Forecasting Human African Trypanosomiasis Prevalences from Population Screening Data Using Continuous Time Models

To eliminate and eradicate gambiense human African trypanosomiasis (HAT), maximizing the effectiveness of active case finding is of key importance. The progression of the epidemic is largely influenced by the planning of these operations. This paper introduces and analyzes five models for predicting HAT prevalence in a given village based on past observed prevalence levels and past screening activities in that village. Based on the quality of prevalence level predictions in 143 villages in Kwamouth (DRC), and based on the theoretical foundation underlying the models, we consider variants of the Logistic Model—a model inspired by the SIS epidemic model—to be most suitable for predicting HAT prevalence levels. Furthermore, we demonstrate the applicability of this model to predict the effects of planning policies for screening operations. Our analysis yields an analytical expression for the screening frequency required to reach eradication (zero prevalence) and a simple approach for determining the frequency required to reach elimination within a given time frame (one case per 10000). Furthermore, the model predictions suggest that annual screening is only expected to lead to eradication if at least half of the cases are detected during the screening rounds. This paper extends knowledge on control strategies for HAT and serves as a basis for further modeling and optimization studies.Author Summary: The primary strategy to fight gambiense human African trypanosomiasis (HAT) is to perform extensive population screening operations among endemic villages. Since the progression of the epidemic is largely influenced by the planning of these operations, it is crucial to develop adequate models on this relation and to employ these for the development of effective planning policies. We introduce and test five models that describe the expected development of the HAT prevalence in a given village based on historical information. Next, we demonstrate the applicability of one of these models to evaluate planning policies, presenting mathematical expressions for the relationship between participation in screening rounds, sensitivity of the diagnostic test, endemicity level in the village considered, and the screening frequency required to reach eradication (zero prevalence) or elimination (one case per 10000) within a given time-frame. Applying these expressions to the Kwamouth health zone (DRC) yields estimates of the maximum screening interval that leads to eradication, the expected time to elimination, and the case detection fraction needed to reach elimination within five years. This paper serves as a basis for further modeling and optimization studies.