Diagnostic Test Accuracy of Deep Learning Prediction Models on COVID-19 Severity: Systematic Review and Meta-Analysis (Preprint)

• DOI: 10.2196/preprints.46340
• Publication Date: 2023-02-10T22:58:17Z
• Authors: Changyu Wang; Siru Liu; Yu Tang; Hao Yang; Jialin Liu
• Keywords: Funnel plot; Triage

<sec> <title>BACKGROUND</title> Deep learning (DL) prediction models hold great promise in the triage of COVID-19. </sec> <title>OBJECTIVE</title> We aimed to evaluate diagnostic test accuracy DL for assessing and predicting severity <title>METHODS</title> searched PubMed, Scopus, LitCovid, Embase, Ovid, Cochrane Library studies published from December 1, 2019, April 30, 2022. Studies that used assess or predict COVID-19 were included, while those without analysis dichotomies excluded. QUADAS-2 (Quality Assessment Diagnostic Accuracy 2), PROBAST (Prediction Model Risk Bias Tool), funnel plots estimate bias applicability. <title>RESULTS</title> A total 12 retrospective involving 2006 patients reported cross-sectionally assessed value on severity. The pooled sensitivity area under curve 0.92 (95% CI 0.89-0.94; &lt;i&gt;I&lt;/i&gt;&lt;sup&gt;2&lt;/sup&gt;=0.00%) 0.95 0.92-0.96), respectively. 13 3951 longitudinal predictive disease 0.76 0.74-0.79; 0.80 0.76-0.83), <title>CONCLUSIONS</title> can help clinicians identify potentially severe cases early triage. However, high-quality research is lacking. <title>CLINICALTRIAL</title> PROSPERO CRD42022329252; https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD 42022329252