Developing a Sensitive Search Strategy in MEDLINE to Retrieve Studies on Assessment of the Diagnostic Performance of Imaging Techniques

Abstract

PURPOSE: To prospectively develop a search strategy in MEDLINE for identifying studies on the diagnostic performance of any imaging modality, with maximized and minimized retrieval of relevant and irrelevant studies, respectively. MATERIALS AND METHODS: Predefined inclusion criteria were used to conduct a hand search of two sets of radiologic journal articles for studies on assessment of the diagnostic performance of imaging techniques. These two sets of articles formed independent derivation and validation record sets for developing and evaluating the search strategy. The sensitivity and positive predictive values (PPVs) of search terms from the derivation reference-standard set of records were used to select terms and develop two components of the search strategy. The first component was used to identify any study (from the imaging literature) in which diagnostic test performance was assessed. The second component was used to identify studies of any imaging modality. The two components were combined in the final search strategy. The sensitivity, specificity, and PPV of the search strategy in the derivation and validation record sets were calculated. RESULTS: The final search strategy had a sensitivity of 92.8%, a specificity of 58.5%, and a PPV of 25.1% for retrieval of the derivation set of records. Validation with an independent set of records gave a sensitivity of 91.9% (95% confidence interval [CI]: 87.1%, 95.1%), a specificity of 52.2% (95% CI: 49.2%, 55.2%), and a PPV of 25.1% (95% CI: 22.0%, 28.5%). Removal of irrelevant publication types further improved specificity and PPV in the validation set: to 77.6% (95% CI: 75.0%, 80.0%) and 40.9% (95% CI: 36.2%, 45.7%), respectively. The volume of imaging literature retrieved from MEDLINE by using the described search strategy has tripled since 1975. CONCLUSION: A sensitive search strategy to identify studies of the diagnostic performance of any imaging test was developed and validated. The retrieval estimates of this strategy in MEDLINE are adequate to develop a register of studies.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 22 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.