Capturing Adolescent Studies: Developing a Targeted Search Hedge to Enhance Systematic Reviews

Abstract

PURPOSE The objective of this project is to develop and validate a librarian-created search hedge (a standardized, pre-tested search strategy) to improve the identification of adolescent-focused studies. Adolescents represent a minority population in the medical literature and are frequently difficult to identify with Medical Subject Headings (MeSH), as they are often included only as a small subgroup within adult or pediatrics studies. Existing database search strategies often lack the precision and recall needed to comprehensively capture adolescent health literature. This hedge is designed to enhance both sensitivity and specificity in identifying adolescent-focused studies within databases such as PubMed, supporting researchers, clinicians, and trainees conducting systematic reviews as well as bedside searching. METHODS A gold standard dataset was developed by selecting five health topics: Asthma, attention-deficit/hyperactivity disorder, athletic injuries, dermatitis, and hepatitis. These topics were intentionally chosen to include conditions both highly relevant to adolescence and less age-specific, in order to rigorously test the search strategy’s ability to distinguish adolescent-focused studies. From PubMed, 1,000 articles across these topics were retrieved and screened in Covidence by a team of eight librarian reviewers to determine inclusion as adolescent-focused studies. Guided by existing hedges and preliminary testing, the team iteratively developed new strings (combinations of keywords, Medical Subject Headings, and age filters) through librarian peer review, combining MeSH terms, keywords, and age filters relevant to adolescents aged 13–17 years. Each string was tested in PubMed and evaluated against the gold standard set. For each iteration, we calculated true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN), and derived sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall accuracy. RESULTS Using the gold standard dataset, the core hedge retrieved TP=182, TN=619, FP=58, and FN=141, corresponding to 56.3% sensitivity, 91.4% specificity, 75.8% PPV, 81.4% NPV, and 80.1% overall accuracy. Compared to the PubMed Adolescent filter (50.2% sensitivity, 93.2% specificity), the hedge modestly improved sensitivity with only a small tradeoff in specificity. A bedside-oriented version achieved very high specificity (98.4%) and precision (88.7%) but much lower sensitivity (26.6%). Testing of additional variants, including puberty, young adult, and adolescent parents, did not improve performance beyond the core hedge. Further refinements will focus on increasing sensitivity while maintaining acceptable specificity. CONCLUSIONS A librarian-developed search hedge can improve sensitivity compared to the standard PubMed Adolescent filter, though overall sensitivity remains modest. Different versions may be tailored to systematic reviews (favoring sensitivity) or bedside use (favoring specificity), offering flexibility depending on the research or clinical context. This study represents the first validated search hedge specifically designed for adolescent health, providing a reproducible approach to strengthen evidence synthesis in this understudied population.