Table 3.

Comparison of previously published search strategies with search strategies developed our database. Values are percentages.

Search Strategy OVID search*	Sensitivity	Specificity	Precision	Accuracy
Bachmann's most sensitive search [22] sensitiv:.tw. OR detect:.tw. OR accura:.tw. OR specific:.tw. OR reliab:.tw. OR positive.tw. OR negative.tw. OR diagnos:.tw.	100.0	Not reported		Not reported
Bachmann's strategy tested in our database	96.9	72.3	1.2	72.4
Our most sensitive search di.fs. OR predict:.tw. OR specificity.tw.	100.0	70.4	1.2	70.5
Difference (our strategy – Bachmann) 95% CI for the difference	3.1 -0.8 to 8.7†	-1.9 -2.7 to -1.2‡	0	-1.9 -2.7 to -1.2‡
Bachmann's most specific search sensitiv:.tw. OR detect:.tw.	73.7	Not reported		Not reported
Bachmann's strategy tested in our database	79.4	90.9	3.0	90.8
Our most specific search specificity.tw.	62.9	98.2	11.0	98.1
Difference (our strategy – Bachmann) 95% CI for the difference	-16.5 -28.9 to -3.8‡	7.3 7.0 to 7.7‡	8.0 5.6 to 11.0‡	7.3 6.9 to 7.7‡

*Search strategies are reported using Ovid's search engine syntax for EMBASE. †Differences are not statistically significant. ‡Differences are statistically significant. : = truncation; tw = textword (word or phrase appears in title or abstract); di = diagnosis; fs = floating subheading. Sensitivity = the proportion of high quality articles for that topic that are retrieved; specificity = the proportion of low quality articles not retrieved; precision = the proportion of retrieved articles that are of high quality; accuracy = the proportion of all articles that are correctly classified.