Tag: statistics

Biostatistics. 2024 Dec 31;26(1):kxaf013. doi: 10.1093/biostatistics/kxaf013.

ABSTRACT

Object-oriented data analysis is a fascinating and evolving field in modern statistical science, with the potential to make significant contributions to biomedical applications. This statistical framework facilitates the development of new methods to analyze complex data objects that capture more information than traditional clinical biomarkers. This paper applies the object-oriented framework to analyze physical activity levels, measured by accelerometers, as response objects in a regression model. Unlike traditional summary metrics, we utilize a recently proposed representation of physical activity data as a distributional object, providing a more nuanced and complete profile of individual energy expenditure across all ranges of monitoring intensity. A novel hybrid Fréchet regression model is proposed and applied to US population accelerometer data from National Health and Nutrition Examination Survey (NHANES) 2011 to 2014. The semi-parametric nature of the model allows for the inclusion of nonlinear effects for critical variables, such as age, which are biologically known to have subtle impacts on physical activity. Simultaneously, the inclusion of linear effects preserves interpretability for other variables, particularly categorical covariates such as ethnicity and sex. The results obtained are valuable from a public health perspective and could lead to new strategies for optimizing physical activity interventions in specific American subpopulations.

PMID:40408136 | DOI:10.1093/biostatistics/kxaf013

JAMA Netw Open. 2025 May 1;8(5):e2511922. doi: 10.1001/jamanetworkopen.2025.11922.

ABSTRACT

IMPORTANCE: Large language models (LLMs) and text-embedding models have shown potential in assessing mental health risks based on narrative data from psychiatric patients.

OBJECTIVE: To assess whether LLMs and text-embedding models can identify depression and suicide risk based on sentence completion test (SCT) narratives of psychiatric patients.

DESIGN, SETTING, AND PARTICIPANTS: This cross-sectional study, conducted at Seoul Metropolitan Government-Seoul National University Boramae Medical Center, analyzed SCT data collected from April 1, 2016, to September 30, 2021. Participants included psychiatric patients aged 18 to 39 years who completed SCT and self-assessments for depression (Beck Depression Inventory-II or Zung Self-Rating Depression Scale) and/or suicide (Beck Scale for Suicidal Ideation). Patients confirmed to have an IQ below 70 were excluded, leaving 1064 eligible SCT datasets (52 627 completed responses). Data processing with LLMs (GPT-4o, May 13, 2024, version; OpenAI [hereafter, LLM1]; gemini-1.0-pro, February 2024 version; Google DeepMind [hereafter, LLM2]; and GPT-3.5-turbo-16k, January 25, 2024, version; OpenAI) and text-embedding models (text-embedding-3-large, OpenAI [hereafter, text-embedding 1]; text-embedding3-small; OpenAI; and text-embedding-ada-002; OpenAI) was performed between July 4 and September 30, 2024.

MAIN OUTCOMES AND MEASURES: Outcomes included the performance of LLMs and text-embedding models in detecting depression and suicide, as measured by the area under the receiver operating characteristic curve (AUROC), balanced accuracy, and macro F1-score. Performance was evaluated across concatenated narratives of SCT, including self-concept, family, gender perception, and interpersonal relations narratives.

RESULTS: Based on SCT narratives from 1064 patients (mean [SD] age, 25.4 [5.5] years; 673 men [63.3%]), LLM1 showed strong performance in zero-shot learning, with an AUROC of 0.720 (95% CI, 0.689-0.752) for depression and 0.731 (95% CI, 0.704-0.762) for suicide risk using self-concept narratives. Few-shot learning for depression further improved the performance of LLM1 (AUROC, 0.754 [95% CI, 0.721-0.784]) and LLM2 (AUROC, 0.736 [95% CI, 0.704-0.770]). The text-embedding 1 model paired with extreme gradient boosting outperformed other models, achieving an AUROC of 0.841 (95% CI, 0.783-0.897) for depression and 0.724 (95% CI, 0.650-0.795) for suicide risk. Overall, self-concept narratives showed the most accurate detections across all models.

CONCLUSIONS AND RELEVANCE: This cross-sectional study of SCT narratives from psychiatric patients suggests that LLMs and text-embedding models may effectively detect depression and suicide risk, particularly using self-concept narratives. However, while these models demonstrated potential for detecting mental health risks, further improvements in performance and safety are essential before clinical application.

