A Machine Learning Model for Evaluating Imported Disease Screening Strategies in Immigrant Populations

Area of Science:

• Tropical Medicine
• Public Health
• Machine Learning Applications

Background:

• Imported diseases pose a significant health challenge in immigrant populations.
• Early diagnosis and treatment are crucial for managing these conditions effectively.
• Existing screening programs may not be optimally targeted.

Purpose of the Study:

• To develop a machine learning model for predicting imported diseases in immigrants.
• To identify key prognostic variables for diseases such as HIV, malaria, and hepatitis.
• To aid in the design and development of more precise screening programs.

Main Methods:

• Retrospective cross-sectional study of immigrant patients (n=759) attending a Tropical Medicine Unit (2009-2016).
• Development of a mathematical model using machine learning to predict disease onset.
• Analysis of prognostic variables for HIV, malaria, hepatitis B/C, schistosomiasis, Chagas, syphilis, and strongyloidiasis.

Main Results:

• The model achieved high predictive accuracy for various diseases: HIV (84.9%), Chagas (92.9%), chronic hepatitis B (85.4%), schistosomiasis (86.9%), hepatitis C (85.6%), malaria (93.3%), syphilis (79.4%), and strongyloidiasis (88.4%).
• Predicted screenings needed to detect the first case varied by disease, e.g., 26 for HIV and Chagas, 12 for hepatitis B and schistosomiasis.
• Machine learning identified key variables for predicting disease burden in immigrants.

Conclusions:

• Machine learning models can effectively predict the prevalence of imported diseases in immigrant populations.
• These models enhance the precision of screening programs, identifying individuals most likely to benefit.
• The study contributes to the optimized design of screening strategies for tropical diseases in immigrants.