Table 3.

Estimates of odds ratios (OR) from multilevel regression (binomial logit) model for the association between introduction of named GP and risk of an emergency hospital admission, England 2012 to 2016 (38 500 observations)

	Unadjusted model			Adjusted model 1*			Adjusted model 2†
	OR	P value	95% CI	OR	P value	95% CI	OR	P value	95% CI
Period (ref.=pre)	1.206	<0.001	1.111 to 1.309	1.156	0.001	1.064 to 1.257	1.137	0.007	1.035 to 1.254
Age (ref.=<75)	1.887	<0.001	1.736 to 2.048	1.594	<0.001	1.464 to 1.735	1.680	<0.001	1.530 to 1.846
Period* Age	1.179	0.003	1.059 to 1.314	1.191	0.002	1.066 to 1.330	1.228	0.001	1.086 to 1.388

*Covariates set to average: gender, number of chronic conditions, level of deprivation (quintiles), number of GPs in practice (quartiles) and urban/rural practice location. For complete table with B-coefficients, see online supplementary table 3.

†Included also standardised covariates: number of GP consultations (quartiles) and continuity of care (Bice & Boxerman index-score, quartiles) till first emergency hospital admission or the end of the observation period when not admitted. Number of observations: 37 207.

GP, general practitioner.