Development of a segmentation model to identify healthcare demand and related costs attributable to population's characteristics

OBJECTIVES. The assessment of individuals’ demand and related costs can guide to allocate healthcare resources and improve access to healthcare. We created a segmentation demand model and we estimated healthcare costs and their determinants within specific and distinguished groups of general population. METHODS. The target population included subjects living in an area of northern Italy registered at one Local Healthcare Unit in 2005 (1,031,684 subjects). On the basis of clinical judgments and literature, we identified eight different segments: subjects unknown to HS, maternity and infancy, elderly, people with one chronic disease (CD), people with more CDs, people with probable or not severe CDs, subjects with acute event, healthy people. To describe these groups and their health demand, we used demographic and healthcare demand data (hospital admissions, drug’s prescriptions, medical specialist visits, diagnostic tests) from administrative databases available at the Lombardy Health System (HS). These were merged adopting probabilistic record linkage (DENALI Data Warehouse) to optimize correct matching of data. RESULTS. Overall, healthcare cost €834 million in 2005 (€809 per-capita). Healthy people were 53% of the sample and cost €180 per-capita. Next more frequent subjects were those with one CD (14%) and cost €916 per-capita, then those with more CDs (13%) who cost €3,457 per-capita. Hospitalizations were the cost driver in five segments (maternity and infancy, elderly, one CD, more CDs, acute event), accounting for 42-89% of total costs. Diagnostic tests and medical visits contributed to 42-45% of total costs among healthy subjects and those with probable CD. Overall, drugs accounted for 24% of costs of sample. CONCLUSIONS. This pioneer demand segmentation model shows an example of how merging different administrative databases makes possible understanding effects of their characteristics on healthcare demand and costs. Research is encouraged to improve the model and study specific variants that can be applied in different healthcare contexts.