Informing the Design and Evaluation of Superuser Care Management Initiatives: Accounting for Regression-to-the-Mean.

Abstract

Health care spending is concentrated among a small number of high-cost patients, and the popularity of initiatives to improve care and reduce cost among such "superusers" (SUs) is growing. However, SU costs decline naturally over time, even without intervention, a statistical phenomenon known as regression-to-the-mean (RTM). We assess the magnitude of RTM in hospital costs for cohorts of hospital SUs identified on the basis of high inpatient (IP) or emergency department (ED) utilization. We further examine how cost and RTM are associated with patient characteristics including behavioral health (BH) problems, multiple chronic conditions, and indicators of vulnerability. Using longitudinally linked all-payer hospital billing data, we selected patient cohorts with ≥2 IP stays (IP SUs) or ≥6 ED visits (ED SUs) during a 6-month baseline period, and additional subgroups defined by combinations of IP and ED superuse. A total of 289,060 NJ hospital IP and treat-and-release ED patients over 2009-2011. Hospital costs among IP and ED SUs declined 70% and 38%, respectively, over 8 quarters following the baseline period. The decrease occurs more quickly for IP SUs compared with ED SUs. Presence of BH problems was positively associated with costs among patients overall, but the relationship varied by SU cohort. Understanding patterns of RTM among SU populations is important for designing intervention strategies, as there is greater potential for savings among patients with more persistent costs (less RTM). Further, as many SU initiatives lack resources for rigorous evaluation, quantifying the extent of RTM is vital for interpreting program outcomes.