PMID:40408109 | DOI:10.1001/jamanetworkopen.2025.11922

JAMA Netw Open. 2025 May 1;8(5):e2512064. doi: 10.1001/jamanetworkopen.2025.12064.

ABSTRACT

IMPORTANCE: Depression detection in primary care remains limited in low- and middle-income countries despite increasing use of the World Health Organization Mental Health Gap Action Programme-Intervention Guide (mhGAP-IG).

OBJECTIVE: To test an app version of the mhGAP-IG (e-mhGAP-IG) in Nepal and Nigeria to improve depression detection.

DESIGN, SETTING, AND PARTICIPANTS: In this feasibility cluster randomized clinical trial conducted from February 14, 2021, to March 25, 2022, primary care facilities (unit of clustering) in Nepal and Nigeria were randomized to the standard mhGAP-IG training arm (control) or to training using the e-mhGAP-IG app (intervention). Primary care workers (PCWs) received training based on the arm assignment of their health care facility. Statistical analysis was conducted from July 20, 2022, through September 27, 2024.

INTERVENTION: Training using standard mhGAP-IG vs training using the e-mhGAP-IG.

MAIN OUTCOMES AND MEASURES: Analysis was performed on an intention-to-treat basis. The main outcome was accuracy of depression detection rates by PCWs, evaluated prior to mhGAP training and 5 to 8 months after training, measured as the percentage of patients who received a depression diagnosis by their PCWs compared with the number of patients who scored 10 or more on the locally validated 9-item Patient Health Questionnaire. Costs per patient detected were calculated.

RESULTS: In Nepal, 25 facilities (67 PCWs; mean [SD] age, 35.3 [9.2] years; 52 men [78%]) were randomized: 13 facilities to standard mhGAP-IG training (36 PCWs) and 12 facilities to e-mhGAP-IG (31 PCWs). In Nigeria, 10 facilities (47 PCWs; mean [SD] age, 46.9 [7.5] years; 44 women [94%]) were randomized: 5 facilities to standard mhGAP-IG (25 PCWs) and 5 facilities to e-mhGAP-IG (22 PCWs). In Nepal, depression detection by PCWs in the standard mhGAP-IG arm increased from 0 of 43 patients before training to 15 of 92 patients after training (adjusted mean change [AMC], 16% [95% CI, 5%-28%]), and depression detection in the e-mhGAP-IG arm increased from 0 of 49 before training to 22 of 91 after training (AMC, 24% [95% CI, 12%-36%]). In Nigeria, depression detection in the standard mhGAP-IG arm increased from 5 of 36 patients before training to 25 of 75 patients after training (AMC, 19% [95% CI, 2%-37%]), and depression detection in the e-mhGAP-IG arm increased from 6 of 35 patients before training to 67 of 76 patients after training (AMC, 71% [95% CI, 57%-85%]). In facilities in the e-mhGAP-IG arm, the app was used for 59 of 616 assessments (10% of patients) in Nepal and 883 of 1077 assessments (82% of patients) in Nigeria. Cost per patient with depression detected using the e-mhGAP-IG was Nepali Rupiya (NPR) 1980 (US $14.79) in Nepal and naira (₦) 1462 (US $0.91) in Nigeria.

CONCLUSIONS AND RELEVANCE: This feasibility cluster randomized clinical trial demonstrated that the use, cost, and potential clinical benefit of the e-mhGAP-IG varied by setting, highlighting the importance of multisite feasibility studies when evaluating digital innovations intended for health care systems worldwide.

TRIAL REGISTRATION: ClinicalTrials.gov Identifier: NCT04522453.

PMID:40408108 | DOI:10.1001/jamanetworkopen.2025.12064

JAMA Netw Open. 2025 May 1;8(5):e2512081. doi: 10.1001/jamanetworkopen.2025.12081.

ABSTRACT

IMPORTANCE: Suicide risk may be elevated among individuals who separate from military service with a character of discharge or service that was not honorable (ie, dishonorable, bad conduct, other than honorable, general, or uncharacterized), but a comprehensive evaluation of this possibility has not been previously published.

OBJECTIVE: To examine suicide mortality by character of discharge compared with individuals who received an honorable discharge and the full veteran population recently separated from service.

DESIGN, SETTING, AND PARTICIPANTS: This retrospective cohort study included all individuals who separated from the active component of the US military between January 1, 2002, and December 31, 2021. Data were obtained from the US Department of Veterans Affairs/Department of Defense Mortality Data Repository. Suicide rates for each character of discharge were compared with those among individuals receiving an honorable discharge and the full veteran population recently separated from service. Analyses were conducted between December 21, 2023, and June 1, 2024.

