JAMA Network Open

Health systems increasingly record ICD-10 Z59 social-needs codes for housing, food, transportation, and related hardship to trigger referrals, report quality, and adjust risk. However code documentation occurs at varying rates and no study has separated what the code says about a patient's hardship from what it says about being documented at all. Using a linked electronic health record and claims corpus covering 4.95 million patients in 21 identifiable health systems, we matched 26,153 adults with a first outpatient Z59 code to similar uncoded patients and, as a negative control, repeated the entire design around an administrative code. Coding ranged from 1 percent to 24 percent of patients across systems and remained highly variable after adjustment for patient mix. A hardship code was associated with 67% (95% CI=1.61, 1.74) more emergency department (ED) visits and 43% ( 95% CI=1.38, 1.49) more hospital admissions in the next year (62 versus 49 of every 100 patients with any acute event). The administrative-code reproduced 46 percent of the Z59 association for ED visits and 70 percent for admissions. In the five systems with the most complete stay dates, no hardship-specific admission excess was detectable. Homelessness codes forecast two and a half times the ED visit rate. At national planning prices, excess emergency and preventable hospital care totaled roughly $2,000 to $2,500 per coded patient-year ($2,516 base scenario; about $1,100 under an exploratory assumed-discount scenario (not an estimate of hardship-attributable spending)), with the largest emergency burden among Medicaid patients. The Z code added no increment to clinical and utilization-based risk prediction models in any system.

Abstract Importance Patients with behavioral health presentations leave the emergency department (ED) before evaluation more often than other patients, but outcomes after leaving without being seen (LWBS) have not been described in this population. Objective To compare return visits, admission, and mortality after LWBS vs completed care among behavioral health ED episodes. Design, Setting, and Participants This retrospective episode-level cohort study used a multisystem deidentified electronic health record database of US health systems (data snapshot June 4, 2026). Participants were patients of all ages with an ED visit and same-day behavioral health diagnosis. Index visits (recorded January 26, 2001 to April 13, 2025) each opened a 30-day episode window; in-window were counted as outcomes, and patients could contribute episodes to either group. Data were analyzed in October 2026. Exposure LWBS (diagnosis code Z53.21 on a date with an ED encounter) vs completed care ending in discharge home. Main Outcomes and Measures The primary outcome was ED return visit within 7 days. Secondary outcomes were return visit within 3 and 30 days; admission on day 0, on days 1-7 (principal measure), and within 30 days; and all-cause death within 7 and 30 days, assessed among episodes from 2017 through 2023 using a verified death-record definition. Results The analysis included 1,305,812 episodes (12,248 LWBS; 1,293,564 completed care) from 525,490 patients; 666,930 episodes (51.1%) involved female patients and 66,733 (5.1%) involved patients younger than 18 years. Return visit within 7 days followed 24.5% of LWBS episodes vs 13.2% of completed-care episodes (risk ratio [RR], 1.86; 95% CI, 1.81-1.92), and admission on days 1-7 followed 17.1% vs 3.9% (RR, 4.41; 95% CI, 4.24-4.59). Among 2017-2023 episodes, 30-day mortality was 0.69% (70 of 10,141) after LWBS vs 0.32% (2,426 of 756,140) after completed care (RR, 2.15; 95% CI, 1.70-2.73). Results were similar under alternative episode definitions. Conclusions and Relevance Among patients with behavioral health ED presentations, LWBS was followed by higher rates of return visits, admission, and short-term mortality than completed care. These unadjusted findings describe outcome burden, not causal effects, and highlight patients who leave as a target for rapid follow-up.

