	Patient-specific scales

Outcome	Square root of UK-MAI		SF-36 mental component score		SF-36 physical component score
Effect	Estimate (SE)	P-value	Estimate (SE)	P-value	Estimate (SE)	P-value

Intercept	1.445 (1.195)	0.215	63.4 (8.00)	<0.001	63.4 (6.98)	<0.001

PCT A	0.645 (0.220)	0.004	–4.36 (1.56)	0.005	–1.95 (1.37)	0.155

PCT C	1.380 (0.207)	<0.001	–0.58 (1.36)	0.667	–0.80 (1.19)	0.504

PCT D	0.603 (0.184)	0.001	–0.41 (1.29)	0.754	0.035 (1.13)	0.975

PCT E	0.758 (0.221)	0.001	–4.36 (1.57)	0.006	–1.26 (1.37)	0.359

PCT B	0	–	0	–	0	–

Drugs	0.406 (0.011)	<0.001	Not fitted	–	Not fitted	–

Age	–0.009 (0.015)	0.545	–0.176 (0.099)	0.076	–0.370 (0.086)	<0.001

Intervention	0.078 (0.093)	0.402	–0.500 (0.684)	0.465	0.888 (0.471)	0.060

Time	–0.004 (0.003)	0.284	–0.057 (0.027)	0.039	–0.039 (0.024)	0.074

Time–intervention	–0.008 (0.005)	0.080	0.004 (0.036)	0.901	–0.039 (0.024)	0.104
Number of episodes in secondary care per month

	Intervention		Time–intervention

Outcome	Estimate (SE)	P-value	Estimate (SE)	P-value

Total secondary episodes	–0.042 (0.058)	0.464	0.004 (0.007)	0.561

Emergency admission episodes	0.049 (0.290)	0.865	–0.042 (0.038)	0.267

Outpatient episodes	–0.085 (0.077)	0.269	0.016 (0.010)	0.099

Inpatient episodes	–0.081 (0.162)	0.616	–0.016 (0.021)	0.436

As dependent variables in the lower part of the table counted the number of episodes per month, results for Poisson regression models are displayed. The study also estimated more complex negative binomial models to account for an excess of zeros, but this hardly changed parameter estimates or improved the fit of the model. Additional covariates (not shown) were: sex, age, number of repeat prescriptions, and the number of months since baseline and the square of this variable.
UK-MAI = UK Medication Appropriateness Index. SF-36 = Short Form–36 Health Survey. SE = standard error.