EXPOSURE: Character of discharge assigned at separation from the active component of military service.

MAIN OUTCOMES AND MEASURES: Crude suicide rates up to 5 years after separation were examined. Age-standardized rate ratios were used for comparisons within the cohort between strata. Age-standardized mortality ratios (SMRs) were used to compare suicide mortality by character of discharge with the full separation cohort and within specific military service branches.

RESULTS: Among 3 627 653 individuals (mean [SD] age at separation, 28.4 [8.6] years; 83% men), 5599 deaths by suicide occurred during the study period. All character of discharge groups had significantly higher suicide rates than the honorable group, with standardized rate ratios ranging from 1.91 (95% CI, 1.69-2.17) for uncharacterized to 2.77 (95% CI, 2.52-3.05) for general. Veterans who received an honorable discharge were less likely to die by suicide than the overall cohort population (SMR, 0.82 [95% CI, 0.79-0.85]), while veterans with all other characters of discharge were more likely to die by suicide, with SMRs ranging from 1.23 (95% CI, 1.15-1.31) for uncharacterized to 1.84 (95% CI, 1.72-1.97) for general. Common demographic risk factors for suicide (eg, age, race) were replicated within character groups in most cases.

CONCLUSIONS AND RELEVANCE: In this cohort study, individuals who did not receive an honorable character of discharge were at significantly higher risk of suicide compared with veterans who received an honorable discharge and the veteran population recently separated from service. These findings suggest that character of discharge may be a helpful risk factor to consider for ongoing suicide prevention efforts.

PMID:40408107 | DOI:10.1001/jamanetworkopen.2025.12081

JAMA Netw Open. 2025 May 1;8(5):e2512189. doi: 10.1001/jamanetworkopen.2025.12189.

ABSTRACT

IMPORTANCE: A recent increase in state policies targeting gender minority (GM; transgender and gender-diverse) people may affect the mental health of sexual and gender minority (SGM; nonheterosexual and/or GM) people and GM people specifically.

OBJECTIVE: To estimate changes in mental health symptoms associated with enactment of anti-GM state policies among SGM people and GM people specifically.

DESIGN, SETTING, AND PARTICIPANTS: This repeated cross-sectional study used a staggered difference-in-differences analysis to examine the associations between anti-GM policies in the US and mean changes in mental health symptoms among 8733 SGM adults who completed annual questionnaires between April 1, 2020, and June 1, 2023, for The Population Research in Identity and Disparities for Equality (PRIDE) Study, a national, prospective, continuously enrolling online cohort study of SGM adults.

EXPOSURES: Living in a state with 1 or more of the following enacted policies: (1) bathroom restrictions for GM people, (2) sports bans for GM young people participating in school sports, and (3) bans on gender-affirming care for young people.

MAIN OUTCOMES AND MEASURES: Mean levels of anxiety (measured using the 7-item Generalized Anxiety Disorder scale [GAD-7]; total score range, 0-21), depression (measured using the 9-item Patient Health Questionnaire scale [PHQ-9]; total score range, 0-27), and posttraumatic stress disorder (PTSD) symptoms (measured using the 6-item PTSD Checklist scale [PCL-6]; total score range, 6-30). For all 3 scales, higher scores indicate more severe symptoms.

RESULTS: Among all 8733 SGM participants in the sample (median age, 32.5 years [IQR, 26.0-45.0 years]; 2024 cisgender men [23.2%], 2355 cisgender women [27.0%], 2198 gender-diverse adults assigned female at birth [25.2%], 321 gender-diverse adults assigned male at birth [3.7%], 1294 transgender men [14.8%], and 541 transgender women [6.2%]), anti-GM policy enactment was associated with significant increases in anxiety (GAD-7 score, 0.8 points [95% CI, 0.2-1.4 points]) and PTSD (PCL-6 score, 0.8 points [95% CI, 0.1-1.4 points]) symptoms in states that enacted anti-GM policies compared with states that did not but was not associated with significant increases in depression symptoms (PHQ-9 score, 0.6 points [95% CI, -0.1 to 1.4 points]). In the GM subsample (n = 4354), nonsignificant changes in anxiety (GAD-7 score, 0.6 points [95% CI, -0.2 to 1.4 points]), depression (PHQ-9 score, 0.1 points [95% CI, -0.9 to 1.1 points]), and PTSD (PCL-6 score, 0.7 points [95% CI, -0.2 to 1.6 points]) symptoms were observed after policy enactment in states that enacted anti-GM policies compared with states that did not. Gender minority adults had high mental health symptoms across the study period.

