Exhibit 4:

Results from linear probability models examining relationships among staffing specificity scores and hospitalization of assisted living residents

	Summary specificity score	DCW specificity score	LPN specificity score	RN specificity score
Full sample of beneficiaries residing in assisted living
Model 1 (main model)	—a	−0.0018***	0.0011**	0
Models using individual specificity score
Model 2	−0.0001	—a	—a	—a
Model 3	—a	−0.0017***	—a	—a
Model 4	—a	—a	0.001*	—a
Model 5	—a	—a	—a	0
Subsample of beneficiaries with a dementia diagnosis residing in assisted living
Model 1 (main model)	—a	−0.0024***	0.002**	0.0001
Models using individual specificity score
Model 2	0	—a	—a	—a
Model 3	—a	−0.0023***	—a	—a
Model 4	—a	—a	0.0019**	—a
Model 5	—a	—a	—a	0.0003

SOURCE Authors’ analysis of regulations pertaining to staffing in assisted living, states’ assisted living licensure information, the Medicare Master Beneficiary Summary File, and Medicare Provider Analysis and Review data for a cohort of Medicare beneficiaries residing in a nine-digit ZIP code corresponding to an assisted living with twenty-five or more beds in 2007 and 2018. NOTES Data include Medicare beneficiaries residing in assisted living settings with twenty-five or more beds who contributed one or more months of data between 2007 and 2018. Total sample is 38,586,949 person-months among 1,414,792 people residing in 10,664 assisted living communities. Each reported effect represents the coefficient on staffing specificity score in a separate model that includes individual-level health and demographic covariates (see exhibit 3), year fixed effects, and assisted living community fixed effects. Robust standard errors (not shown) are estimated with clustering at the assisted living community level. DCW is direct care worker. LPN is licensed practical nurse. RN is registered nurse.

^aThe model did not include the indicated specificity score.

^*p < 0.1

^**p < 0.05

^***p < 0.01