Quantifying Positive Health Externalities of Disease Control Interventions: Modeling Chikungunya and Dengue

Purpose. Health interventions can generate positive externalities not captured in traditional, single-disease cost-effectiveness analyses (CEAs), potentially biasing results. We illustrate this with the example of mosquito-borne diseases. When a particular mosquito species can transmit multiple diseases, a single-disease CEA comparing disease-specific interventions (e.g., vaccination) with interventions targeting the mosquito population (e.g., insecticide) would underestimate the insecticide’s full benefits (i.e., preventing other diseases). Methods. We developed three dynamic transmission models: chikungunya, dengue, and combined chikungunya and dengue, each calibrated to disease-specific incidence and deaths in Colombia (June 2014 to December 2017). We compared the models’ predictions of the incremental benefits and cost-effectiveness of an insecticide (10% efficacy), hypothetical chikungunya and dengue vaccines (40% coverage, 95% efficacy), and combinations of these interventions. Results. Model calibration yielded realistic parameters that produced close matches to disease-specific incidence and deaths. The chikungunya model predicted that vaccine would decrease the incidence of chikungunya and avert more total deaths than insecticide. The dengue model predicted that insecticide and the dengue vaccine would reduce dengue incidence and deaths, with no effect for the chikungunya vaccine. In the combined model, insecticide was more effective than either vaccine in reducing the incidence of and deaths from both diseases. In all models, the combined strategy was at least as effective as the most effective single strategy. In an illustrative CEA, the most frequently preferred strategy was vaccine in the chikungunya model, the status quo in the dengue model, and insecticide in the combined model. Limitations. There is uncertainty in the target calibration data. Conclusions. Failure to capture positive externalities can bias CEA results, especially when evaluating interventions that affect multiple diseases. Multidisease modeling is a reasonable alternative for addressing such biases.