Estimating cholera incidence with cross-sectional serology

Area of Science:

• Epidemiology
• Immunology
• Machine Learning

Background:

• Accurate cholera epidemiology is crucial for effective control.
• Current incidence data often lack laboratory confirmation, relying on syndromic surveillance.
• Serological markers could enable incidence estimation via cross-sectional serosurveys.

Purpose of the Study:

• To develop and validate machine learning models for identifying recent Vibrio cholerae O1 infections using serological markers.
• To assess the feasibility of using cross-sectional serosurveys for estimating cholera incidence.

Main Methods:

• Trained machine learning models on 1569 serologic samples from Bangladesh to detect recent V. cholerae O1 infections.
• Utilized antibody profiles and varying numbers of serological markers (six and two).
• Validated model performance on North American volunteers and simulated serosurveys.

Main Results:

• Models accurately identified individuals infected within the past year (6-marker AUC: 93.4%, 2-marker AUC: 91.0%).
• The two-marker model showed strong performance in validation cohorts (AUC range: 88.4-98.4%).
• Simulated serosurveys demonstrated accurate annual incidence estimation even with small sample sizes and low incidence rates.

Conclusions:

• Antibody profiles contain information about the timing of recent V. cholerae O1 infections.
• Machine learning models can reliably identify recent infections using serological markers.
• Cross-sectional serosurveys represent a viable method for estimating cholera incidence.