Importance: Clinicians often look for a dental source in Staphylococcus aureus bacteremia (SAB), but how often dental disease is found, compared with other bacteremias, is unclear. Objective: To compare documented dental disease in adults with SAB, Escherichia coli bacteremia, and a general population cohort. Design, Setting, and Participants: Retrospective cohort study in the Protege database of 48 975 adults with bacteremia from January 1, 2015, through December 31, 2023, assigned by earliest in-window organism, and 81 945 general population adults. Exposures: SAB, with E coli bacteremia as the primary comparator and the general population as the secondary comparator. Main Outcomes and Measures: Documented dental disease from 7 days before through 30 days after index, defined by a qualifying dental procedure or disease-specific note text and adjusted for age, sex, race, and documentation intensity. Results: Dental disease occurred in 528 of 16 435 patients with SAB (3.21%) and 179 of 24 942 patients with E coli bacteremia (0.72%). The adjusted odds ratio was 3.86 (95% CI, 3.23 to 4.63). By the pathology-only definition, disease occurred in 517 patients with SAB (3.15%) and 399 general population patients (0.49%). The adjusted odds ratio was 3.71 (95% CI, 3.08 to 4.46). Within SAB, rates were highest with prior infective endocarditis (9.8%), opioid use disorder (6.2%), injection drug use (5.3%), and prosthetic valve (5.0%). Prevalence ranged from 0.49% in the general population to 7.98% in other streptococcal bacteremia. Of 437 patients with SAB and an assignable first finding date, 333 (76.2%) had the finding after the index date. Conclusions and Relevance: Dental disease was documented more often in SAB than in E coli bacteremia after adjustment. Most findings were first recorded after bacteremia was recognized, during SAB evaluation. These data do not show that dental disease caused the bacteremia.

Objective To evaluate the association of early addiction consultation with withdrawal pharmacotherapy and complications in hospitalized patients with alcohol withdrawal. Methods Retrospective single-center cohort of adult hospitalizations with alcohol withdrawal (diagnosis code or CIWA-Ar [≥]10 within 24 hours) from July 2020 to January 2026. Exposure was an addiction consult service order within 24 hours of arrival, compared with no or later consultation. Primary outcomes were intensive care unit (ICU) transfer after the first 24 hours and delirium tremens or withdrawal seizure not present on admission. Process outcomes included phenobarbital and thiamine administration within 24 hours and benzodiazepine dose in the first 48 hours. Secondary outcomes included length of stay and 30-day death or return to hospital. Results Of 4,663 hospitalizations, 1,379 (29.6%) received early consultation. After adjustment, early consultation was associated with lower odds of ICU transfer (odds ratio [OR] 0.73; 95% CI 0.54-0.99) and fewer episodes of delirium tremens (risk difference -1.09 per 100; -1.45 to -0.73), with no significant difference in withdrawal seizure. Early consultation was associated with phenobarbital (OR 1.57; 1.32-1.86) and thiamine (OR 1.23; 1.07-1.41) within 24 hours, higher 48-hour lorazepam-equivalent dose (ratio of geometric means 1.44; 1.32-1.57), shorter stays (ratio 0.92; 0.88-0.97), and lower odds of 30-day death or return to hospital (OR 0.78; 0.66-0.91). Conclusions Early addiction consultation was associated with earlier, more intensive withdrawal pharmacotherapy, lower odds of ICU transfer, and fewer episodes of delirium tremens, although the complication associations were sensitive to cohort definition.

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.

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.

Background. Fluid collections are common after kidney transplantation, but their prevalence at scheduled imaging and their management are inconsistent from canter to canter. We describe the cross sectional prevalence and management of perigraft collections, including an exploratory deepdrain subgroup. Methods. In this retrospective single center cohort, 146 consecutive adult kidney transplant recipients (Cleveland Clinic Weston Hospital, July 2024 to June 2025) were imaged on postoperative day 0/1 and at approximately 1 and 3 months. Routine deep drainage was avoided (n=124), 22 clinically selected recipients with perioperative anticoagulation received a deep drain. The primary outcome was cross sectional prevalence of a perigraft collection at each visit. Secondary outcomes were management among patients with a collection. Drain group comparisons were exploratory. Results. Collections were present in 29 of 146 recipients at postoperative day 0/1 (19.9%; 95% CI, 14.2 to 27.1), 71 of 146 at 1 month (48.6%; 95% CI, 40.7 to 56.7), and 34 of 146 at 3 months (23.3%; 95% CI, 17.2 to 30.8). Prevalence in the no drain group was 21.0%, 50.0%, and 23.4%, respectively; corresponding prevalence in the deep drain group was 13.6%, 40.9%, and 22.7%. Between group prevalence comparisons were not statistically significant at any time point. BMI was higher in the deep-drain group than in the no-drain group (30.04 vs 27.1 kg/m2; P=0.01). Among patients with a 1-month collection, management did not differ significantly between groups (any invasive management, 11.1% vs 16.1%; P=1.00). Among patients with a 3month collection, invasive management was more common in the deep drain group than in the no drain group (80.0% vs 13.8%; P=0.007). This comparison included only five deep-drain recipients and was confounded by clinical selection, perioperative anticoagulation, and BMI imbalance. Conclusions. Perigraft collection prevalence was highest at 1 month and lower at 3 months in this predominantly no drain cohort. Most collections were managed conservatively, and invasive treatment was uncommon in the no-drain group. The deep drain subgroup provides clinically relevant information about high risk recipients selected for drainage but cannot support causal conclusions regarding drain benefit or harm. Patient level linkage is required to distinguish persistence, resolution, and new collection formation.

