Article type
Year
Abstract
Background: The issue of focusing a systematic review, and thereby choosing the right question, can be problematic from both the reviewer's and consumer's point of view. Variation exists in whether different types of outcomes, patients (e.g. age category, risk level), and interventions are included within a single review (lumping) or are divided into several related, but individual reviews (splitting). From a decisionmaking point of view, clinicians and consumers may find very specific reviews helpful in answering specific clinical questions. However, when looking for a range of information on broader topics, reviews that are grouped or lumped together but still contain relevant subgroups may be more useful. Objective: To gather empirical evidence of the extent to broad or narrow questions are addressed within perinatal systematic review topics.
Methods: All (n=221) completed reviews undertaken by the Pregnancy and Childbirth and Neonatal Review Groups in the Cochrane Database of Systematic Reviews (Issue 4, 1999) were examined. A classification system was developed for assessing the degree of lumping or splitting of review topics in relation to patient problems, outcome measures, participants and intervention types.
Results: Most reviews (94%) included all relevant outcome measures of interest within the review. The majority of reviews (94%) attempted to include a wide range of patients or participants on whom the intervention(s) could appropriately be used. However, of these reviews (n=208), only a minority (n= 45) analysed the results by relevant patient subgroups. Only 12% of reviews compared a variety of interventions for a particular clinical problem (either against each other or against a placebo) within a single review. 17% of reviews addressed interventions that related to multiple patient problems within the one review. Sub group analysis by either: study type (sensitivity analysis), intervention type, participants, or outcomes was undertaken in 44% of reviews (n = 98). Of the reviews that undertook any form of subgroup analysis, only 31% (n = 31) stated the intended subgroups a priori.
Conclusions and recommendations: It seems appropriate to include all relevant outcomes pertaining to an intervention comparison within the scope of one review. Whilst this was attempted in most reviews, it was difficult to assess a complete listing of the desired versus available outcomes either in the text of the review or on the Metaview graphs. The degree of subgroup analysis by relevant patient subgroups was surprising low. A review that compares the effects of several interventions relating to a particular patient problem may be desirable from an end-users point of view. However, for conditions where there are many treatment options, such a task may be large and difficult to manage. In these situations it may be necessary to undertake a series of reviews looking at a variety of interventions and bring them together in an overview that end-users would find easily accessible. In addition, improved search engines and/or keyword systems would help locating all the relevant reviews of interest.
Methods: All (n=221) completed reviews undertaken by the Pregnancy and Childbirth and Neonatal Review Groups in the Cochrane Database of Systematic Reviews (Issue 4, 1999) were examined. A classification system was developed for assessing the degree of lumping or splitting of review topics in relation to patient problems, outcome measures, participants and intervention types.
Results: Most reviews (94%) included all relevant outcome measures of interest within the review. The majority of reviews (94%) attempted to include a wide range of patients or participants on whom the intervention(s) could appropriately be used. However, of these reviews (n=208), only a minority (n= 45) analysed the results by relevant patient subgroups. Only 12% of reviews compared a variety of interventions for a particular clinical problem (either against each other or against a placebo) within a single review. 17% of reviews addressed interventions that related to multiple patient problems within the one review. Sub group analysis by either: study type (sensitivity analysis), intervention type, participants, or outcomes was undertaken in 44% of reviews (n = 98). Of the reviews that undertook any form of subgroup analysis, only 31% (n = 31) stated the intended subgroups a priori.
Conclusions and recommendations: It seems appropriate to include all relevant outcomes pertaining to an intervention comparison within the scope of one review. Whilst this was attempted in most reviews, it was difficult to assess a complete listing of the desired versus available outcomes either in the text of the review or on the Metaview graphs. The degree of subgroup analysis by relevant patient subgroups was surprising low. A review that compares the effects of several interventions relating to a particular patient problem may be desirable from an end-users point of view. However, for conditions where there are many treatment options, such a task may be large and difficult to manage. In these situations it may be necessary to undertake a series of reviews looking at a variety of interventions and bring them together in an overview that end-users would find easily accessible. In addition, improved search engines and/or keyword systems would help locating all the relevant reviews of interest.