Does the accuracy of clinician diagnosis of COVID-19 compared to RT-PCR in adults depend on the number and severity of comorbidities?

Background: Due to the high transmissibility of SARS-Cov-2, the virus causing COVID-19, accurate diagnostic methods are essential for effective infection control, but the gold standard method of real-time polymerase chain reaction (RT-PCR) is costly, slow, and test capacity has at times been insufficient. Method: Diagnosis data were retrieved from registers, based on positive RT-PCR or ICD-10 codes set by clinicians. Through linkage to a population-representative adult cohort in Sweden, we assessed the accuracy of clinician diagnosis against RT-PCR, stratified by number and severity of comorbidities. Results: A total of 42,621 subjects were included. Of these, 6,560 had COVID-19. Clinician diagnosis was found in 5,705 subjects, while 3,936 had a positive RT-PCR and 3,081 got diagnosed with both methods. Of those with at least one comorbidity, sensitivity for clinician diagnosis ranged from 69% (95% CI 44-86) for those with two “severe” comorbidities (diabetes and chronic obstructive pulmonary disease) to 84% (95% CI 73-91) for those with two “moderately severe” comorbidities (asthma and hypertension). Specificity was > 90% for all comorbidity groups. Youden9s index increased slightly with the number of comorbidities in both “severe” and “moderately severe” categories, but for those with “light” comorbidities (eczema, rhinitis, sleep disorders), it was the lowest with ≥2 comorbidities (69% (95% CI 66-72)). Youden9s index was 71% (95% CI 70-72) for those with no comorbidities and 71% (95% 69-73) for the whole cohort. Conclusion: Clinicians identify non-cases to a high degree, but RT-PCR is needed for adequate sensitivity, regardless of comorbidity.