Table 2.

Pre-Post Differences for Service Utilization for BHH and non-BHH Patients with Propensity Score Weighting

	Contrast (BHHPost − ControlPost) − (BHHPre − ControlPre)	Delta SE	p-value	*<.10; **<.05
Service Utilization
% Patients with Any ED Use	.011	.033	.728
Total ED visits	−.428	.174	.014	**
Num. ED visits (excluding zeros)	−.618	.221	.005	**
% Patients with any psychiatric hospitalizations	−.030	.021	.148
Total psychiatric hospitalizations	−.125	.040	.002	**
Num. psychiatric hospitalizations (excluding zeros)	−.685	.210	.001	**
% Patients with any medical hospitalizations	−.005	.025	.826
Total medical hospitalizations	−.067	.051	.182
Num. medical hospitalizations (excluding zeros)	−.292	.178	.101

Note. BHH = Behavioral Health Home Program; PS = propensity score.

Estimated using GEE models (Stata xtgee and margins commands; logit link for binary outcomes and log link and gamma family variance for continuous outcome measures). Models were adjusted for propensity score weights using the Stata svy command. Propensity scores included the following covariates: sex; age; race/ethnicity (non-Hispanic White, non-Hispanic black, Asian, Hispanic); non-native English speaker; Medicare enrollment; Medicaid enrollment; diagnosis of diabetes or bipolar disorder; area-level variables (percent foreign-born, percent living below federal poverty level (FPL), percent with female head of household, percent less than high school graduate).