A Predictive Model for 30-Day Mortality of Fungemia in ICUs

BACKGROUND: Few predictive models have been established to predict the risk of 30-day mortality from fungemia. This study aims to create a nomogram to predict the 30-day mortality of fungemia in ICUs. METHODS: Data of ICU patients with fungemia from both the Medical Information Mart for Intensive Care (MIMIC-III) database and the Grade-III Class-A hospital in China were collected. The data extracted from the MIMIC-III database functioned as the training dataset, which was used to construct a predictive model for 30-day mortality risk in ICU patients with fungemia; the data from the hospital functioned as the validation dataset, which was used to validate the model. A predictive model for 30-day mortality risk in ICU patients with fungemia was then built based on R software. Such indicators as C-index and calibration curve were utilized to evaluate the prediction ability of the model. Data of ICU patients with fungemia from the hospital were used as a validation dataset to validate the model. RESULTS: Predictive models were constructed by age, international normalized ratio (INR), renal failure, liver disease, respiratory rate (RR), glucocorticoid therapy, antifungal therapy, and platelets. The C-index value of the models was 0.838 (95% CI: 0.79096–0.88504). Attested by external validation results, the model has satisfactory predictive ability. CONCLUSION: The 30-day mortality risk predictive model for ICU patients with fungemia constructed in this study has good predictive ability and may hopefully provide a 30-day mortality risk screening tool for ICU patients with fungemia.