Article type
Year
Abstract
Background: The EPPI-Centre maintains the specialised register of trials (TRoPHI) for the Cochrane Collaboration Health Promotion and Public Health (HP&PH) Field. Trials are sourced both through EPPI-Centre reviews, and quarterly updates using highly sensitive searches of PubMed, EMBASE and CENTRAL. Searching for HP&PH trials, and other study designs, poses particular difficulties. Public health is a broad topic area with varied definitions, and there are related problems with accuracy of public health indexing in bibliographic databases. Current search terms describe the concepts of public health and study design (RCT and CCT). Analysis of previous searches reveals that:
The exploded MeSH term, 'Public Health' includes terms relating to study design, e.g. 'double-blind methodÂ’ and 'prospective studiesÂ’, resulting in a high percentage of records recalled on the basis of study design alone, and not relevant to public health
Many records are grouped around areas not relevant to public health, e.g. surgical techniques, drug efficacy trials.
Several modifications to deal with these issues have been considered, but uncertainty exists regarding the loss of relevant records.
Objectives: To develop and test modified searches for HP&PH trials in order to identify searches which most appropriately balance the need for sensitivity and specificity.
Methods: A prospective comparison in PubMed of our usual search strategy with three experimental approaches:
1. Searches where the MeSH heading 'Public Health' was exploded, but not in identified trees, such as 'Epidemiologic Research Design' and 'Epidemiologic Study Characteristics'
2. Searches using the Cochrane Trials filter and searches using the PubMed limit, Clinical Trial
3. Searches using the NOT operator to avoid retrieval of pre-specified non public health topics.
Searches were compared in terms of sensitivity and specificity, and impact on resource issues.
Results: Analysis has shown that 93% of records were excluded as not relevant to public health, of which approximately 80% were recalled on the basis of study design terms. Data will be presented on all aspects of the approaches tested to demonstrate their relative effectiveness.
Conclusions: This study will provide evidence for, future methods of updating TRoPHI and what trade-offs are being made between sensitivity and specificity.
The exploded MeSH term, 'Public Health' includes terms relating to study design, e.g. 'double-blind methodÂ’ and 'prospective studiesÂ’, resulting in a high percentage of records recalled on the basis of study design alone, and not relevant to public health
Many records are grouped around areas not relevant to public health, e.g. surgical techniques, drug efficacy trials.
Several modifications to deal with these issues have been considered, but uncertainty exists regarding the loss of relevant records.
Objectives: To develop and test modified searches for HP&PH trials in order to identify searches which most appropriately balance the need for sensitivity and specificity.
Methods: A prospective comparison in PubMed of our usual search strategy with three experimental approaches:
1. Searches where the MeSH heading 'Public Health' was exploded, but not in identified trees, such as 'Epidemiologic Research Design' and 'Epidemiologic Study Characteristics'
2. Searches using the Cochrane Trials filter and searches using the PubMed limit, Clinical Trial
3. Searches using the NOT operator to avoid retrieval of pre-specified non public health topics.
Searches were compared in terms of sensitivity and specificity, and impact on resource issues.
Results: Analysis has shown that 93% of records were excluded as not relevant to public health, of which approximately 80% were recalled on the basis of study design terms. Data will be presented on all aspects of the approaches tested to demonstrate their relative effectiveness.
Conclusions: This study will provide evidence for, future methods of updating TRoPHI and what trade-offs are being made between sensitivity and specificity.