Predicting Leptospirosis Using Baseline Laboratory Tests and Geospatial Mapping of Acute Febrile Illness Cases...

Area of Science:

• Infectious Diseases
• Epidemiology
• Medical Informatics

Background:

• Leptospirosis, a zoonotic infection caused by *Leptospira* bacteria, is reemerging globally.
• The disease presents diagnostic challenges due to nonspecific symptoms and can lead to severe outcomes with high mortality.
• Poor sanitation and urbanization are linked to increased risk in affected regions.

Purpose of the Study:

• To investigate associations between laboratory parameters and leptospirosis diagnosis.
• To identify spatial patterns and high-risk areas using geographic mapping.
• To evaluate the utility of machine learning models for leptospirosis prediction.

Main Methods:

• An observational retrospective study analyzed 325 patients with suspected leptospirosis over one year.
• Laboratory investigations, geographic mapping, and machine learning (k-nearest neighbors - KNN) were employed.
• IgM ELISA was used for laboratory confirmation of leptospirosis.

Main Results:

• Of 325 patients, 43 (13.2%) tested positive for leptospirosis.
• Geographic mapping revealed case clusters in West Bengal, India, with some cases from Tripura and Bangladesh.
• No significant association was found between individual laboratory parameters and diagnosis; KNN showed 74% accuracy (AUC 0.6).

Conclusions:

• Geographic mapping identified leptospirosis case clusters but no strong links with individual lab parameters.
• Machine learning models, particularly KNN, offer moderate predictive accuracy for leptospirosis.
• Overlapping clinical features with dengue and scrub typhus complicate diagnosis in endemic areas like West Bengal.