Predicting outcomes in undifferentiated prolonged fever presenting to the emergency department: A prospective study from North India

Abstract

Undifferentiated prolonged fever presenting to the emergency department (ED) poses significant diagnostic and prognostic challenges, particularly in low- and middle-income countries. We aimed to describe the clinical spectrum and in-hospital outcomes of adults presenting to the ED with undifferentiated prolonged fever and to identify the baseline predictors of mortality. In this prospective observational study, consecutive patients aged ≥13 years who fulfilled Durack and Street’s criteria for fever of unknown origin were enrolled between March 2022 and September 2024 at the ED of a tertiary care center in North India. Patients with immunosuppressive states or nosocomial fever were excluded. All patients underwent a standardized diagnostic evaluation guided by potentially diagnostic clues. Baseline demographic, clinical, and laboratory variables recorded at ED presentation, including complete blood count-derived inflammatory indices, were evaluated as potential predictors of in-hospital mortality using receiver operating characteristic (ROC) analysis and multivariable logistic regression. A total of 100 patients were enrolled (mean age: 39 years; 57% male), with a median fever duration of 45 days (interquartile range: 30–90). Infections were the most common etiology (37%), with tuberculosis predominating, followed by malignancies (21%), predominantly advanced hematological malignancies, and noninfectious inflammatory diseases (19%), which frequently presented with multisystem involvement, most commonly systemic lupus erythematosus and Still’s disease. Despite extensive evaluation, 19 patients (19%) remained undiagnosed. Overall, 25 patients (25%) died during hospitalization, with the highest mortality observed among undiagnosed cases (8/19, 42%). In the univariate analysis, neutrophil percentage, neutrophil-to-lymphocyte ratio (NLR), aspartate transaminase, and lactate dehydrogenase levels were significantly higher in the mortality group, whereas total protein and albumin levels were significantly lower. ROC analysis demonstrated modest discriminatory ability for NLR (area under the curve: 0.651). In multivariable analysis, admission NLR emerged as the only independent baseline predictor of in-hospital mortality (odds ratio: 1.089 per one-unit increase; 95% confidence interval: 1.006–1.178; P = 0.034). Undifferentiated prolonged fever presenting to the ED is associated with substantial in-hospital mortality and diagnostic uncertainty. Admission NLR was independently associated with mortality, but discrimination was modest; it may have adjunctive value for early risk stratification pending external validation.