Evaluation of the usefulness of a predictive model for the diagnosis of COVID-19 in relation to the TR-PCR test and rapid test, in patients at the Univalle Sud Hospital

Abstract

Objective: Early diagnosis for COVID-19 with rapid antigen test (PRA) and real-time polymerase chain reaction (TR-PCR), applying a predictive model. Material and methods: Retro-prospective, observational, longitudinal study; in 2020, a PRA and TR-PCR indication score for COVID-19 was applied to 98 patients (age, fever, respiratory symptoms, chest X-ray, comorbidities, place of residence and work activity); > 4 indicated taking TR-PCR (χ2=0.531, p>0.05), area under the curve (AUC) of 0.5; 3 regression models were performed, the third was more effective (vital signs and clinical signs) to decide positivity; in 2023, 284 patients were included to validate the model. Results: Mean age 47.7±16.7, frequent symptoms with cough, fever and asthenia; mean systolic pressure 119.7±15.6 mmHg and diastolic 74.8±10.9 mmHg; TR-PCR with a sensitivity of 85 %, a validity rate of 49 %; positive predictive value of 48 %; negative predictive value 55 %, positive likelihood ratio 1.019; Cohen's Kappa index, (κ=0.016, 95% CI: -0.073 to 0.1). Conclusions: Molecular tests were contrasted with vital signs and clinical signs; We consider that our model can be useful, but its validation still needs to continue and be applied to more patients.