Evidence-based and epidemiological interpretation of the diagnosis of SARS-COV-2

Abstract

INTRODUCTION AND OBJECTIVES: Diagnostic tests are never perfect; hence cannot definitely tell whether the disease is present or absent which leads to inaccurate decisions by healthcare workers. The key metrics in interpreting the accuracy are sensitivity and specificity and the test performance measures of positive-predictive-value(PPV) and negative-predictive-value (NPV). Positive and negative predictive values of diagnostic tests vary depending upon the pre-test probability. This paper aims to compile the evidence generated and facilitate the epidemiological interpretation of diagnostic laboratory testing available for COVID-19 pandemic. METHODS: We searched systematic reviews published, especially Cochrane reviews and evidence summaries. Diagnostic accuracy studies were searched when there were no published systematic reviews. PPV, NPV and its 95% confidence intervals (CI) were calculated for the reported sensitivities and specificities and assumed pre-test probability levels. RESULTS: At a low pre-test probability level, PPV is 50%(95%CI: 28%-72%) when we assume 56% sensitivity of the antigen test. If the sensitivity is low, PPV will be low and vice versa; however, the NPV is 69.3% (95%CI: 67%-71%). Considering RT-PCR test PPV is 28.6%(95%CI: 13%-52% at a low pre-test probability level) when we assume 40% sensitivity. Assuming RT-PCR sensitivity is 95%, PPV is 47.4% (95%CI: 32%-63%), NPV is 62.3%(95%CI: 60.6% to 63.9%). Further, assuming that the sensitivity of the RT-PCR is 95%, NPV is 95.2%(95%CI: 93.1% to 96.7%). CONCLUSION: In low resource settings symptoms, signs, basic laboratory and other investigations can be used to screen COVID-19 patients and interpretation of the results of the antigen and RT-PCR test for better diagnosis.