Does the Covichem Score Accurately Predict COVID-19 Disease Severity?

The Covichem score may not accurately predict COVID-19 disease severity among patients admitted to the emergency department (ED), according to results of an external validation analysis reported in the American Journal of Emergency Medicine.

The Covichem score (an estimate ranging from 0 to 1) was developed to predict a patient’s risk for severe COVID-19 infection upon hospital admission, using clinical parameters and commonly available laboratory results. Researchers sought to externally validate the diagnostic utility of the Covichem score and to assess its value in predicting intensive care unit (ICU) admission and inhospital mortality. The prospective, single-center, observational, diagnostic accuracy study was conducted at a university hospital in Istanbul, Turkey. The participants presented at the hospital ED from October 1, 2021, to January 1, 2022.

All participants were at least 18 years old and either had a positive SARS-CoV-2 polymerase chain reaction test result or were considered to have a high probability of COVID-19 infection based on computed tomography images and the presence of symptoms such as dry cough, fever, chills, fatigue, dyspnea, chest pain, myalgia, diarrhea, anosmia, and/or ageusia. The study cohort included 507 patients, 313 (61.7%) of whom were classified as having severe COVID-19. The median age for all patients was 63 years (interquartile range [IQR], 28), and the median age of those in the severe group was significantly higher than for those in the nonsevere group (67 years [IQR, 25] vs 55.5 years [IQR, 28], respectively; P <.001). A higher percentage of male patients were in the severe group vs the nonsevere group (55% vs 39%; P <.001).

After calculating patients’ Covichem scores using a threshold of 0.5, the researchers assessed how well these scores predicted disease severity, ICU admission, and mortality. Among all patients, the median Covichem score was 0.093 (IQR, 0.238, n/N = 491/507, 96.8%). The median Covichem score was significantly increased in the severe group vs the nonsevere group (Covichem score: 0.170; IQR, 0.298; n = 300 vs Covichem score: 0.026; IQR, 0.065; n = 191; P < .001).

Using a threshold of 0.5, researchers estimated that 12.4% of participants were at risk of developing severe disease based on their Covichem scores; in actuality, 61.7% of patients studied did develop severe disease. The investigators determined that the Covichem score’s true positive rate for predicting severe disease was 18%, whereas the false positive rate was 3.7%.

The accuracy of the Covichem score was poor at a threshold of 0.5, with an area under the curve (AUC) of 48.5%. When outpatients were excluded and the cutoff was 0.5, the accuracy of the Covichem score had an AUC of 38.8 (34.1-43,7), sensitivity of 18.1 (13.9-22.9), and specificity of 93.8 (87.7-97.5). The researchers also calculated the accuracy of the Covichem score using an AUC of 82% (standard error, 2%; P <.001). The Youden J Index showed that the Covichem score had the highest sensitivity (79.7%) and specificity (71.2%) at a threshold of 0.059.

The Covichem score predicted ICU admission and mortality with an AUC of 72.3% (95% CI, 68.0%-76.2%) and 75.0% (95% CI, 70.8%-78.8%), respectively.

Study limitations include recruitment of study participants from a single referral hospital where most had relatively severe disease, along with researchers’ limited ability to monitor certain measures of disease severity.

“In this external validation of the Covichem score, we found that it performed worse than in the original derivation and validation study, even with the assistance of a new cutoff,” concluded the study authors.

Reference

Ozpolat C, Altunbas E. Diagnostic utility of the Covichem score in predicting COVID-19 disease. Am J Emerg Med. Published online July 16, 2022. doi:10.1016/j.ajem.2022.07.025