Homelessness and alcohol-related inpatient diagnoses across intersecting social positions in the United States, 2016-2023.

INTRODUCTION: Homelessness is a structural determinant of health associated with elevated morbidity and healthcare utilisation. Evidence remains limited on how homelessness relates to alcohol-related inpatient diagnoses across intersecting sociodemographic groups. METHODS: We analysed adult inpatient discharge records (n=91 709 874) from 18 US states (2016-2023) in the Healthcare Cost and Utilisation Project State Inpatient Databases. Homelessness was identified using International Classification of Diseases, Tenth Revision, Clinical Modification Z codes. Outcomes included any alcohol-related mental or behavioural disorder, alcohol abuse, alcohol dependence and severe alcohol-induced psychiatric conditions. Bayesian multilevel logistic regression models with intersectional stratification estimated predicted prevalence and average marginal effect risk ratios (AMEs) comparing discharges coded with homelessness to those without, accounting for sociodemographic heterogeneity and temporal trends. RESULTS: Alcohol-related diagnoses were substantially more prevalent among discharges coded with homelessness. Predicted prevalence of any alcohol-related disorder was 34.7% versus 6.6% among those without. Overall AMEs ranged from 3.41 for alcohol abuse to 4.57 for alcohol dependence. Although male individuals and adults aged 50-64 years with homelessness had higher absolute prevalence, larger relative effects were observed among female individuals (AME 5.30 vs 3.56) and highest among adults aged 65 years and older (7.31), despite lower absolute prevalence compared with adults aged 50-64 years (prevalence 43.06%; AME 3.51). Between-stratum heterogeneity was substantial across intersectional strata (ΔR² 0.30-0.39). CONCLUSIONS: Associations between homelessness and inpatient alcohol disorders varied across intersecting social positions. Average population effects masked meaningful heterogeneity, highlighting the utility of an intersectionality-informed approach for characterising socially patterned differences in recorded inpatient diagnosis burden.