EXHIBIT 4. Relationships among receipt of Veteran-Directed Care (VDC) and changes in hospital use and costs over time.

SOURCE Authors’ analysis of aggregated encounter and all-cause hospitalization cost data for 2015–18 from the Veterans Health Administration (VHA). NOTES Costs include only those attributed to an acute hospitalization in the VHA. We regressed outcomes on the indicators for receipt of VDC in the post period, indicators for receipt of other purchased care services in the post period in comparison sites without an active VDC program in fiscal year 2017, and month. In matched models, we used coarsened exact matching to create groups of recipients and nonrecipients of VDC with similar sociodemographic and clinical characteristics. Full details of the models are in the online appendix (see note 26 in text). CI is confidence interval.

	Hospital admissions		Hospitalization costs ($)
Model	Odds ratio	95% Cl	Average incremental effect	95% Cl
ALL-CAUSE HOSPITALIZATIONS
Unmatched fixed effects	0.83	(0.68, 1.02)*	−358	(−687, −29)**
Matched fixed effects	0.86	(0.75, 0.99)**	−274	(−468, −79)***
Unmatched population-averaged	0.92	(0.80, 1.06)	−149	(−452, 153)
Matched population-averaged	0.90	(0.81, 1.01)*	−163	(−325, −2)**
AMBULATORY CARE-SENSITIVE HOSPITALIZATIONS
Unmatched fixed effects	1.09	(0.71, 1.39)	—a	—a
Matched fixed effects	1.15	(0.83, 1.58)	—a	—a
Unmatched population-averaged	1.03	(0.76, 1.40)	—a	—a
Matched population-averaged	1.05	(0.83, 1.33)	—a	—a

^aSparse data on ambulatory care–sensitive hospital admission costs precluded multivariable analyses.

^*p < 0:1

^**p < 0:05

^***p < 0:01