Article type
Year
Abstract
Background: EbM Ärzteinformationszentrum (Ebminfo.at) is a rapid review (RR) service for doctors working in Lower Austrian hospitals. Reviews generally focus on narrow clinical questions and use systematic reviews and randomized controlled trials as evidence base. To streamline the search process, a standardized approach is used: systematic searches of Medline and the Cochrane Library complemented by reference list checking of included records. Additional information sources and study designs are added on a case-by-case basis. Since February 2018, we have been evaluating the value of using the PubMed Similar Articles search function as either a replacement or an addition to the main information sources.
Objectives:
1) to evaluate if the combination of a focused Ovid Medline search (Mf) and a PubMed Similar Articles search (PSA) is as sensitive as the combination of a more comprehensive search using Ovid Medline and the Cochrane Library (Mc+CLib);
2) to identify if PSA identifies unique references.
Methods: for each RR we develop two Ovid Medline strategies, one pritiorizing comprehensiveness, one giving greater weight to precision. We also conduct a comprehensive search of the Cochrane Library. Additionally, we use the one to six references found by preliminary searches as seed articles for a PubMed Similar Articles search. For each seed article, the first 100 similar references are limited by study design and exported. After the conclusion of the RR, we check which search approach(es) identified the studies included in the review.
Results: preliminary results based on 17 RR show that the combination Mc+CLiB retrieved on average 208 references (median 148, interquartile range (IQR) 96 to 222) and had an average sensitivity of 92% (median 100%, IQR 86% to 100%). Mf+PSA retrieved on average 216 records (median 102, IQR 103 to 247) with an average sensitivity of 84% (median 100%, IQR 67% to 100%). In four RR, PSA retrieved one unique reference that was not identified by any other search method. Combining all three sources retrieved on average 88.5 (median 93, IQR 35 to 136) additional records and raised the average sensitivity to 97% (median 100%, IQR 100% to 100%). In three RR, we identified one additional reference by reference list checking or additional searching.
This is an ongoing evaluation, further results will be available at the Colloquium.
Conclusions: based on the preliminary results, Mf+PSA is not an acceptable replacement for Mc+CLib as the sensitivity of the approach varied widely. However, PSA proved a valuable addition to Mc+CLib. The unique references found by PSA lacked information about the intervention or the population in their abstracts or MeSH terms, making them difficult to retrieve by traditional database searching.
Patient or healthcare consumer involvement: Ebminfo.at is a service that answers clinical questions posed by doctors. Their questions arise from daily clinical practice, but patients are only indirectly involved.
Objectives:
1) to evaluate if the combination of a focused Ovid Medline search (Mf) and a PubMed Similar Articles search (PSA) is as sensitive as the combination of a more comprehensive search using Ovid Medline and the Cochrane Library (Mc+CLib);
2) to identify if PSA identifies unique references.
Methods: for each RR we develop two Ovid Medline strategies, one pritiorizing comprehensiveness, one giving greater weight to precision. We also conduct a comprehensive search of the Cochrane Library. Additionally, we use the one to six references found by preliminary searches as seed articles for a PubMed Similar Articles search. For each seed article, the first 100 similar references are limited by study design and exported. After the conclusion of the RR, we check which search approach(es) identified the studies included in the review.
Results: preliminary results based on 17 RR show that the combination Mc+CLiB retrieved on average 208 references (median 148, interquartile range (IQR) 96 to 222) and had an average sensitivity of 92% (median 100%, IQR 86% to 100%). Mf+PSA retrieved on average 216 records (median 102, IQR 103 to 247) with an average sensitivity of 84% (median 100%, IQR 67% to 100%). In four RR, PSA retrieved one unique reference that was not identified by any other search method. Combining all three sources retrieved on average 88.5 (median 93, IQR 35 to 136) additional records and raised the average sensitivity to 97% (median 100%, IQR 100% to 100%). In three RR, we identified one additional reference by reference list checking or additional searching.
This is an ongoing evaluation, further results will be available at the Colloquium.
Conclusions: based on the preliminary results, Mf+PSA is not an acceptable replacement for Mc+CLib as the sensitivity of the approach varied widely. However, PSA proved a valuable addition to Mc+CLib. The unique references found by PSA lacked information about the intervention or the population in their abstracts or MeSH terms, making them difficult to retrieve by traditional database searching.
Patient or healthcare consumer involvement: Ebminfo.at is a service that answers clinical questions posed by doctors. Their questions arise from daily clinical practice, but patients are only indirectly involved.
PDF