Beyond Average: Who Benefits from Integrated Care? Heterogeneous Treatment Effects on Healthcare Utilization and Costs

Although integrated care has attracted increasing attention, there is limited causal evidence on heterogeneous treatment effects across patient groups. This gap is important for health- economic evaluation as average effects can miss clinically and economically relevant variation in healthcare utilization and cost responses. For resource-intensive integrated care programs, the economic value may depend on which patient groups experience changes in hospital-based care, outpatient utilization, and healthcare costs. We evaluate an integrated care initiative in a socially deprived urban area in Germany using claims data for 199,448 insured individuals observed from 2015 to 2024. Because enrolment in the program is staggered and selective, we estimate heterogeneous treatment effects using a staggered Difference-in-Differences Causal Forest, comparing enrolled individuals to not-yet- enrolled and never -enrolled counterparts and flexibly recovering how treatment effects vary with pre-treatment patient characteristics. Heterogeneity in treatment effects is most pronounced for inpatient costs, hospital outpatient visits, and ambulatory care-sensitive condition admissions. Estimated reductions in inpatient costs and ambulatory care-sensitive condition admissions are concentrated among patients with higher pre-treatment hospital utilization, a history of avoidable hospitalizations, and elevated inpatient costs. Hospital outpatient visits increase more among patients with greater baseline ambulatory engagement, and individuals with limited prior acute-care use exhibit smaller effects. These findings indicate that the economic implications of integrated care are contingent on how utilization and cost effects are distributed across patient groups, particularly with respect to prior acute-care use, healthcare costs, and care intensity. Health-economic evaluations of integrated care should incorporate heterogeneous treatment effect estimation to distinguish between groups for whom such interventions may reduce hospital-based care and those for whom the primary effect is increased ambulatory engagement.