Effects of a Community-Based Care Management Program on Utilization and Spending Among High-Utilizers

Background

High-utilizers are among the 5 percent of patients that account for more than half of the US health care expenditure. Many health systems continue to experiment with new ways to care for these high-risk, high-cost patients. The purpose of this study was to evaluate the effectiveness of a high-intensity care management program in Rhode Island, in which a Community Health Team (CHT), consisted of social workers, behavioral health specialists, and nurse care managers, provided both medical and social assistance to high-utilizers.

Methods

Using the state All-Payer Claims Database, we identified 2,282 participants who ever enrolled in the CHT program between 2014 and 2017. A total of 10,514 control patients were selected based on propensity scores constructed from baseline demographic characteristics (age, gender, ZIP Code-level poverty rate, and Charlson comorbidities), insurance status (Medicaid coverage, Medicaid eligibility basis, Medicare coverage, and Dual Eligibility), and past utilization trends (monthly ED visit rates, monthly hospitalization rates, and 3 or more ED visits or hospitalizations in the past 6 months).

To estimate the impact of the CHT program, we used linear regression models with a difference-in-difference framework to compare outcomes between program participants and control patients, 12 months before and 24 months after program enrollment. The outcomes of interest were rates and spending of ED visits and hospitalizations. The unit of analysis was person-month. For each model, a binary treatment variable indicated whether a patient ever enrolled in the CHT program. Binary post-intervention variables indicated whether an observation was 0 (prior to CHT enrollment), or 1 or 2 years after enrollment. The estimates of interest were the interaction terms between the treatment and post-intervention indicators. All models were adjusted for the demographic and coverage variables described above. Standard errors were corrected at person level. To examine if treatment effects varied among participants with different treatment intensities, we stratified participants into 3 subgroups based on the number of CHT staff visits they had: Subgroup 1 had 1-2 CHT visits (0-50 percentile); Subgroup 2 had 2-6 CHT visits (50-75 percentile); Subgroup 3 had more than 6 CHT visits (above 75 percentile).

Results

During the 24-month follow-up period, we did not find significant changes to overall utilization and spending of ED visits and hospitalizations among CHT participants. However, we found substantial variation in treatment effects when we stratified the participants by the number of face-to-face visits they had with CHT staff. Among Subgroup 1, there was a significant decrease in ED visits (DID: -0.026; p-value: 0.018), ED cost (DID: $-10; p-value: 0.08), and hospitalizations (DID: -0.011; p-value: 0.004) during Year 1 after program enrollment. Some of these benefits continued into Year 2. The same effects were not observed among Subgroups 2 or 3, suggesting that the needs among the subgroups may be different.

Conclusions

While we did not find any overall impact of the CHT program on participants, we found that the treatment effects varied among subgroups, suggesting that care management programs may need to further tailor to the needs of their participants.