Importance: Pediatric obesity affects 1 in 5 US children and has been linked to brain structural differences. Most evidence is cross-sectional and models body mass index (BMI) linearly, with metrics that compress severe obesity. Objective: To examine associations between brain structure and BMI for age across the weight spectrum and Centers for Disease Control and Prevention (CDC) weight status categories, distinguishing between and within person associations, and their age and sex moderation. Design, Setting, and Participants: Longitudinal cohort study of youth aged 8 to 10 years at baseline in the US community based Adolescent Brain Cognitive Development Study (4 waves, 2016 to 2025; ages 8 to 18 years). Exposures: Extended CDC BMI-for-age z-score, grouped by wave-mean z-score: underweight (<5th percentile, n=297), healthy weight (5th to <85th, n=7,452), overweight (85th to <95th, n=1,919), and obesity ([&ge;]95th, n=1,842). Main Outcomes and Measures: Cortical thickness (12 regions) and subcortical volume (7 regions) from T1 weighted magnetic resonance imaging (MRI), defined a priori. Results: Among 11,510 participants contributing 32,212 observations (mean [SD] age, 12.6 [2.5] years; 52.2% male), higher between person BMI for age z was associated with lower frontal and temporal thickness (6 regions) and larger subcortical volume (5 regions; all q < .001), but associations were nonlinear (17 of 19 regions). Youth with obesity had lower thickness than healthy weight youth in 8 frontal and temporal regions (largest, -0.21 SD; q < .001), with dorsal frontal differences decreasing across adolescence; underweight youth had smaller subcortical volumes. Higher BMI for age z was associated with larger subcortical volume (all 7 regions) within the healthy weight category and lower cortical thickness (11 of 12 regions; fusiform: {beta} = -0.27; q < .001) within the obesity category, steeper than in healthy weight youth in 10 regions, but not in overweight or underweight youth. Within person associations did not reproduce the obesity associations, reversed from negative at age 10 to positive at age 16, and were more negative in female than male participants. Conclusions and Relevance: Youth with obesity had lower cortical thickness than healthy weight peers, but associations between brain structure and BMI for age differed between CDC weight status groups, suggesting a complex relationship between adiposity and brain structure.

Background: The relationship between recent viral infection and subsequent susceptibility to, or protection against, other viruses remains unclear, particularly across ages and symptom status. Methods: A prospective cohort study enrolled participants aged 6 months to 49 years in Washington and Oregon, USA, 2022-2024. Weekly nasal swabs were tested for SARS-CoV-2, respiratory syncytial virus (RSV), and influenza A/B; additional swabs were collected during symptomatic illness and were also tested for parainfluenza (PIV) 1-4, endemic human coronaviruses (hCOV), adenovirus, human metapneumovirus, human rhinovirus (hRV), and enteroviruses. Proportional hazards regression models were used to characterize associations between incident symptomatic infection (illness) and history of recent illness due to other viruses (within 1-7, 8-14, 15-30, 31-60, 61-90, 91-180 days), adjusted for geographic region, age, sex, daycare/school/work attendance, household size, and number of children in the household. For SARS-CoV-2, influenza, and RSV, these analyses were repeated to study incidence of infection (including asymptomatic cases). Findings: Of 3,560 participants, 2,969 (83%) had [&ge;]1 illness over 4,299 person-years. Adjusted hazard ratios (aHRs) in the short term ([&le;]2 weeks) after an illness were predominantly below 1, indicating that recent viral illness was often associated with lower risk of subsequent illness with other viruses. Across age groups, recent hRV illness was associated with lower risk of hCOV, PIV, and RSV illness (aHRs 0.07-0.27). SARS-CoV-2 illness was associated with a higher short-term risk of adenovirus and PIV illness (aHRs 3.20-3.33). In children aged 6 months-4 years, recent influenza illness was associated with higher short-term risk of SARS-CoV-2 illness. Including asymptomatic infections generally attenuated these associations. Interpretation: Some viral illnesses may alter short-term susceptibility to subsequent viral illnesses. Interactions may differ by age and symptom status. Funding: US Centers for Disease Control and Prevention; Washington State Department of Health.

