Identifying key aspects to enhance predictive modeling for early identification of schistosomiasis hotspots to guide mass drug administration

Abstract

Background: Schistosomiasis, a neglected tropical parasitic disease, threatens the lives of over 250 million people worldwide. In schistosomiasis prevention, high-transmission areas that do not respond to treatments, known as hotspots, pose extreme challenges to the elimination of the disease. Accurate and early identification of such hotspots is crucial for timely intervention, but this is hindered by the limited availability of effective prediction methods.

Methods: Based on the Schistosomiasis Consortium for Operational Research and Evaluation (SCORE) project over a 5-year period, this study developed prediction methods for the first (baseline) year to identify hotspots. Three key aspects were considered: (i) collecting secondary data from public sources to complement baseline schistosomiasis infection data and constructing spatially weighted predictors to incorporate neighboring information; (ii) categorizing predictors to mitigate overfitting and quantifying the importance of each category in hotspot predictions; and (iii) investigating the hotspot imbalance distribution and addressing the imbalance with a sampling-based technique to improve prediction performance.

Results: Compared to the approach using only baseline infection data, the spatially weighted data fusion method achieved relative improvements (RIs) in hotspot prediction accuracy by fusing baseline infection data with each predictor category: 10% with biology, 8.6% with geography, 6.6% with society, 3.5% with baseline infection data around villages, 3.3% with environment, 1.8% with agriculture, and 7.2% with all predictors. Furthermore, across the same predictor combinations, applying the sampling-based technique with the proposed method yielded RIs of 6.5%-37.9%, compared to the approach that did not address the imbalance.

Conclusion: Spatially weighted data fusion using secondary data improved the early identification of schistosomiasis hotspots. Addressing the imbalance of hotspots can further improve the early identification of the hotspots.

Fig 1. Study framework for enhancing early identification of schistosomiasis hotspots (early identification refers to the use of infection data only from the first year to develop prediction models).

Fig 2. Map of hotspot areas around Lake Victoria, highlighted in red points, in Tanzania (top) and Kenya (bottom) using persistent hotspot definitions I–II.

The map layers were created using publicly available world map data from Natural Earth, accessed via the R package rnaturalearth [24].

Fig 3. Under PHS definition I, the RIs, in prediction accuracy on unprocessed test sets, obtained from the models trained using pre-processed synthetic data based on the proposed different predictor configurations, compared to models using unprocessed original imbalanced training data, where the best model with the highest prediction accuracy was considered for each method (i.e., y-axis).