CONCLUSIONS AND RELEVANCE: In this study of 8733 SGM adults using difference-in-differences analysis, anti-GM policies were associated with worse mental health symptoms among SGM adults but no changes in mental health symptoms among GM adults. As these policies proliferate, it is important to consider how they may affect mental health.

PMID:40408106 | DOI:10.1001/jamanetworkopen.2025.12189

JAMA Netw Open. 2025 May 1;8(5):e2512198. doi: 10.1001/jamanetworkopen.2025.12198.

ABSTRACT

IMPORTANCE: Life expectancy is a composite health measure reflecting acute and life-course exposures. Identifying conditions behind disparities in life expectancy can guide policy, planning, and financing to battle the most urgent health problems.

OBJECTIVE: To examine the contribution of 33 causes of death to life expectancy disparities, highlighting 2 sets of priority conditions-8 infectious and maternal and child health conditions (I-8) and 7 noncommunicable diseases and injuries (NCD-7).

DESIGN, SETTING, AND PARTICIPANTS: This cross-sectional study examined life expectancy disparities in 7 global regions and 165 countries from 2000 to 2021. Western Europe and Canada (hereafter referred to as the North Atlantic) in 2019 were used as a benchmark for life expectancy achievable with advanced health care and living standards. Life expectancy gaps in locations with life expectancy lower than the benchmark were decomposed by cause of death using the Pollard decomposition on the Global Health Estimates from the World Health Organization. Data were analyzed from February to March 2025.

EXPOSURE: Geographic location (countries and regions).

MAIN OUTCOME AND MEASURE: Life expectancy at birth.

RESULTS: In the median country in 2019, the I-8 and NCD-7 together accounted for 80% (IQR, 71%-88%) of the life expectancy gap compared with the North Atlantic. Outside sub-Saharan Africa, the NCD-7 accounted for the largest share of the gap; for example, more than the total life expectancy gap in China, or 5.5 (95% uncertainty bounds [UB], 5.0-6.0) years of a 4.3-year life expectancy gap; and 6.4 (95% UB, 5.9-6.8) years of a 11.5-year gap in India. However, reduced mortality from the I-8 contributed to enormous improvements in sub-Saharan Africa, accounting for 21.4 (95% UB, 20.6-22.2) years of a 31-year gap in 2000 and 11.4 (95% UB, 10.9-11.8) years of a 22-year gap in 2019. India transitioned from having most of the gap accounted for by the I-8 in 2000, or 11.9 (95% UB, 11.0-13.0) years of a 19.6-year life expectancy gap, to having a larger share accounted for by the NCD-7 in 2019.

CONCLUSIONS AND RELEVANCE: This cross-sectional study suggests that a limited number of causes account for most life expectancy disparities. Together with current information on risk factors, interventions, and morbidity not yet reflected in life expectancy, the varying contributions of these causes to gaps in life expectancy can help focus health policy and guide interventions to reduce risk factors and treat conditions.

PMID:40408105 | DOI:10.1001/jamanetworkopen.2025.12198

JAMA Netw Open. 2025 May 1;8(5):e2512209. doi: 10.1001/jamanetworkopen.2025.12209.

ABSTRACT

IMPORTANCE: Postpartum hemorrhage (PPH) remains the leading cause of maternal mortality globally. Although traditional management relies on blood loss quantification, identifying cases resistant to conventional treatment remains challenging, potentially delaying crucial interventions.

OBJECTIVE: To determine the prevalence and clinical implications of PPH that is resistant to treatment showing arterial contrast extravasation on dynamic computed tomography (CT) (PRACE) and its association with intervention requirements.

DESIGN, SETTING, AND PARTICIPANTS: A multicenter retrospective case-control study of PPH cases was conducted at 43 tertiary facilities across Japan, including 30 university hospitals, between January and December 2021. The study included patients with PPH who were either transported to the centers or who delivered there. Data analysis was performed from September 2023 to November 2024.

EXPOSURE: A diagnosis of PPH, defined as blood loss exceeding 2000 mL or requiring more than 10 units of red blood cell transfusion.