Clinical records preserve patient findings and final recommendations but rarely capture how specialists select evidence, develop alternative interpretations, disagree, and converge on a management strategy. We describe a dataset derived from ten retrospectively selected, completed cases of histologically confirmed colorectal cancer. For each case, an administrator uploaded de-identified textual clinical records to an authenticated web-based annotation platform. A medical oncologist, surgical oncologist, and radiation oncologist, each with more than ten years of clinical experience, independently linked relevant source fragments to patient-specific interpretations and documented clinical problems, differential hypotheses, uncertainty, and diagnostic and treatment strategies. After the three records were locked, a large language model generated candidate questions for review by the tumour-board chair. The four physicians then discussed each case online; recordings were automatically transcribed and verified by two medical residents. The dataset comprises 30 source-linked specialist reasoning records, 10 verified case-level discussion transcripts, 10 final tumour-board decisions, and associated provenance and quality-control metadata. It supports research on evidence-grounded reasoning, cross-specialty disagreement, multidisciplinary decision-making, and multi-agent clinical decision support.

Background: This study compared functional and cognitive trajectories in older women treated for early breast cancer with matched women without cancer. Methods: Data were derived from two Dutch cohorts: the CLIMB study (women aged 75 years and older with operable breast cancer) and the ISCOPE study (community-dwelling women aged 75 years and older without cancer). Breast cancer patients were matched 1:2 to controls on age and baseline functional or cognitive status. Functional status was assessed using the Groningen Activity Restriction Scale (GARS), and cognitive function using the Mini-Mental State Examination (MMSE). Longitudinal changes over 12 months were analyzed using both unadjusted and confounder-adjusted linear mixed models. Results: 168 breast cancer patients and 336 matched breast cancer-free controls were included with a mean age of 80 years. Most breast cancer patients had stage 0-I or II disease and received surgery and/or radiotherapy; 48% received adjuvant endocrine therapy, chemotherapy was rare (<1%). At baseline, both cohorts were comparable in age and functional or cognitive status, although there was a higher comorbidity burden in the control cohort. No significant differences in functional or cognitive decline were observed between the cohorts over 12 months. Conclusions: In older women, early breast cancer and its treatment did not lead to accelerated decline in functional status or cognitive function within the first year compared to matched older women without cancer. These findings suggest that, for relatively fit older patients, breast cancer treatment may not accelerate functional decline in the short term.

