Table 5.

Results for staff reported individual incidents.

	Staff reported individual incidents (n = 1395)
	Estimate	Std. Error	p-value
Intercept	-0.7716	0.3404	0.0283
CARE Diagnostic	-0.0313	0.0225	0.1646
Randomization Group	0.1613	0.4104	0.6944
Time	-1.5985	0.5611	0.0044
Time by Randomization Group	1.6540	0.5986	0.0058

The total number of R-REM incidents from the R-REM staff report measure was used. A higher number indicates more incidents reported by the staff. The CARE Diagnostic scale was coded in the deviant direction. Poisson regression analyses were performed with a log link function, assuming a compound symmetry covariance structure, adjusted for clustering within unit. Up to three waves of data were included in the analyses. The original sample size requirements were based on posited differences as small as 0.25 between the experimental and comparison groups in the average incident event rates per resident in one year (different λ₀, λ₁), given power =0.80; α=0.05, R=.90, Variance Inflation Factor due to clustering=1.87. This set of calculations yielded sample sizes as high as 322 per group. The observed estimated adjusted differences reflected in the analyses above were much larger: λ₀, = 0.35; λ₁ = 2.06.