Leptospirosis is an important cause of febrile illness in tropical settings, but early clinical recognition is difficult because its signs and symptoms overlap with other infections and serological confirmation is frequently delayed. The authors performed a secondary analysis of archived clinical trial data to derive a simple, bedside decision model based on routinely available clinical variables, with the goal of improving early diagnostic probability for leptospirosis in adults presenting with undifferentiated fever.
The analysis used data from a randomized controlled trial that enrolled adults with undifferentiated febrile illness lasting 5–15 days between October 2020 and February 2021. From the trial cohort, 190 patients were available for this secondary analysis. The reference standard for case ascertainment was serological positivity by IgM-ELISA. The investigators applied multivariable logistic regression to identify independent clinical predictors of IgM-ELISA–confirmed leptospirosis and then derived a simplified additive score assigning one point per predictor.
Independent predictors retained in multivariable analysis were conjunctival suffusion, icterus, and acute kidney injury. These three clinical features formed the basis of a pragmatic bedside score with possible values from 0 to 3, where each present predictor contributed one point. The source reports that model derivation focused on easily observable clinical signs and basic laboratory evidence of renal dysfunction, aiming for a tool usable at the point of care.
Discrimination of the 0–3 score was assessed using receiver operating characteristic analysis. The score demonstrated good overall discrimination with area under the ROC curve (AUC) of 0.801. Using a threshold of ≥2 points, the score yielded sensitivity of 68.8% and specificity of 82.4% for detecting IgM-ELISA–positive leptospirosis among the study population. These operating characteristics indicate reasonable rule-in utility at the chosen cut-off, while retaining moderate sensitivity.
In an exploratory analysis, the authors evaluated a model that included proteinuria as a predictor. That proteinuria-based model showed moderate discrimination with an AUC of 0.741. The source notes this as an alternative or supplementary approach but presents it as exploratory rather than the primary bedside score.
Baseline seroprevalence of leptospirosis in the analysed cohort was 25.3% (48 of 190 patients). The authors modelled post-test probabilities for clinical combinations and for simulated integration with rapid diagnostic testing. When all three clinical predictors (conjunctival suffusion, icterus, acute kidney injury) were present, the probability of IgM-ELISA positivity rose to 91% from the 25% baseline. When the bedside score was combined with a positive rapid diagnostic test, modelled probabilities increased further to 97%–99%, illustrating how clinical assessment plus point-of-care testing might substantially strengthen diagnostic certainty.
This secondary analysis produced a concise bedside score using three readily ascertained clinical features that meaningfully increases the likelihood of diagnosing leptospirosis among adults with 5–15 days of undifferentiated fever. The score has good discrimination (AUC 0.801) and appears particularly useful for raising post-test probability when used together with rapid diagnostic tests. The simplicity of assigning one point per predictor supports potential uptake in frontline and resource-limited settings where rapid decision-making is required.
The source is a secondary analysis of trial data; details about external validation, prospective application in other populations, or calibration metrics were not reported in the abstract. The full text may include additional methodological details, but those specifics were not provided in the source content presented here. Clinicians should interpret the reported sensitivity and specificity within the context of the study population (seroprevalence 25.3%) and consider local epidemiology and available diagnostics before implementation.
Clinicians and implementers seeking to adopt the score should consult the full article for complete methodology and consider local validation before routine use.