Article type
Year
Abstract
Background:
Medical knowledge grows quickly and is often accessed via online database such as PubMed by clinical healthcares to solve clinical questions. PubMed is the most popular database due to the characteristics of being easy to use, free and continued improvement to meet user’s requirement especially in evidence-based medicine. Clinical queries are online search filters designed to improve the retrieval of relevant and evidence-based articles for the purpose categories of therapy, diagnosis, prognosis, etiology, clinical prediction guides and systematic review (SR) from PubMed database. The research methodology filters use MeSH and tags to improve the sensitivity and specificity of search results. However, the usage of MeSH and tags may loss the newest evidences those just supplied by publishers, indexed was in process or not Medline format.
Objectives:
The aim of this study is to develop innovative search filter for SR and compare sensitivity, specificity and precision to the clinical queries for SR in PubMed database.
Methods:
We created a customized search strategy to accommodate the newestly not-indexed or in-process systematic reviews. Covid-19 was trending topic recently and was used as the references to compare the sensitivity, specificity and precision of retrieval of SRs by clinical queries in PubMed database or our customized search strategy. Sensitivity and specificity were defined as the proportion of SRs detected and non-SRs excluded by the given search filters, respectively. Precision was defined as the proportion of retrieved articles that are SRs.
Results:
Textword of Covid-19 was used for search in PubMed on 5 April, 2020 and 2575 articles founded. Fourteen (0.5%) and 25 (1.0%) articles were recognized as systematic reviews using PubMed clinical queries of SR and our customized filter, respectively. The PubMed search strategy for SR had a sensitivity of 52.0% (95% confidence interval 33.5% to 70.0% ) and a specificity of 99.96% (95% confidence interval 99.80% to 1). Our customized search strategy for SR had a sensitivity of 96.0% (95% confidence interval 80.5% to 99.3% ) and a specificity of 99.96% (95% confidence interval 99.80% to 1). The precision of these 2 search strategies were 92.9% and 96.0%, respectively. Most of the missed articles (91.7%) by using PubMed clinical queries for SRs were retrieved by our customized search strategy of not-indexed filter.
Conclusions:
Systematic reviews can be retrieved from PubMed by this innovative and optimal search strategy with perfect sensitivity or specificity, and high precision even for the newest topics or articles. Sometimes, the newest and not-indexed articles in PubMed may be missed if clinical queries or "limit" function was used.
Patient or healthcare consumer involvement: yes.
The precision searching provided in this study can compromised the bugs for retrieving systematic reviews in PubMed. Clinical health-carers can easily search the best evidence of systematic reviews for solving patients-centered questions.
Medical knowledge grows quickly and is often accessed via online database such as PubMed by clinical healthcares to solve clinical questions. PubMed is the most popular database due to the characteristics of being easy to use, free and continued improvement to meet user’s requirement especially in evidence-based medicine. Clinical queries are online search filters designed to improve the retrieval of relevant and evidence-based articles for the purpose categories of therapy, diagnosis, prognosis, etiology, clinical prediction guides and systematic review (SR) from PubMed database. The research methodology filters use MeSH and tags to improve the sensitivity and specificity of search results. However, the usage of MeSH and tags may loss the newest evidences those just supplied by publishers, indexed was in process or not Medline format.
Objectives:
The aim of this study is to develop innovative search filter for SR and compare sensitivity, specificity and precision to the clinical queries for SR in PubMed database.
Methods:
We created a customized search strategy to accommodate the newestly not-indexed or in-process systematic reviews. Covid-19 was trending topic recently and was used as the references to compare the sensitivity, specificity and precision of retrieval of SRs by clinical queries in PubMed database or our customized search strategy. Sensitivity and specificity were defined as the proportion of SRs detected and non-SRs excluded by the given search filters, respectively. Precision was defined as the proportion of retrieved articles that are SRs.
Results:
Textword of Covid-19 was used for search in PubMed on 5 April, 2020 and 2575 articles founded. Fourteen (0.5%) and 25 (1.0%) articles were recognized as systematic reviews using PubMed clinical queries of SR and our customized filter, respectively. The PubMed search strategy for SR had a sensitivity of 52.0% (95% confidence interval 33.5% to 70.0% ) and a specificity of 99.96% (95% confidence interval 99.80% to 1). Our customized search strategy for SR had a sensitivity of 96.0% (95% confidence interval 80.5% to 99.3% ) and a specificity of 99.96% (95% confidence interval 99.80% to 1). The precision of these 2 search strategies were 92.9% and 96.0%, respectively. Most of the missed articles (91.7%) by using PubMed clinical queries for SRs were retrieved by our customized search strategy of not-indexed filter.
Conclusions:
Systematic reviews can be retrieved from PubMed by this innovative and optimal search strategy with perfect sensitivity or specificity, and high precision even for the newest topics or articles. Sometimes, the newest and not-indexed articles in PubMed may be missed if clinical queries or "limit" function was used.
Patient or healthcare consumer involvement: yes.
The precision searching provided in this study can compromised the bugs for retrieving systematic reviews in PubMed. Clinical health-carers can easily search the best evidence of systematic reviews for solving patients-centered questions.