High-cost high-need patients are typically defined by risk or cost thresholds which aggregate clinically diverse subgroups into a single 'high-need high-cost' designation. Programs have had limited success in reducing utilization or improving quality of care for high-cost high-need Medicaid patients, which may be due to the underlying clinical heterogeneity of patients meeting high-cost high-need designations.
Our objective was to segment a population of high-cost high-need Medicaid patients (N = 676,161) eligible for a national complex case management program between January 2012 and May 2015 to disaggregate clinically diverse subgroups. Patients were eligible if they were in the top 5 % of annual spending among UnitedHealthcare Medicaid beneficiaries. We used k-means cluster analysis, identified clusters using an information-theoretic approach, and named clusters using the patients' pattern of acute and chronic conditions. We assessed one-year overall and preventable hospitalizations, overall and preventable emergency department (ED) visits, and cluster stability.
Six clusters were identified which varied by utilization and stability. The characteristic condition patterns were: 1) pregnancy complications, 2) behavioral health, 3) relatively few conditions, 4) cardio-metabolic disease, and complex illness with relatively 5) low or 6) high resource use. The patients varied by cluster by average ED visits (2.3-11.3), hospitalizations (0.3-2.0), and cluster stability (32-91%).
We concluded that disaggregating subgroups of high-cost high-need patients in a large multi-state Medicaid sample identified clinically distinct clusters of patients who may have unique clinical needs. Segmenting previously identified high-cost high-need populations thus may be a necessary strategy to improve the effectiveness of complex case management programs in Medicaid.