Table 3.

Parameter estimates of mixed models for physical component summary (PCS) fitted in four polypharmacy strata for people with stable/decrease in comorbidities over four years

	No polypharmacy N = 913			Baseline only polypharmacy N = 87			Incident polypharmacy N = 130			Persistent polypharmacy N = 280
	β (SE)	P value	95% CI	β (SE)	P value	95% CI	β (SE)	P value	95% CI	β (SE)	P value	95% CI
Fixed effects
Intercept	45.3 (0.3)	< 0.001	44.7, 46.0	40.9 (1.3)	< 0.001	38.3, 43.6	42.5 (1.0)	< 0.001	40.6, 44.3	38.7 (0.9)	< 0.001	37.0, 40.5
Slope (per year)	− 0.2 (0.1)	0.009	− 0.3, − 0.1	− 0.3 (0.2)	0.177	− 0.7, 0.1	− 0.9 (0.2)	< 0.001	− 1.3, − 0.5	− 0.2 (0.2)	0.214	− 0.5, 0.1
Random effects
Variance in intercept	55.8 (3.03)	< 0.001	50.1, 62.0	84.9 (14.1)	< 0.001	61.4, 117.5	83.2 (12.0)	< 0.001	62.7, 110.3	70.5 (6.6)	< 0.001	58.6, 84.8
Variance in slope	0.7 (0.2)	< 0.001	0.5, 1.1	0.18 (0.5)	0.696	0.001, 27.8	2.01 (0.6)	0.001	1.1, 3.7	0.4 (0.3)	0.238	0.1, 1.7

Intercepts and slopes for all models are adjusted for age, sex, education, social life, marital status, Socio-Economic Indexes for Areas decile, physical activity, smoking and alcohol intake, and baseline comorbidities; All covariates were mean centred and included as intercept and slope predictors; PCS: physical component summary; Polypharmacy defined as ≥ 5 medications. Covariate estimates are reported in Table S11