MAIN OUTCOMES AND MEASURES: The primary outcome was the prevalence of PRACE and its association with uterine artery embolization (UAE) requirement. Secondary outcomes included risk factors for severe coagulopathy (fibrinogen <150 mg/dL). Odds ratios (ORs) were estimated using multivariable logistic regression to evaluate the association between types of PPH (atonic uterus, tissue, and PRACE) and PPH severity.

RESULTS: Among 352 patients (median [IQR] age, 33.0 [30.0-37.0] years; 211 [60.0%] primiparous), 205 (58.2%) underwent CT scans, with PRACE detected in 58 (32.2%) of evaluable cases. Patients with PRACE had significantly higher total blood loss (median [IQR], 3455 [2000-5070] mL vs 2500 [1500-2650] mL) and greater UAE requirement (50 of 58 patients [86.2%] vs 35 of 122 patients [28.7%]) compared with patients without PRACE. PRACE was the primary factor associated with the need for UAE (OR, 27.74; 95% CI, 10.52-83.14).

CONCLUSIONS AND RELEVANCE: In this retrospective case-control study of patients with severe PPH undergoing dynamic CT, PRACE represented a distinct and common pathology in severe PPH and was associated with the need for interventional procedures. These findings suggest that dynamic CT imaging should be considered as an essential diagnostic tool in managing treatment-resistant PPH cases.

PMID:40408104 | DOI:10.1001/jamanetworkopen.2025.12209

JAMA Netw Open. 2025 May 1;8(5):e2512262. doi: 10.1001/jamanetworkopen.2025.12262.

ABSTRACT

IMPORTANCE: In most countries, children in the same grade can be up to 12 months apart in age, which represents a considerable relative age difference, particularly in the early years. A relative age effect has been frequently reported in the diagnosis and treatment of attention-deficit/hyperactivity disorder and, more rarely, in the diagnosis of other neurodevelopmental disorders. The relative age effect has never been studied for speech therapy, which is frequently prescribed to treat specific language and learning disorders.

OBJECTIVE: To quantify the association of relative age with the initiation of speech therapy.

DESIGN, SETTING, AND PARTICIPANTS: This cohort study used data from the French National Health Data System and included all children born in France between 2010 and 2016 from September of the year of their 5th birthday until July of the year of their 10th birthday or July 31, 2022 (end of study).

EXPOSURE: Children were categorized according to month of birth.

MAIN OUTCOMES AND MEASURES: Risk of speech therapy initiation was estimated using Cox models adjusted for other known risk factors.

RESULTS: In total, 4 188 985 children (mean [SD] age, 5.2 [0.3] years; 50.8% girls) were included. Speech therapy was initiated for 692 086 children (incidence rate, 53.1 per 1000 person-years, 54.5% boys) during a mean (SD) follow-up of 3.1 (1.6) years. Among children of the same schooling level, the risk of initiating speech therapy steadily increased according to quarter and month of birth, with adjusted hazard ratios of 1.51 (95% CI, 1.50-1.52) for children born in the last quarter vs the first quarter and 1.64 (95% CI, 1.62-1.66) for children born in December vs those born in January. The relative age effect observed for speech therapy was of the same magnitude as that observed for methylphenidate in attention-deficit/hyperactivity disorder (positive control outcome). No relative age effect was observed for the initiation of desmopressin for nocturnal enuresis (negative control outcome).

CONCLUSIONS AND RELEVANCE: This cohort study found that among children in the same grade, an age difference of a few months was associated with the frequency of initiation of speech therapy. These findings may be attributable to a mismatch between the expectations of the school system and the relative age and level of maturity of the children or earlier identification of neurodevelopmental disorders in the youngest children in each age group.

PMID:40408103 | DOI:10.1001/jamanetworkopen.2025.12262

Invest Ophthalmol Vis Sci. 2025 May 1;66(5):32. doi: 10.1167/iovs.66.5.32.

ABSTRACT

PURPOSE: Mutations affecting the CRB1 gene produce retinal dystrophies including early onset severe retinal dystrophy/Leber congenital amaurosis (EOSRD/LCA), retinitis pigmentosa (RP), cone-rod dystrophy (CORD), and macular dystrophy (MD). As treatment strategies advance toward clinical translation, there is a need to establish reliable outcome metrics and to better understand the visual deficits associated with CRB1 retinopathies. To this end, we measured visual acuity (VA) and crowding (the disruptive effect of clutter on object recognition), both key functions in spatial vision, using child-friendly computer-based tests, and gold-standard clinical measures.

