Table 1

Results from mixed-effects linear regression models: time trends in the effectiveness of five strategies*

Analysis	No of studies	No of effect sizes	Latest month of follow-up	Coefficient	P value	95% CI
(A) Training alone
ITS studies only	2	29	11	NA	NA	NA	NA
i. ITS+non-ITS studies with >2 follow-up measurements	17	192	30	−0.19	0.023	−0.36	−0.03
Sensitivity analyses
ii. ITS+non-ITS studies with >2 follow-up measurements from >3 studies†	16	177	10.5	−0.23	0.228	−0.59	0.14
iii. ITS+all non-ITS studies	65	549	30	−0.35	0.001	−0.56	−0.14
(B) Supervision alone
i. ITS studies only	3	39	33	0.88	<0.0001	0.72	1.03
ii. ITS+non-ITS studies with >2 follow-up measurements	5	47	33	0.86	<0.0001	0.67	1.04
Sensitivity analyses
iii. ITS+non-ITS studies with >2 follow-up measurements from >3 studies†	5	33	6	−0.54	0.652	−2.89	1.81
iv. ITS+all non-ITS studies	16	90	33	0.82	<0.0001	0.58	1.05
(C) Training plus supervision
ITS studies only	0	0	NA	NA	NA	NA	NA
i. ITS+non-ITS studies with >2 follow-up measurements	5	90	16	−0.40	0.005	−0.68	−0.12
Sensitivity analyses
ii. ITS+non-ITS studies with >2 follow-up measurements from >3 studies†	5	82	6	−0.52	0.211	−1.35	0.30
iii. ITS+all non-ITS studies	22	156	16	−0.46	0.200	−1.17	0.25
(D) Group problem-solving alone
i. ITS studies only	8	230	34	0.50	<0.0001	0.37	0.64
ii. ITS+non-ITS studies with >2 follow-up measurements	8	230	34	0.50	<0.0001	0.37	0.64
Sensitivity analyses
iii. ITS+non-ITS studies with >2 follow-up measurements from >3 studies†	8	225	30	0.53	<0.0001	0.39	0.67
iv. ITS+all non-ITS studies	12	245‡	34	0.46	<0.0001	0.31	0.61
(E) Group problem-solving plus training
i. ITS studies only	4	199	24	−0.13	0.485	−0.51	0.24
ii. ITS+non-ITS studies with >2 follow-up measurements	4	199	24	−0.13	0.485	−0.51	0.24
Sensitivity analyses
iii. ITS+non-ITS studies with >2 follow-up measurements from >3 studies†	4	116	12.5	−0.45	0.259	−1.23	0.33
iv. ITS+all non-ITS studies	4	199	24	−0.13	0.485	−0.51	0.24

Roman numerals number each regression analysis performed and are referred to in the STATA code and diagnostic plots on GitHub.
Coefficient is the model coefficient, which is the mean percentage-point change in healthcare provider practice outcomes per month.
*Estimates from random-intercept linear regression models adjusted for healthcare provider baseline performance.
†The sensitivity analysis used the ‘ITS+non-ITS studies with >2 follow-up measurements’ dataset, but only included effect sizes up through the latest follow-up time that involved three or more studies.
‡One effect size was missing baseline performance and was excluded from the regression modelling.
ITS, interrupted time series; NA, not available.