Accuracy of self-diagnosis in conditions commonly managed in primary care: diagnostic accuracy systematic review and meta-analysis

ObjectivesTo assess the diagnostic accuracy of self-diagnosis compared with a clinical diagnosis for common conditions in primary care.DesignSystematic review. Meta-analysis.Data sourcesMedline, Embase, Cochrane CENTRAL, Cochrane Database of Systematic Reviews and CINAHL from inception to 25 January 2021.Study selectionEligible studies were prospective or retrospective studies comparing the results of self-diagnosis of common conditions in primary care to a relevant clinical diagnosis or laboratory reference standard test performed by a healthcare service provider. Studies that considered self-testing only were excluded.Data extractionTwo authors independently extracted data using a predefined data extraction form and assessed risk of bias using Quality Assessment of Diagnostic Accuracy Studies-2.Methods and results5047 records identified 18 studies for inclusion covering the self-diagnosis of three common conditions: vaginal infection (five studies), common skin conditions (four studies) and HIV (nine studies). No studies were found for any other condition. For self-diagnosis of vaginal infection and common skin conditions, meta-analysis was not appropriate and data were reported narratively. Nine studies, using point-of-care oral fluid tests, reported on the accuracy of self-diagnosis of HIV and data were pooled using bivariate meta-analysis methods. For these nine studies, the pooled sensitivity was 92.8% (95% CI, 86% to 96.5%) and specificity was 99.8% (95% CI, 99.1% to 99.9%). Post hoc, the robustness of the pooled findings was tested in a sensitivity analysis only including four studies using laboratory testing as the reference standard. The pooled sensitivity reduced to 87.7% (95% CI, 81.4% to 92.2%) and the specificity remained the same. The quality of all 18 included studies was assessed as mixed and overall study methodology was not always well described.Conclusions and implications of key findingsOverall, there was a paucity of evidence. The current evidence does not support routine self-diagnosis for vaginal infections, common skin conditions and HIV in primary care.PROSPERO registration numberCRD42018110288.