Skip to main content

Table 2 Properties of search strategies evaluated against the gold standard set of CRTs

From: Electronic search strategies to identify reports of cluster randomized trials in MEDLINE: low precision will improve with adherence to reporting standards

Search Strategy Total articles retrieved CRTs retrieved (Sensitivity %) (N = 162) Non-CRTs retrieved
(1-Specificity %) (N = 25545)
Precision
Existing strategy for randomized controlled trials:     
   1. randomized controlled trial.pt.     
   2. animals/     
   3. humans/     
   4. 2 NOT (2 AND 3)     
   5. 1 NOT 4 1697 152 (93.8%) 1545 (6.0%) 9.0%
Cluster design-related terms:     
   6. cluster$ adj2 randomi$.tw. 87 59 (36.4%) 28 (0.11%) 67.8%
   7. ((communit$ adj2 intervention$) OR (communit$ adj2 randomi$)).tw. 96 16 (9.9%) 80 (0.31%) 16.7%
   8. group$ randomi$.tw. 34 9 (5.6%) 25 (0.10%) 26.5%
   9. 6 OR 7 OR 8 208 78 (48.1%) 130 (0.51%) 37.5%
   10. intervention?.tw. 2636 139 (85.8%) 2497 (9.8%) 5.3%
   11. cluster analysis/ 150 30 (18.5%) 120 (0.47%) 20.0%
   12. health promotion/ 796 38 (23.5%) 758 (3.0%) 4.8%
   13. program evaluation/ 560 36 (22.2%) 524 (2.1%) 6.4%
   14. health education/ 428 32 (19.8%) 396 (1.6%) 7.5%
   15. 10 OR 11 OR 12 OR 13 OR 14 3658 149 (92.0%) 3509 (13.7%) 4.1%
   16. 9 OR 15 3680 155 (95.7%) 3525 (13.8%) 4.2%
Highest sensitivity:     
   17. 16 OR 5 4583 161 (99.4%) 4422 (17.3%) 3.5%
Highest precision:     
   18. 16 AND 5 794 146 (90.1%) 648 (2.5%) 18.4%
  1. "$" allows for truncation of words so that variations such as "randomization", "randomisation", "randomized" are included; adj refers to the adjacency operator to accommodate terms such as "community-based randomized trial"; pt refers to publication type; ? refers to optional wildcard character retrieving 1 or 0 characters;/refers to MeSH; tw refers to text words in the title and abstract.
\