectory matching for complex mechanistic transmission models, particularly for higher-dimensional inference problems.

Abstract Background Community-based exercise may support physical and cognitive recovery after stroke, but the preferences of stakeholders involved in receiving, supporting and delivering these programmes are poorly understood. This study piloted a discrete choice experiment (DCE) that aimed to evaluate stroke survivors, carers, clinicians, and fitness instructors preferences for community-based exercise programmes for stroke survivors. The pilot aimed to assess the feasibility of conducting and analysing data for a larger-scale study and inform the sample size for a definitive DCE. Methods A DCE design was developed using literature reviews, stakeholder consultations, and evaluations of current services. Six attributes were included in the design: location, price, session frequency, individualisation, carer involvement, and recovery. Sixteen Bayesian D-optimal choice sets were developed, then blocked into two sets of eight to reduce cognitive burden. Main-effects were analysed using a mixed logit model, and stakeholder differences were explored using a conditional logit interaction model. Willingness to pay was also calculated. Results Overall, 161 respondents started the survey and 106 (66%) fully completed it. Stakeholders significantly preferred programmes that were delivered at a leisure centre, lower priced, individualised, had carer-specific sessions and offered greater chances of recovery. The largest willingness-to-pay estimates were {pound}49.41/ month for 80% chance of meaningful recovery vs 20%, and {pound}18.61/ month for leisure centre delivery vs online. Significant preference heterogeneity was identified in the main-effects model. Exploratory stakeholder interactions suggested clinicians were more price sensitive and placed greater importance on recovery and carer-specific sessions than stroke survivors, while fitness instructors placed greater importance on programme location and recovery. Carer interaction estimates were unstable due to the small subgroup sample. Conclusions The pilot supported the feasibility of the DCE design and analytical approach but identified limitations in the recruitment methods and survey completion. Addition of face-to-face recruitment methods are recommended for a larger-scale study. A larger sample size is required to investigate differences in preferences between stakeholder groups.

In many natural dynamical systems, tipping points emerge where slowly changing external factors induce shifts from one state to another, known as critical transitions or bifurcations. In epidemic systems, detecting these transitions is important for outbreak preparedness. Various statistical indicators like lag-1 autocorrelation and variance have been employed as early warning signals (EWS) of critical transitions by showing how the dynamics of the system exhibit critical slowing down (CSD) near the transition. More recently, deep learning-based algorithms have been developed that attempt to more accurately detect and classify bifurcations; however, they are limited by the systems simulated in the training dataset, the length of the time series, and have not been rigorously evaluated on real-world epidemic data spanning multiple geographic regions. In this paper, we evaluate existing deep learning-based bifurcation prediction models, on a CDC-curated dataset containing weekly influenza hospital admissions across all fifty U.S. states and the nation as a whole. We develop a pipeline for data preparation, reproduction number (Rt) estimation using the EpyEstim package (a Python implementation of the EpiEstim framework [4]), and bifurcation point identification. We systematically evaluate the models under several windowing strategies, including rolling, expanding, and fixed-length windows, and report accuracy, precision, recall, F1 score, and specificity. Our best-performing configuration (expanding window, length 100) achieves an accuracy of 90.09%, precision of 72.86%, recall of 96.23%, and F1 score of 82.93%. We additionally discuss Markov Regime Switching models as an alternative framework for detecting epidemic transitions. Our findings indicate that these models can generalize to real-world, state-level influenza data and are capable of online/real-time prediction.

Accurate prediction of individual medical outcomes is essential for optimizing treatment allocation amid rising costs, coverage denials, and limited clinical resources. Traditional predictive models, including linear regression and neural networks, rely on average effects and cannot tailor predictions to the specific circumstances of individual cases. We present relevance-based prediction (RBP), a model-free and task-specific method that predicts outcomes as weighted averages of observed cases, with weights determined by a rigorously defined and theoretically justified measure of relevance. Unlike model-based methods that rely on fixed calibrated parameters, relevance-based prediction revisits the original data for each prediction and customizes both the cases and variables used. We illustrate RBP by applying it to predict opioid treatment outcomes and demonstrate that it provides case-specific insights unavailable from conventional models, including how each prior case informs a prediction, how each variable affects each predictions reliability and value, and how reliable each prediction is before it is made. These individualized insights may prevent misleading average-based decisions and reduce harmful or suboptimal treatment.

