Skip to main content
. 2021 Nov 17;7(3):e001818. doi: 10.1136/rmdopen-2021-001818

Table 2.

Summary of results for reporting and analysing for confounding

Main analysis DS1 DS2
N 211 131 115
Adjusted and crude analysis presented 181 (86%) 112 (85%) 100 (87%)
Crude analysis presented only 22 (10%) 12 (9%) 9 (8%)
Adjusted analysis presented only 8 (4%) 8 (6%) 6 (5%)
Method of selection for adjustment covariates*
A priori/wisely 114 (54%) 77 (59%) 69 (60%)
Stepwise method 37 (17%) 19 (15%) 15 (13%)
Bivariate selection 45 (21%) 25 (19%) 21 (18%)
Unknown 20 (9%) 12 (9%) 11 (10%)
Other 1 (0%) 1 (1%) 1 (1%)
Method of adjustment for confounding*
Multivariable model 146 (69%) 88 (67%) 88 (77%)
Stratification 9 (4%) 5 (4%) 5 (4%)
Matching 17 (8%) 14 (11%) 14 (12%)
Inverse probability weighting 7 (3%) 5 (4%) 5 (4%)
Propensity score 26 (12%) 23 (18%) 23 (20%)
Restriction 0 0 0
Other 1 (0%) 0 0

DS1: Dataset 1 includes only studies where the comparison of at least two treatments was the main exposure of interest (head to head studies).

DS2: Dataset 2 includes only studies from DS1, which had no other outcome than effectiveness.

*Sum of the methods may be greater than the numbers of studies as some studies used several methods.