Primary care connectedness and healthcare spending in Switzerland

Objectives: In healthcare systems with high provider fragmentation, integrated care could curb spending and reduce inefficiency. This paper evaluates whether primary care network topology and the connectedness of providers affect patient-level healthcare expenditures and utilization across six spending categories. Methods: Using Swiss mandatory health insurance claims from 2015 to 2024 structured into three triennia (2015—2017, 2017—2019, 2022—2024), I construct patient-sharing networks among primary healthcare providers over baseline two-year windows. Communities are detected using a spatially weighted Leiden algorithm under the Constant Potts Model objective function. Patients are assigned to communities via their main primary care provider. Imposing community stability in the first two years of the triennium, patients switching communities in year three are identified as network movers. To identify supply-side network effects. I implement a within-movers ANCOVA design evaluating the change in comm unity degree centralisation, the change in (log) degree centrality of the main primary care provider, and the change in local clustering, across both extensive (propensity to incur spending) and intensive (log expenditure) margins.Results: Within-mover estimates point in the same direction at both scales: greater connectedness is associated with lower utilisation and spending. For communities comprised of 3 or more providers, a one-standard-deviation increase in comm unity degree centralisation lowers the probability of incurring expenditure in every category examined, by between 0.5 percentage points (outpatient care) and 3.6 percentage points (emergency care), and reduces conditional spending on laboratory tests by 14.3% and on prescription drugs by 9.7%. Switching to a m ain primary care provider with higher degree centrality (i.e. one who collaborates more with other providers) likewise reduces spending across all expenditure classes, with the largest drops observed for laboratory tests ( — 1.1 p.p. extensive, —5.0% intensive per doubling of the provider’s degree). Higher local clustering around the main primary care provider reduces the propensity to incur expenses on prescription drugs and emergency care and reduces conditional drug spending, but leaves the remaining categories unaffected.Discussion : Connectedness in primary care, whether measured by the structural position of a patient’s own provider or by the coordination architecture of the community that provider belongs to, is associated with lower spending, concentrated in the diagnostic and pharmaceutical categories where duplication is easiest to avoid. Whether network centralisation is a valid cost-containment instrument should be confirmed by further accompanying evidence on quality, for example using health outcomes to distinguish lower spending from under-provision.