Table 3.
Number (%) of trials using different methods of analysis to account for clustering
| All trials (n = 86) |
Trials with <40 clustersa
(n = 56) |
Trials with 40+ clustersa
(n = 29) |
|
|---|---|---|---|
| Cluster-level or individual-level analysis (if adjusted for clusters) | |||
| Cluster-level | 8 (9) | 6 (11) | 2 (7) |
| Individual-level | 77 (90) | 50 (89) | 26 (90) |
| Unclear | 1 (1) | 0 (0) | 1 (3) |
| Analysis type if at individual level | |||
| Mixed-effects model | 46/77 (60) | 31/50 (62) | 15/26 (58) |
| GEE | 21/77 (27) | 14/50 (28) | 6/26 (23) |
| Other | 2/77 (3) | 2/50 (4) | 0/26 (0) |
| Unclear | 8/77 (10) | 3/50 (6) | 5/26 (19) |
| Used a degree-of-freedom correction for mixed-effects model | |||
| Yes | 0/46 (0) | 0/31 (0) | 0/15 (0) |
| No | 0/46 (0) | 0/31 (0) | 0/15 (0) |
| Unclear | 46/46 (100) | 31/31 (100) | 15/15 (100) |
| Used a small-sample correction for GEEs – number (%) | |||
| Yes | 0/21 (0) | 0/14 (0) | 0/6 (0) |
| No | 0/21 (0) | 0/14 (0) | 0/6 (0) |
| Unclear | 21/21 (100) | 14/14 (100) | 6/6 (100) |
aOne trial was excluded as the number of clusters was not reported
GEE generalised estimating equation