METHODS: Patients with molecularly confirmed biallelic CRB1 pathogenic variants were compared with age-matched controls (n = 20 in each). Best-corrected visual acuity (BCVA) was measured with both Early Treatment Diabetic Retinopathy Study (ETDRS) and the computerized VacMan procedures (using an unflanked/isolated VacMan target), which also allowed measurement of crowding when surrounding flanker elements were added.

RESULTS: Both acuity and crowding were significantly elevated in individuals with CRB1 retinopathy compared with controls. ETDRS acuity correlated with both the unflanked (r = 0.868, P < 0.001) and flanked VacMan thresholds (r = 0.748, P < 0.001). No statistically significant changes in crowding were observed with respect to CRB1 phenotype (EOSRD/LCA, CORD, or MD) or age of onset.

CONCLUSIONS: This study demonstrates for the first time that individuals with CRB1 retinopathy exhibit elevated crowding in their foveal vision compared with controls. Measuring crowding offers valuable insights into understanding functional visual deficits in CRB1 retinopathy and could be a useful metric for monitoring disease progression and treatment outcomes in inherited retinal diseases.

PMID:40408095 | DOI:10.1167/iovs.66.5.32

JAMA Health Forum. 2025 May 2;6(5):e250820. doi: 10.1001/jamahealthforum.2025.0820.

ABSTRACT

IMPORTANCE: Demographic and policy changes have occurred since the last large, representative study of homeless adults in the 1990s, which may affect health care access and use.

OBJECTIVE: To describe the prevalence of poor health care access and short-term health care and use the Gelberg-Andersen Behavioral Model for Vulnerable Populations to assess the association between changes in homelessness with health care access and short-term care use.

DESIGN, SETTING, AND PARTICIPANTS: This representative survey of adults experiencing homelessness in California from October 2021 to November 2022 used multistage, venue-based, and respondent-driven sampling. Data were analyzed from May 2023 to December 2024.

EXPOSURES: Shelter status (predisposing vulnerable), insurance (enabling), impairment with activities of daily living (ADL; need), and illicit substance use during the previous 6 months (need).

MAIN OUTCOMES AND MEASURES: The study assessed self-reported no prior-year ambulatory care use and prior 6-month unmet health care need, unmet medication need, emergency department (ED) use, and hospitalization. Population prevalence estimates with Wald 95% CIs and multivariable Poisson regressions were calculated to compute prevalence ratios (PRs).

RESULTS: Thirty-two hundred adults completed the survey (mean age, 46.1 [95% CI, 45.3-46.9] years; 1965 cisgender men [67.2%], 1148 cisgender women [31.2%], and 57 transgender and gender queer individuals [1.6%]), of whom 2016 (77.6%) were unsheltered, 2609 (82.6%) were insured, 1056 (34.4%) had an ADL impairment, and 911 (37.1%) reported illicit substance use 3 or more times a week. A total of 1121 (39.1%) reported no ambulatory care use; 765 (24.3%) reported an unmet health care need and 714 (23.3%) an unmet medication need; 1252 (38.9%) used the ED; and 668 (22.0%) were hospitalized. Lack of ambulatory care use (PR, 1.71; 95% CI, 1.51-1.94) and unmet health care needs (PR, 1.19; 95% CI, 1.02-1.40) were more prevalent for those who were unsheltered. Lack of ambulatory care use (PR, 0.63; 95% CI, 0.57-0.70) and unmet health care needs (PR, 0.80; 95% CI, 0.67-0.95) were less prevalent for those with insurance. Unmet health care needs (PR, 2.13; 95% CI, 1.79-2.55), ED use (PR, 1.15; 95% CI, 1.02-1.30), and hospitalization (PR, 1.74; 95% CI, 1.40-2.17) were more prevalent for those with an ADL impairment. Lack of ambulatory care use (PR, 1.46; 95% CI, 1.19-1.79) and unmet health care needs (PR, 1.30; 95% CI, 1.08-1.55) were more prevalent for those who used illicit substances 3 or more times a week.

CONCLUSIONS AND RELEVANCE: This cross-sectional study found that adults experiencing homelessness reported poor access to ambulatory care and a high prevalence of short-term care use, despite high rates of insurance. Changes in homelessness during the past 30 years were associated with worsened health care access and use.

PMID:40408090 | DOI:10.1001/jamahealthforum.2025.0820