Purpose: To compare the performance of three large language models (LLMs), ChatGPT-o1, Grok 3, and Gemini 2.0 Pro, on the American College of Prosthodontists' National Prosthodontic Resident Examination (NPRE). Materials and Methods: Official NPRE examinations from years 2015-2024 were retrieved from the American College of Prosthodontists website and used to prompt each LLM. Paid access tiers and advanced reasoning were enabled. Responses from LLMs were recorded and compared against the official answer keys for accuracy, variability across years, and consistency of responses. Duplicate questions were retained for the primary analysis and removed in a sensitivity analysis. Accuracy (percent correct) was compared using generalized estimating equations binary logistic regression (primary); Cochran's Q with Bonferroni-adjusted McNemar tests (sensitivity). Year-to-year variability (SD of annual accuracy) was compared using weighted Levene tests. Consistency across duplicate items was quantified as the percentage of perfect matches and compared using Cochran's Q with Bonferroni-adjusted McNemar tests. Results: Across 1,487 questions, overall accuracy differed among models (global P < 0.001). ChatGPT-o1 achieved 78.4% correct, Grok 3 76.4%, and Gemini 2.0 Pro 73.2%; ChatGPT-o1 and Grok 3's accuracy significantly exceeded Gemini 2.0 Pro (both P [&le;] 0.016), with no significant difference between ChatGPT-o1 and Grok 3 (P = 0.099). After removing duplicates (n = 1,200), results were similar: ChatGPT-o1, Grok 3, and Gemini 2.0 Pro were 77.1%, 76.6%, and 71.5% correct, respectively; ChatGPT-o1 and Grok 3's accuracy was significantly higher than Gemini 2.0 Pro (both P < 0.001). Variability across years did not differ significantly (P [&ge;] 0.097). Grok 3 showed significantly higher perfect-match consistency (79.2%) than ChatGPT-o1 (69.4%) and Gemini 2.0 Pro (64.5%) (P [&le;] 0.010). Conclusions: Contemporary LLMs showed 70-80% accuracy on postgraduate prosthodontic multiple-choice content. While promising for educational support, residual error and inconsistency warrant human oversight and transparent methods reporting before curricular adoption.

BACKGROUND Traditional organ dysfunction classifications after severe trauma rely on binary categories (e.g., chronic critical illness [CCI] versus rapid recovery), which likely oversimplify the true heterogeneity of postinjury organ dysfunction trajectories and their underlying immune-biologic drivers. We hypothesized that severely injured trauma patients demonstrate multiple, distinct organ dysfunction phenotypes distinguished by unique immune signatures. METHODS We performed a prospective, longitudinal, observational cohort study of severely injured trauma patients at high risk for multiple organ dysfunction at a single Level 1 trauma center. Organ dysfunction was quantified daily through 28 days, death, or discharge using a trauma-specific modified Sequential Organ Failure Assessment (mSOFA) score excluding the neurologic subdomain to isolate systemic organ dysfunction from traumatic brain injury. We utilized group-based trajectory modeling to identify longitudinal mSOFA phenotypes. Plasma immune protein trajectories were compared across phenotypes using mixed-effects models with time interaction. RESULTS Sixty-nine patients were enrolled. Trajectory modeling identified five mSOFA phenotypes (c1-c5) with differences in resuscitation requirements, shock severity, and 28-day mortality, but not baseline demographics, injury severity, or early physiologic parameters. The phenotypes revealed organ dysfunction heterogeneity not fully captured by CCI/rapid recovery classification. Forty of 92 measured immune markers showed phenotype-differentiated trajectories (FDR<0.1). The most fulminant phenotype (c5) was distinguished by IL-17C, CXCL5, and CXCL6, implicating epithelial-barrier and neutrophil-trafficking mechanisms. The phenotype with highest CCI incidence (c3) was distinguished by rising CXCL9, ARTN, and TWEAK, implicating Th1 polarization and tissue-inflammation/repair signaling. The clinically distinct, older-age phenotype (c4) carried the largest differentiating signature (29 proteins), spanning renal injury/repair, endothelial activation, and adaptive lymphocyte modulation. CONCLUSIONS Organ dysfunction after severe trauma is more biologically heterogeneous than traditional clinical classifications suggest. Distinct mSOFA trajectory phenotypes correspond to distinct immune signatures, supporting separable pathobiology rather than a single axis of injury severity, with implications for phenotype-targeted precision immunomodulatory approaches in critical illness.

