Predictive Performance of Scoring Systems for Mortality Risk in Patients with Cryptococcemia: An Observational Study

Abstract

Cryptococcal infection is usually diagnosed in immunocompromised individuals and those with meningeal involvement, accounting for most cryptococcosis. Cryptococcemia indicates a poor prognosis and prolongs the course of treatment. We use the scoring systems to predict the mortality risk of cryptococcal fungemia. This was a single hospital-based retrospective study on patients diagnosed with cryptococcal fungemia confirmed by at least one blood culture collected from the emergency department covering January 2012 and December 2020 from electronic medical records in the Taichung Veterans General Hospital. We enrolled 42 patients, including 28 (66.7%) males and 14 (33.3%) females with a mean age of 63.0 ± 19.7 years. The hospital stay ranged from 1 to 170 days (a mean stay of 44.4 days), and the overall mortality rate was 64.3% (27/42). In univariate analysis, the AUC of ROC for MEWS, RAPS, qSOFA, MEWS plus GCS, REMS, NEWS, and MEDS showed 0.833, 0.842, 0.848, 0.846, 0.846, 0.878, and 0.905. In the multivariate Cox regression analysis, all scoring systems, older age, lactate, MAP, and DBP, indicated significant differences between survivor and non-survivor groups. Our results show that all scoring systems could apply in predicting the outcome of patients with cryptococcal fungemia, and the MEDS displays the best performance. We recommend a further large-scale prospective study for patients with cryptococcal fungemia.