Article type
Year
Abstract
Background:
Systematic reviews (SRs) that synthesise prognosis research are increasingly important due to rising numbers of, and evidence regarding, people with long-term health conditions. Prognosis SRs are directly relevant to patients, clinicians and policymakers who need high-quality information regarding likely future outcomes on which to base healthcare decisions. Search strategies are challenging due to poor reporting of prognosis studies and inconsistent electronic database indexing. Performance metrics of prognosis search filters (hedges) in published development and evaluation studies show high sensitivity but poor precision. Consequently, researchers conducting prognosis SRs spend considerable time screening titles and abstracts, which delays completion of SRs and increases the funding required.
Objective:
To evaluate the efficiency of the most sensitive search filters in Medline and EMBASE (Ovid) for identifying prognosis studies in a SR and assess the impact on screening burden for reviewers.
Methods:
A recent SR identified development and evaluation studies of prognosis search filters. We will combine those with sensitivity > 95%, including Clinical Queries and Yale filters, with subject-specific terms in a search strategy for a SR of predictors of unplanned hospital admissions in older people. We will use a 'super' filter including all terms from the individual filters as a comparison of efficiency.
Results:
Searches for this SR will begin in March 2018. We will calculate relative recall, the proportion of the studies included in the SR retrieved by each search filter combined with the same set of subject terms. We will note the total number of references that need to be screened and calculate the number needed to read (NNR) for each search filter combined with subject terms. We will compare metrics from the SR for each filter to those published in their original development or evaluation studies. We will present the results.
Conclusions:
We will share the results on the efficiency of currently available search filters with the Cochrane Prognosis Group. This work forms first stage in a PhD that aims to test and, if needed, develop search filters to improve identification of prognosis studies for SRs.
Patient involvement:
Keele Research User Group were involved in the conduct of the SR of unplanned hospital admissions.
Systematic reviews (SRs) that synthesise prognosis research are increasingly important due to rising numbers of, and evidence regarding, people with long-term health conditions. Prognosis SRs are directly relevant to patients, clinicians and policymakers who need high-quality information regarding likely future outcomes on which to base healthcare decisions. Search strategies are challenging due to poor reporting of prognosis studies and inconsistent electronic database indexing. Performance metrics of prognosis search filters (hedges) in published development and evaluation studies show high sensitivity but poor precision. Consequently, researchers conducting prognosis SRs spend considerable time screening titles and abstracts, which delays completion of SRs and increases the funding required.
Objective:
To evaluate the efficiency of the most sensitive search filters in Medline and EMBASE (Ovid) for identifying prognosis studies in a SR and assess the impact on screening burden for reviewers.
Methods:
A recent SR identified development and evaluation studies of prognosis search filters. We will combine those with sensitivity > 95%, including Clinical Queries and Yale filters, with subject-specific terms in a search strategy for a SR of predictors of unplanned hospital admissions in older people. We will use a 'super' filter including all terms from the individual filters as a comparison of efficiency.
Results:
Searches for this SR will begin in March 2018. We will calculate relative recall, the proportion of the studies included in the SR retrieved by each search filter combined with the same set of subject terms. We will note the total number of references that need to be screened and calculate the number needed to read (NNR) for each search filter combined with subject terms. We will compare metrics from the SR for each filter to those published in their original development or evaluation studies. We will present the results.
Conclusions:
We will share the results on the efficiency of currently available search filters with the Cochrane Prognosis Group. This work forms first stage in a PhD that aims to test and, if needed, develop search filters to improve identification of prognosis studies for SRs.
Patient involvement:
Keele Research User Group were involved in the conduct of the SR of unplanned hospital admissions.