Baseline hospital performance and the impact of medical emergency teams: Modelling vs. conventional subgroup analysis

Table 3 Weighted quadratic or linear regression models to predict changes in incidence of primary and secondary outcomes during the study period

	Change of primary outcome			Change of unexpected cardiac arrests		Change of unplanned ICU admission		Change of unexpected death
	Control: quadratic effect	Control: linear effect	MET ^†	Control	MET ^†	Control	MET ^†	Control: quadratic effect	Control: linear effect	MET ^†
Baseline incidence (per 1000 admissions)	-2.168 (0.017)*	-0.135 (0.562)	-0.592 (0.001)**¶	-0.945 (<0.001)**	-0.736 (<0.001)**	-0.161 (0.313)	-0.556 (0.002)**	-2.076 (0.002)**	-1.039 (<0.001)**	-0.676 (<0.001)**
Baseline incidence squared	0.125 (0.020)*							0.256 (0.048)*
Constant	6.632 (0.041)*	-0.518 (0.753)	2.789 (0.005)**	1.487 (0.009)**	0.932 (0.002)**	0.215 (0.804)	2.214 (0.007)**	2.046 (0.001)**	1.182 (0.002)**	0.587 (0.004)**
R-squared	0.532	0.039	0.710	0.782	0.851	0.112	0.643	0.923	0.870	0.819

Note: P values (in the parentheses) for the regression coefficients and the constant; control hospitals showed a quadratic effect for only the primary outcome and unexpected deaths.
* P < 0.05; ** P < 0.01;
¶The sensitivity analysis by removing the two hospitals with highest baseline incidence in MET hospitals produced a regression coefficient for baseline incidence as -0.416 with p = 0.034;
^†Only the MET hospitals showed a linear effect for all of the four outcomes.

ISSN: 1745-6215