Objective: To estimate Long COVID undercoding in electronic health records (EHRs) against federal survey prevalence, and to evaluate a predictor of later coding that anyone can rebuild from counts a research enclave releases. Materials and Methods: Among 23,790 All of Us adults with COVID-19, we compared coded shares with Household Pulse Survey prevalence. Among 23,300 adults uncoded at day 90, we predicted a first U09.9 code by day 455. We fitted naive Bayes and linear discriminant models outside the enclave from released counts and compared them with an elastic net fitted inside to the same patients and features. Tests were a random 20% of patients (4,680, 50 coded) and 18 held-out EHR sites (5,760, 60 coded). Results: Under survey-comparability assumptions, about 1 in 9 adults who ever had Long COVID carried its code, and the women-to-men coding ratio was 0.79 (95% sensitivity interval, 0.65-0.95). The released models' areas under the receiver operating characteristic curve (AUROCs) were 0.68 (95% CI, 0.60-0.75) on the patient test and 0.64 (0.55-0.72) at held-out sites, 0.049 (0.001-0.104) above the elastic net. On the patient test, the elastic net had a higher area under the precision-recall curve (difference 0.062, 0.003-0.122). Released probabilities overestimated coding risk at held-out sites. Discussion: Fitting from released counts did not consistently lower discrimination. Coded outcomes inform disease discrimination only under stated coding assumptions. Conclusion: Long COVID is substantially and unevenly undercoded, and a model rebuilt from public counts predicts its later coding with moderate discrimination.

Background: Self-supervised learning may improve segmentation when labelled data are limited, but its benefit may depend on model and initialisation strategy. This study evaluated in-domain masked-reconstruction pretraining across limited annotation budgets. Methods: The International Skin Imaging Collaboration (ISIC) 2016 development set was partitioned into 765 images for self-supervised pretraining, 94 for fine-tuning and 41 for validation. Five nested annotation budgets (12-94 images) and five seeds were evaluated on the 379-image test set. A custom U-Net compared random and in-domain self-supervised initialisation; a ResNet-34 U-Net compared random, ImageNet, in-domain self-supervised and sequential ImageNet-to-self-supervised initialisation. Models trained with 94 labelled images were evaluated without adaptation on ISIC 2017. A separate experiment examined restoration-based pretraining. Results: In the custom U-Net, self-supervised pretraining achieved higher mean Dice than random initialisation at all five budgets, including 0.664 versus 0.615 with 12 labels and 0.811 versus 0.768 with 94 labels. Seed-level 95% confidence intervals for paired differences excluded zero at 47 and 94 labels but not at the other budgets. With 47 labels, the self-supervised model achieved mean Dice of 0.789, compared with 0.768 for random initialisation with 94 labels. In ResNet-34, ImageNet initialisation achieved higher mean Dice than in-domain self-supervised pretraining at all five budgets, and adding self-supervised pretraining after ImageNet reduced mean Dice at four of five budgets. All six 94-label configurations achieved lower mean Dice on ISIC 2017 than ISIC 2016. In the restoration experiment, mean Dice was 0.686 with random initialisation and 0.638-0.642 following pretraining. Conclusions: In-domain masked-reconstruction pretraining improved mean Dice relative to random initialisation in the custom U-Net, but not consistently in the ResNet-34 U-Net, where ImageNet initialisation performed better at every budget. Adding self-supervised pretraining after ImageNet provided no consistent gain. These findings are specific to the dermoscopic lesion-segmentation experiments and do not establish melanoma-specific diagnostic or clinical performance.

Importance: Preventing substance use disorder (SUD) is a top public health priority, but clinical practice currently lacks validated prognostic tools to identify risk before the onset of heavy use/problems. Objective: To determine whether models can predict SUD using genetic, behavioral, and environmental factors and to evaluate performance across substance types (alcohol, cannabis, tobacco), sex, and genetic similarity groups. Design: This prognostic study developed and validated a multi-domain risk model using data from four longitudinal cohorts (The National Longitudinal Study of Adolescent to Adult Health, Avon Longitudinal Study of Parents and Children, The Collaborative Study on the Genetics of Alcoholism, and FinnTwin12) with follow-up from early adolescence into young adulthood. Setting: Population-based and family-based samples from the United States, United Kingdom, and Finland. Participants: The analysis included participants with available genomic and behavioral data. The cohorts represented a spectrum of risk, ranging from nationally representative community samples to high-risk populations. Exposure(s) (for observational studies): Predictors included polygenic scores for substance use-related phenotypes and 10 established early-life behavioral and environmental risk factors. Main Outcome(s) and Measure(s): The outcomes were a diagnosis of moderate-to-severe SUD based on the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and positive and negative predictive value. Results: A total of 16,362 participants were included in the study (54.7% female; mean [SD] age 26.73 [3.36]). Genetic similarity group composition was 70.4% European, 16.8% African, 9.8% admixed American, 2.2% East Asian, and 0.7% South Asian. SUD prevalence was 22.0%. The integrated model achieved an AUC of 0.78 for predicting any SUD, with a sensitivity of 69% and a positive predictive value of 45%. Performance was comparable for specific substances (AUC range 0.74-0.82) and across genetic similarity groups (AUC range 0.61-0.82). The model demonstrated higher AUC in females compared with males. Conclusions and Relevance: Findings highlight the potential utility of an SUD prediction model to inform clinicians and patients about risk for developing a SUD.

