Validation of Search Filters for Identifying Pediatric Studies in PubMed - 26/02/13
, Mariska M.G. Leeflang, PhD 2, Elvira C. van Dalen, MD, PhD 1, Leontien C.M. Kremer, MD, PhD 1
Reprint requests: Edith Leclercq, PhD, Cochrane Childhood Cancer Group, Department of Paediatric Oncology, Emma Children’s Hospital/Academic Medical Center, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands.Abstract |
Objective |
To identify and validate PubMed search filters for retrieving studies including children and to develop a new pediatric search filter for PubMed.
Study design |
We developed 2 different datasets of studies to evaluate the performance of the identified pediatric search filters, expressed in terms of sensitivity, precision, specificity, accuracy, and number needed to read (NNR). An optimal search filter will have a high sensitivity and high precision with a low NNR.
Results |
In addition to the PubMed Limits: All Child: 0-18 years filter (in May 2012 renamed to PubMed Filter Child: 0-18 years), 6 search filters for identifying studies including children were identified: 3 developed by Kastner et al, 1 developed by BestBets, one by the Child Health Field, and 1 by the Cochrane Childhood Cancer Group. Three search filters (Cochrane Childhood Cancer Group, Child Health Field, and BestBets) had the highest sensitivity (99.3%, 99.5%, and 99.3%, respectively) but a lower precision (64.5%, 68.4%, and 66.6% respectively) compared with the other search filters. Two Kastner search filters had a high precision (93.0% and 93.7%, respectively) but a low sensitivity (58.5% and 44.8%, respectively). They failed to identify many pediatric studies in our datasets. The search terms responsible for false-positive results in the reference dataset were determined. With these data, we developed a new search filter for identifying studies with children in PubMed with an optimal sensitivity (99.5%) and precision (69.0%).
Conclusion |
Search filters to identify studies including children either have a low sensitivity or a low precision with a high NNR. A new pediatric search filter with a high sensitivity and a low NNR has been developed.
Le texte complet de cet article est disponible en PDF.Keyword : CCG, CCTs, CHF, HSSS, MeSH, NNR, RCTs
Plan
| E.L. and E.vD. are funded in part by KiKa, Children Cancer Free (Stichting Kinderen Kankervrij). The authors declare no conflicts of interest. |
Vol 162 - N° 3
P. 629 - mars 2013 Retour au numéro
Article précédent
- Determinants of Health Care Use in a Population-Based Leukodystrophy Cohort
- Clint Nelson, Michael B. Mundorff, E. Kent Korgenski, Cameron J. Brimley, Rajendu Srivastava, Joshua L. Bonkowsky
- Clinical Presentation of Cervical Ribs in the Pediatric Population
- Kenny H. Chan, Sarah A. Gitomer, Jonathan N. Perkins, Conan Liang, John D. Strain
Bienvenue sur EM-consulte, la référence des professionnels de santé.
L’accès au texte intégral de cet article nécessite un abonnement.
Déjà abonné à cette revue ?