Article type
Year
Abstract
Background: Systematic reviews often investigate the effectiveness of interventions for one sex but identifying reports of interventions with data presented according to the sex of study participants can be challenging due to suboptimal indexing in bibliographic databases and poor reporting in titles and abstracts.
Objectives: To investigate the performance of a range of search terms in identifying studies where data are presented specifically separately for men and women, or for men only.
Methods: Several comprehensive electronic searches were undertaken across a range of databases to inform a series of systematic reviews investigating obesity management in men. The included studies formed a reference standard set. Combinations of controlled vocabulary terms and free text search strings, designed to identify studies with male participants, were investigated in MEDLINE and Embase.
Results: The reference standard set comprised 87 papers. Searching without sex-specific terms returned 31 897 results in MEDLINE and 37 351 in Embase and identified 82% of the reference standard set in both databases. Excluding records indexed with the MESH/Emtree term ‘female’ reduced the yield by 71% (MEDLINE) and 62% (Embase) and achieved a sensitivity of 47% and 51% respectively. Excluding records indexed with the MESH/Emtree term ‘female’ but not with ‘male’ reduced the yield by 18% (MEDLINE) and 17% (Embase) and achieved 100% and 98% sensitivity respectively. The free text search string ‘male or males ormen’ reduced the yield by 76% and achieved 77% sensitivity in both databases.
Conclusions: Excluding records indexed female but not male is more sensitive than searching specifically for male. Using sex-specific index terms substantially reduces the amount of records to be screened but can compromise sensitivity and should be used with caution. Further analysis will demonstrate which combinations of sex-specific terms achieve the best balance between sensitivity and specificity.
Objectives: To investigate the performance of a range of search terms in identifying studies where data are presented specifically separately for men and women, or for men only.
Methods: Several comprehensive electronic searches were undertaken across a range of databases to inform a series of systematic reviews investigating obesity management in men. The included studies formed a reference standard set. Combinations of controlled vocabulary terms and free text search strings, designed to identify studies with male participants, were investigated in MEDLINE and Embase.
Results: The reference standard set comprised 87 papers. Searching without sex-specific terms returned 31 897 results in MEDLINE and 37 351 in Embase and identified 82% of the reference standard set in both databases. Excluding records indexed with the MESH/Emtree term ‘female’ reduced the yield by 71% (MEDLINE) and 62% (Embase) and achieved a sensitivity of 47% and 51% respectively. Excluding records indexed with the MESH/Emtree term ‘female’ but not with ‘male’ reduced the yield by 18% (MEDLINE) and 17% (Embase) and achieved 100% and 98% sensitivity respectively. The free text search string ‘male or males ormen’ reduced the yield by 76% and achieved 77% sensitivity in both databases.
Conclusions: Excluding records indexed female but not male is more sensitive than searching specifically for male. Using sex-specific index terms substantially reduces the amount of records to be screened but can compromise sensitivity and should be used with caution. Further analysis will demonstrate which combinations of sex-specific terms achieve the best balance between sensitivity and specificity.