Importance: Conditions that are hard to diagnose, several more common in women, may be missing from electronic health records (EHRs), hiding patients from research cohorts and case finding. Objective: To measure how often the EHR contains a diagnosis code for chronic conditions that participants report, and to compare this with national survey benchmarks. Design, Setting, and Participants: This cross-sectional study used the All of Us Research Program Registered Tier Dataset v9. Participants were adults with EHR visits in both 12-month periods from July 2021 to June 2023 who completed a health history survey. We analyzed the data in October 2026. Exposures: Self-reported clinician diagnosis of 10 conditions with ongoing visits or treatment, as well as sex at birth, race and ethnicity, age, and years of EHR records. Main Outcomes and Measures: Share of participants reporting current care who had a qualifying diagnosis code, alone and relative to hypertension and type 2 diabetes, and the ratio of age-standardized coded shares to national survey prevalence. Results: Of 161,480 respondents, 114,610 (73,870 female [64.5%], median [IQR] age 63 [50-72] years) had at least 5 years of EHR records before the survey. Among those reporting current care, 84.6% (95% CI, 84.2%-85.0%) with hypertension and 87.0% (95% CI, 86.5%-87.6%) with type 2 diabetes had a code, compared with 41.3% (95% CI, 38.9%-43.8%) for endometriosis, 39.4% (95% CI, 34.3%-44.8%) for Lyme disease, and 24.4% (95% CI, 23.2%-25.6%) for chronic fatigue. Counting ICD-9-CM and other mapped codes, 55.3%, 51.5%, and 28.3%, respectively, had any qualifying record. Women had lower coded shares than men for hypertension, type 2 diabetes, obesity, depression, and anxiety (crude ratios 0.93-0.96, within-health-system ratios 0.98-1.02) and higher shares for migraine and fibromyalgia (within health systems, 1.11 [95% CI, 1.07-1.15] and 1.41 [95% CI, 1.29-1.57]). Age-standardized coded shares among active patients were 1.1 to 2.4 times national prevalence of the same self-reported diagnoses. Conclusions and Relevance: In this cross-sectional study, fewer than half of participants reporting ongoing care for chronic fatigue, Lyme disease, or endometriosis had a qualifying ICD-10-CM code. Linking each participant's report to their own record measured this gap directly.

Objective Reducing household air pollution exposure during gestation and infancy may have longer-term benefits for child linear growth. We examined whether a liquefied petroleum gas (LPG) cookstove and fuel intervention delivered during gestation and infancy could improve height-for-age Z scores (HAZ) and reduce the risk of stunting at ages 2-5 years in India, Rwanda, and Guatemala. Materials and method The Household Air Pollution Intervention Network (HAPIN) was a randomized controlled trial that enrolled ~3200 pregnant women across Guatemala, India, Peru, and Rwanda. Women in the intervention group received an LPG cookstove and free fuel during pregnancy and until the infant reached 12 months of age. Following the HAPIN trial, we continued following children in Guatemala, India, and Rwanda (N=2169) and measured child height at ages 2, 3, 4, and 5 years (HAPIN 2.0). As prespecified in the study protocol, we used mixed-effects models, as well as linear and log-binomial regression, within the intention to treat (ITT) framework, to evaluate the interventions effect on HAZ and stunting. Subgroup analyses were performed to assess effect modification by child sex, maternal height, and intervention country. Results The intervention reduced gestational exposures to particulate matter (PM2.5: 80.7 vs 117.5 {micro}g/m3), carbon monoxide (CO: 1.3 vs 2.0 ppm), and black carbon (BC: 8.6 vs 12.1 {micro}g/m3) and infancy exposures (PM2.5: 36.9 vs 113.2 {micro}g/m3; CO: 0.60 vs 2.2 ppm). Post-trial PM2.5 exposure were similar between arms: 99.9 vs 98.7 {micro}g/m3. We found no statistically significant effect of the intervention on repeated measures of HAZ ({beta} = -0.06, 95% CI: -0.16, 0.03 p=0.204), annual HAZ, stunting, nor severe stunting at any age during the follow-up. There was no statistically significant effect modification by subgroups. Conclusion Despite substantial reductions in air pollution exposures during the trial, the LPG intervention did not improve growth or reduce stunting risk from ages 2 through 5 years in this multi-country cohort. These findings, consistent with prior results of the lack of effects on growth through 12 months, provide additional evidence that a clean cooking intervention alone may not be sufficient to address childhood stunting in low-resource settings. Future exposure-response analyses will supplement these ITT analyses.

