Skip to main content

Table 4 Overview and evaluation of the NRS filters identified

From: Study filters for non-randomized studies of interventions consistently lacked sensitivity upon external validation

Study filters (developers) Study types targeted Codinga) Hits in MEDLINE Sensitivity (interval)b)
Clinical trials (University of Texas) [27] Clinical trials 2 1.445.276 39%
Observational Studies – Medline (SIGN) [20] Observational studies 5–10 2.492.125 49–90%
MEDLINE precision (Fraser 2000) [24] Observational studies 5–8,10 9.509.757 73–88%
MEDLINE specificity (Fraser 2000) [24] Observational studies 5–8,10 8.423.107 53–85%
MEDLINE cohort, case-control, and case series strategy (BMJ) [28] Observational studies 5–8,10 2.517.309 55–92%
MEDLINE cohort, case-control, case series, and case study strategy (BMJ) [28] Observational studies 5–8,10 4.441.461 61–93%
Search terms for finding non-RCTs (Royle 2003) [29] Non-RCT 2–12 8.073.091 46–98%
MEDLINE cohort study strategy (BMJ) [28] Cohort studies 5–7 1.982.782 58–69%
Cohort studies (University of Texas) [27] Cohort studies 5–7 2.204.911 52–72%
Case-control studies_1 (University of Texas) [27] Case-control studies 8 660.864 78%
Case-control studies_2 (University of Texas) [27] Case-control studies 8 1.284.387 80%
Medline cohort and case-control strategy (BMJ) [28] Cohort, case-control 5–8 2.430.887 61–92%
Fixed method A for MEDLINE (Furlan 2006) [21] Cohort, case-control, cross-sectional 5–9 4.184.894 49–83%
Fixed method B for MEDLINE (Furlan 2006) [21] Cohort, case-control, cross-sectional 5–9 6.559.073 69–85%
  1. a) Coding of study types; see Table 2; b) Presentation as an interval if the study filter covers more than one study type
  2. SIGN Scottish Intercollegiate Guidelines Network, BMJ British Medical Journal Clinical Evidence