Simple clinical decision model for predicting leptospirosis: a secondary analysis of a randomized controlled trial

Background
Leptospirosis is a major cause of febrile illness in tropical regions, but early diagnosis is challenging due to overlapping clinical features and delayed serological confirmation. We developed a simple bedside decision model using readily available clinical variables.

Methods
This secondary analysis used archived data from a randomized controlled trial of adults with undifferentiated febrile illness (October 2020–February 2021) of 5–15 days’ duration. Independent predictors of IgM-ELISA positive leptospirosis were identified using multivariable logistic regression. A simplified score assigning one point per predictor was derived. Discrimination was assessed using ROC analysis, and post-test probabilities were modelled for clinical combinations and simulated rapid diagnostic testing.

Results
Among 190 patients, 48 (25.3%) were serologically positive for leptospirosis. Conjunctival suffusion, icterus and acute kidney injury were independent predictors. The score (0–3) showed good discrimination (AUC 0.801), with sensitivity 68.8% and specificity 82.4% at ≥2. An exploratory proteinuria-based model demonstrated moderate discrimination (AUC 0.741). Probability increased from 25% at baseline to 91% when all three predictors were present and to 97%–99% when combined with a positive rapid test.

Conclusions
A simple bedside score substantially increases the probability of diagnosing leptospirosis in patients with undifferentiated febrile illness of 5–15 days' duration, particularly when integrated with rapid testing.