Patients with alcohol use disorders (AUD) contribute to overall higher levels of health service utilization (HSU), including acute care in general medical hospitals (White et al., 2018) and significant proportions of emergency department (ED) encounters and hospitalizations (Chavez et al., 2016, Suen et al., 2022). In the U.S., the frequency of AUD-related HSU in the ED has risen to 61 % over the past decade, even as overall rates of AUDs have decreased (White et al., 2018). AUDs are often present in patients with both the highest rates of healthcare service use and the lowest access to healthcare services outside of the ED, where use is disproportionate (Rinehart et al., 2018).

In the U.S., AUD-related HSU in general medical hospital settings often consist of “treat-and-release” visits, characterized by ED encounters, minimal medical complications, and limited options for non-medical treatment (Hsia and Niedzwiecki, 2017, Schmidt et al., 2018, Sutton and Jutel, 2016, Wu et al., 2018). Hospitals may be disincentivized to allocate resources towards the treatment of AUDs in the absence of a medical or psychiatric diagnosis for which treatment costs which are reimbursable. Decisions to allocate resources (i.e., staff time, funding) based on the severity of medical needs may fail to address the diverse clinical needs of patients with AUDs.

AUDs have been shown to include highly heterogenous patient subgroups. Subgroups have been identified based on patterns of alcohol consumption, comorbidities, and sociodemographic characteristics in community samples (Casey et al., 2013, Jackson et al., 2014, Litten et al., 2015, Müller et al., 2020, Moss et al., 2007) and samples drawn from addiction treatment centers (Fernandez et al., 2019). This heterogeneity may limit the utility of existing models to predict HSU. Previous models have been criticized for limited incorporation of sociodemographic factors, despite clear associations between factors such as ethnicity and homelessness with AUD and HSU (Mahmoudi et al., 2020). Given the heterogeneity among patients with AUD, both in terms of the type and severity of clinical needs and potential structural and economic barriers to treatment, a single uniform pattern of HSU for this class of disorders is unlikely.

Patient profile heterogeneity and relations to HSU can be understood within the Andersen model of HSU (Andersen, 1995), a widely used theoretical framework to study health services. The Andersen model presents HSU as a function of a patient’s predisposition to use services (e.g., sociodemographic and health behavior variables), their clinical needs for care (e.g., diagnosis), and their ability (or inability) to access services (e.g., insurance status, types of available services, service settings; Andersen, 1995; Babitsch et al., 2012). These factors operate together to shape differences in patterns of HSU.

Some studies have deployed cluster analysis to identify distinct and readily interpretable subgroups of patients based on a wide range of sociodemographic and clinical factors and examined their relations to patterns of HSU (Fleury, Grenier, Cao and Huỳnh, 2022; Fleury et al., 2019). Fleury et al. (2022)). For example, Fleury et al. (2022) found that patterns of HSU differed among patients with cannabis-related disorders: some patients relied on the ED for care; others had high HSU across multiple healthcare settings. Few studies have examined rates of service use for any cause among AUD patients presenting to general medical hospitals in the U.S., settings where AUD patients are increasingly likely to receive services.

Understanding this heterogeneity in patient profiles and patterns of HSU may help to address gaps in this literature. Prior studies of predictors of HSU have focused on outcomes including single instances of hospital readmission, typically within 30 days (e.g., Zeyu, et al., 2023). However, this focus on hospitalization re-admission may fail to account for patients with high frequency of visits (i.e., re-encounters) in the ED or other acute care settings.

Models accounting for the frequency (i.e., number of hospital visits) and acuity (i.e., total days spent using services) are commonly used in actuarial research and can clarify trajectories of HSU (e.g., Kurz, 2020; Frees et al., 2014, Lally and Hartman, 2016). These models may be especially helpful when applied to heterogenous patient populations with limited access to care, such as individuals with AUD who are served at publicly funded medical hospitals (Kurz, 2020). To our knowledge, only one study in the U.S has examined both frequency and acuity of HSU using hospital records (Zeyu, et al., 2023). The resulting model found the strongest predictors related to sociodemographic characteristic and previous HSU.

Building on work by Fleury et al. (2022) and others, we apply the Andersen framework and cluster analysis to examine patterns of HSU among patients with AUDs accessing services in a general hospital setting. The aim of the current study is two-fold. First, we aim to model the heterogeneity of patients with AUD using clustering methods adopted from previous studies. Second, we will predict patterns of HSU over time (i.e., at 30 days and 12 months) using subgroups identified through clustering. We will examine the frequency of hospital readmission & ED reencounter while also comparing differences in the acuity of HSU, or the amount of services received, over this period. We will test the hypothesis that repeated unmet needs (vs. the severity of needs at a single encounter) may be a strong driver of service use. Understanding differences among subgroups of patients with AUD in their sociodemographic profiles, clinical needs and patterns of HSU can improve the quality of patient care and guide efforts to allocate hospital resources. Better targeting of resources may reduce ED overcrowding and excess burden on clinical staff.