Improving quality of stroke care through benchmarking center performance: why focusing on outcomes is not enough

Table 3 Results from the random effect proportional odds regression analysis using the inverse of the modified Rankin Scale at 90 days as the dependent variable

	Model 1	Model 2	Model 3	Model 4
Center volume [OR (95% CI)]	–	–	1.01 (0.98–1.04)	1.01 (0.98–1.05)
Year of admission [OR (95% CI)]	–	–
2014			ref	ref
2015			1.32 (0.95–1.85)	1.20 (0.86–1.67)
2016			1.67 (1.22–2.28)	1.36 (0.98–1.90)
2017			1.60 (1.17–2.19)	1.27 (0.91–1.77)
Time from arrival at the ED of intervention center to groin puncture (every 30 min) [OR (95% CI)]	–	–	–	0.87 (0.82–0.92)
Use of general anesthesia [OR (95% CI)]	–	–	–	0.72 (0.57–0.90)
Variance of random center intercept [tau²(95% CI)]^#	0.040 (0.012–0.113)	0.086 (0.042–0.261)	0.081 (0.028–0.227)	0.089 (0.033–0.254)
AIC	11,158	10,050	10,043	10,012

ED Emergency department, AIC Akaike’s Information Criterion. A lower AIC value indicates a better model fit
Model 1 (‘unadjusted model’) includes a random center intercept only
Model 2 (‘case-mix adjusted model’) includes a random center intercept and the case-mix indicators (Table 1). See Supplementary Figure 3 for the estimated fixed effects of each case-mix indicator
Model 3 (‘case-mix and structure indicators adjusted model’) includes a random center intercept, case-mix indicators, and structure indicators
Model 4 (‘case-mix, structure and process indicators adjusted model’) includes a random center intercept, case-mix indicators, structure indicators, and process indicators
^#95% CIs around τau² were estimated using single imputed data with bootstrap

ISSN: 1472-6963