Analysis of risk factors for dismal prognosis of cryptococcal meningitis and construction of a Nomogram predictive model

Abstract

Objective To analyze the risk factors for dismal prognosis in patients with cryptococcal meningitis and construct a prediction scoring system Nomogram model. Methods A total of 100 patients with cryptococcal meningitis who treated with anticryptococcal therapy admitted to Affiliated Hospital of Zunyi Medical University from January 2010 to August 2022 were selected. The patients were divided into favorable prognosis group (n = 19) and dismal prognosis group (n = 81) according to the results of cerebrospinal fluid (CSF) cryptococcal culture during hospitalization and clinical symptoms and signs at the time of discharge. Risk factors were screened by using univariate and multivariate stepwise Logistic regression analyses. A Nomogram model was constructed based on the risk factors, the receiver operating characteristic (ROC) curve and calibration curves of the model were plotted, and Hosmer ‐ Lemeshow goodness‐of‐fit test was performed. Results The proportion of patients in the dismal prognosis group with Nutritional Risk Screening 2002 (NRS 2002) score (Z = ‐ 3.898, P = 0.000), CSF pressure > 250 mm H2O (χ2 = 9.512，P = 0.002) and duration of antifungal treatment < 14 d (χ2 = 17.847，P = 0.000) on admission were higher than those in the favorable prognosis group, and the blood routine red blood cell count (t = ‐ 2.802, P = 0.006) and lymphocyte count (Z = ‐ 2.878, P = 0.004), plasma albumin (t = ‐ 4.332, P = 0.000), and the proportion of amphotericin B application (χ2 = 4.597, P = 0.032) were lower than those in the favorable prognosis group. Logistic regression analysis showed the admission high NRS 2002 score (OR = 3.258, 95%CI: 1.337-7.940; P = 0.009), CBF pressure > 250 mm H2O (OR = 0.108, 95%CI: 0.018-0.659; P = 0.016), and the duration of antifungal treatment < 14 d (OR = 0.092, 95%CI: 0.011-0.742; P = 0.025) were risk factors for dismal prognosis of cryptococcal meningitis. A Nomogram model was constructed based on the above 3 risk factors, and the area under the ROC curve was 0.927 (95%CI: 0.873-0.980, P = 0.000), which predicted a cut ‐ off value of 53.50 points for dismal prognosis in cryptococcal meningitis; the calibration curve (with good consistency), and the Hosmer‐Lemeshow goodness‐of‐fit test (χ2 = 2.694, P = 0.912) indicated that the model had good discrimination, calibration and stability. Conclusions Patients with cryptococcal meningitis with a high NRS 2002 score, CSF pressure > 250 mm H2O, and antifungal treatment < 14 d had a dismal prognosis, and the Nomogram model constructed accordingly has a high predictive value of dismal prognostic risk.

DOI: 10.3969/j.issn.1672‐6731.2024.04.006