Article type
Year
Abstract
Background: Contributorship of searching and statistical expertise is poorly reported in Cochrane reviews, even when a statement of contributorship is present.
Objectives: To examine how the contribution of the person responsible for searching was reported, whether the contribution is recognized through authorship or acknowledgement, and the association between the type of contributorship and characteristics of the search strategy including quality, reproducibility and the presence of certain types of errors.
Methods: Data sources: The Cochrane Library, Issue 3, 2002. Inclusion criteria: we included the review if one or more sections of the highly sensitive search strategy were used, if primary studies were either RCT or quasi RCT, and if included and excluded studies were listed. Data extraction: Two librarians assessed the searches for errors, establishing consensus on discordant ratings.
Results: We screened 169 reviews and 105 met all eligibility criteria. Authors fulfilled the searching role in 41.9% of reviews studies, acknowledged persons or groups in 13.3%, a combination in 9.5% and the role was not reported in 35.2% of reviews. For the subset of 78 reviews in which meta-analysis was performed, the person responsible for statistical decisions was an author in 47.4% of cases, and unreported in most other cases (47.4%). We could tell professional qualifications too infrequently to yield analyzable data (4.8%).
We were able to assess search quality for 66 searches across 74 reviews. We were more likely to be able to assess the search when the role was reported (Pearson correlation =.266, p=.006). We found an association between whether the role of searcher was reported and the number of consequential errors (chi-square = 14.851, df=4 p=.005). There was no association between the number of consequential errors and who the search was attributed to (chi-square = 21.802, df=16 p=.150).
Conclusions: Contributorship for the key roles of searching and statistical decision-making is poorly reported in Cochrane reviews, but more complete role reporting is associated with greater reproducibility of searches and fewer substantive errors in search strategies.
Objectives: To examine how the contribution of the person responsible for searching was reported, whether the contribution is recognized through authorship or acknowledgement, and the association between the type of contributorship and characteristics of the search strategy including quality, reproducibility and the presence of certain types of errors.
Methods: Data sources: The Cochrane Library, Issue 3, 2002. Inclusion criteria: we included the review if one or more sections of the highly sensitive search strategy were used, if primary studies were either RCT or quasi RCT, and if included and excluded studies were listed. Data extraction: Two librarians assessed the searches for errors, establishing consensus on discordant ratings.
Results: We screened 169 reviews and 105 met all eligibility criteria. Authors fulfilled the searching role in 41.9% of reviews studies, acknowledged persons or groups in 13.3%, a combination in 9.5% and the role was not reported in 35.2% of reviews. For the subset of 78 reviews in which meta-analysis was performed, the person responsible for statistical decisions was an author in 47.4% of cases, and unreported in most other cases (47.4%). We could tell professional qualifications too infrequently to yield analyzable data (4.8%).
We were able to assess search quality for 66 searches across 74 reviews. We were more likely to be able to assess the search when the role was reported (Pearson correlation =.266, p=.006). We found an association between whether the role of searcher was reported and the number of consequential errors (chi-square = 14.851, df=4 p=.005). There was no association between the number of consequential errors and who the search was attributed to (chi-square = 21.802, df=16 p=.150).
Conclusions: Contributorship for the key roles of searching and statistical decision-making is poorly reported in Cochrane reviews, but more complete role reporting is associated with greater reproducibility of searches and fewer substantive errors in search strategies.