Host immunity acquired through prior infections is a key factor shaping the spread of emerging variants. However, accounting for complete infection histories in mathematical models is challenging because the number of possible histories grows exponentially with the number of variants. Here, we present a flexible framework for modeling the co-circulation of multiple variants emerging at arbitrary times while incorporating cross-immunity and variant-specific characteristics. We introduce three tractable approximations: truncation, which retains only infection histories up to a given number of strains; one variant at a time, which is exact when variants circulation does not overlap; and two variants at a time, which is exact when at most two strains co-circulate. We systematically evaluate these approximations across a range of cross-immunity levels and inter-variant emergence delays. Truncation approximations perform best under short emergence delays and strong cross-immunity, whereas the other approaches outperform them at longer delays. When applied to SARS-CoV-2 variant dynamics, the second-order truncation and the two variants at a time approximation accurately capture incidence patterns, whereas the other approaches perform well only for specific subsets of variants. These results identify the parameter regimes in which simplifying assumptions remain valid, provide guidance for model selection, and highlight the key role of infection history in shaping epidemic trajectories.

Background: Quantifying behavioral drivers of protective action during infectious disease outbreaks remains challenging. App-based experimental epidemic games (epigames) provide a framework for studying these decisions in naturalistic settings, but correspondence between in-game measures and real-life beliefs or behaviors requires evaluation. We examined whether an activity-based opportunity cost influenced voluntary quarantine decisions within an epigame and whether real-life health beliefs transferred into the game and predicted observed choices. Methods: We conducted a two-week randomized controlled trial from May 4 to May 17, 2026, at VinUniversity (VinUni) in Hanoi, Vietnam. Participants using the Epigames app installed in their smartphone were randomized to a low-barrier group, earning 1 point per qualifying contact with other participant when remaining active, or a high-barrier group, earning 2 points per contact. Of 439 enrolled participants, 283 were successfully randomized to one of the two groups, made at least one valid daily quarantine decision, and their data was used in the analysis. Real-life (S1) and in-game (S2) beliefs measured susceptibility, severity, quarantine self-efficacy, and perceived benefits. Belief correspondence was assessed using Holm-adjusted Spearman correlations and prospective cumulative-logit models. Quarantine choices were analyzed using participant-day binomial generalized estimating equations (GEE) and participant-level quasi-binomial models. Results: Participants made 1,823 valid daily decisions and chose quarantine on 458 participant-days (25.1%). Quarantine rates were 30.1% in the low-barrier group and 20.2% in the high-barrier group. The high barrier was associated with 41% lower population-averaged odds of quarantine (GEE odds ratio [OR] 0.59, 95% confidence interval [CI] 0.40-0.86, = 0.0059); the quasibinomial estimate was identical (OR 0.59, 95% CI 0.40-0.86, = 0.0055). Among 251 participants with matched S1-S2 data, all four corresponding beliefs were positively associated (Spearman's = 0.259-0.418; all Holm-adjusted < 0.001). Prospective ordinal models confirmed positive matched-domain associations (ORs 1.39-2.18 per one-point S1 increase). Beliefs showed little robust association with quarantine behavior, and neither S1 nor S2 beliefs jointly modified the randomized effect. Self-efficacy, network degree, and the preregistered gender main effect were also unsupported. Conclusion: A higher activity-based opportunity cost reduced voluntary quarantine within the VinUni epigame, while participants carried relative differences in stated health beliefs from real-life to in-game framing. These beliefs did not reliably predict observed quarantine choices within the game framing. Epigames can support controlled study of incentives and belief correspondence in naturalistic settings such as college campuses, but behavioral validity or real-world generalizability requires further work.

Objective: We explored the implementation considerations of machine learning (ML) suicide risk identification in a health system that serves primarily American Indian patients in the Southwestern United States. Previous work developed and reported the performance of suicide risk models, but traditional performance metrics do not directly reflect clinical workflows. Materials and Methods: Using EHR data collected for Emergency Department (ED) visits between 10/01/2019 and 10/02/2021, we conducted Decision Curve Analysis (DCA) to identify the standardized net benefit (SNB) of all models previously evaluated and plotted the potential workload implications of the model with the greatest SNB. Finally, we evaluated suicide screening practices during the study period to inform implementation barriers. Results: Among 9,244 patients, 139 (1.5%) had a suicide attempt or death within 90 days of an included visit. The population-specific logistic regression model had the greatest SNB compared to other approaches. When operationalized, the 95th percentile cutoff did not substantially increase clinical workloads, while the 75th percentile cutoff substantially impacted workloads. The secondary analysis on suicide screening depicted variations in screening administration over time. Discussion: While DCA identified advantages of ML suicide risk identification over current standards, the complexities of suicide-related care limited its clinical interpretability. A stepped-care framework may mitigate potential workload burdens. Training initiatives and dynamic implementation strategies will support high-quality implementation of ML suicide risk identification. Conclusion: Our analyses support the feasibility and utility of augmenting existing suicide screening practices with ML suicide risk identification, and demonstrate the importance of nuanced analyses to inform implementation considerations.