Table 3.
Proportions of trials (n=34 unless otherwise stated) following procedures to enhance internal and external validity. Figures are numbers (percentages)
| Procedure | Followed procedure | Unclear if followed | Not followed |
|---|---|---|---|
| Internal validity | |||
| Accounting for clustering in sample size calculation | 21 (62) | 8 (24)* | 5 (15) |
| Accounting for clustering in analysis | 30 (88) | 2 (6) | 2 (6) |
| Protected against recruitment/identification bias when identifying/recruiting patients | 21 (62)† | 6 (18) | 7 (21) |
| Blinding of individual participants to allocation status | 19 (56) | 11 (33) | 4 (12) |
| Assessment of primary outcome blind to allocation status | 15 (44) | 13 (38) | 6 (18) |
| External validity | |||
| Full information on number of clusters approached, recruited, and analysed | 20 (59) | NA | 14 (41) |
| Comparison of characteristics of clusters recruited and those not recruited | 6 (33)‡ | NA | 12 (67) |
| Discussion of cluster generalisability | 18 (53) | NA | 16 (47) |
| Discussion of how the clusters analysed might differ from other clusters | 4 (12) | NA | 30 (88) |
| Some information about acceptability and/or feasibility | 19 (59)§ | NA | 13 (41) |
NA=not applicable (for external validity we assessed whether or not certain information was reported in trial report; by definition, it was never unclear whether information was reported).
*Five reports did not include sample size calculations; three did not provide adequate information in the sample size calculation.
†Includes those trials where we judged that selection bias was impossible or unlikely.
‡n=18 (judged only for those recruiting <99% of clusters).
§n=32 (judged only for trials in which clusters had option to opt out of intervention targeted at them or had to deliver part of intervention to patients).