Background and Objective: Adding belzutifan to adjuvant pembrolizumab improved disease-free survival (DFS) in LITESPARK-022, but its durability beyond the 1-year treatment is unknown. Methods: Following a registered plan, we reconstructed individual patient data from published Kaplan-Meier curves of ten phase 3 adjuvant renal cell carcinoma trials. Within each trial, we estimated restricted mean survival time differences ({Delta}RMST), hazard ratios (HRs) before and after planned treatment ended, and DFS differences. Key Findings and Limitations: In LITESPARK-022 (central review), the 24-month {Delta}RMST was 1.34 months (95% confidence interval 0.73-1.95), and the belzutifan arm remained ahead after month 12 ({Delta}RMST increment 0.74 months/yr, 0.32-1.19). The HR was 0.55 (0.43-0.71) before month 12 and 1.16 (0.87-1.54) afterwards (post/pre ratio 2.10); investigator assessment showed similar attenuation. Similar ratios occurred in KEYNOTE-564 at its first interim analysis (2.02), whose benefit persisted, and in SORCE and PROTECT (2.09, 2.59), whose benefit faded. From months 12 to 30 (post hoc), the DFS difference changed by -3.7 percentage points (-8.1 to 0.8) with central review and 0.6 (-4.0 to 5.0) with investigator assessment. Data were reconstructed, and follow-up was short. Conclusions and Clinical Implications: Relative hazard reductions with belzutifan were greater before month 12; durability of the absolute DFS advantage is uncertain and assessment-dependent, so single-HR projections should be avoided.

Purpose Prescribing to lactating women is constrained by limited evidence. Published compatibility classifications provide clinical guidance and a discrete research comparator. How often prescribing diverges from that guidance, and what characteristics are associated with divergence, is unknown at national scale. Methods Retrospective analysis of prescriptions filled during lactation, identified in the 2012-2014 MarketScan Commercial Claims Database. Prescriptions were classified as concordant (Hale's lactation risk category L1-L3) or discordant (L4-L5). Multivariable logistic regression identified factors associated with concordance. Results Of 96,484 new prescriptions, 95,710 (99.2%) were concordant and 774 (0.8%) were discordant. Discordant prescriptions were filled later after delivery (median 59 vs 33 days) and had a day supply three times longer (30 vs 10 days). In multivariable analysis (N=28,732), concordant prescriptions were associated with maternal age under 35 (adjusted odds ratio (aOR)=1.34, 95% confidence interval (CI): 1.02-1.75) and chronic-condition medications (aOR=10.94, 95% CI: 6.47-18.50), and were less likely within 14 days of delivery (aOR=0.53, 95% CI: 0.38-0.73) and with each additional day of supply (aOR=0.96, 95% CI: 0.96-0.96). Compared with family practice, concordance did not differ for obstetric, internal medicine, or emergency medicine prescriptions, but prescriptions from other provider types had lower odds of concordance (aOR=0.57, 95% CI: 0.41-0.79). Conclusions Prescribing during lactation was overwhelmingly concordant with published compatibility guidance; the rare divergence was patterned by prescription and prescriber characteristics. Because the categories are an expert classification rather than consensus guidance, divergence should not be read as inappropriate prescribing.