marginally statistically significantP = 0.0478
The random coefficient model in which both the intercept and the effect of the linear predictor were allowed to vary across hospitals (Model 4) was found to provide a marginally statistically significant improvement in fit compared to the prior model in which only the intercept varied across hospitals (likelihood ratio test: χ 2 = 6.08 ( df = 2), P = 0.0478).