Fine-grained mathematical modeling for cost-effectiveness evaluation of public health policies for cervical cancer, with application to a Colombian case study

BACKGROUND: Cervical cancer (CC) is globally ranked fourth in terms of incidence and mortality among women. Vaccination against Human Papillomavirus (HPV) and screening programs can significantly reduce CC mortality rates. Hence, executing cost-effective public health policies for prevention and surveillance is crucial. However, defining policies that make the best use of the available resources is not easy, as it requires predicting the long-term costs and results of interventions on a changing population. Since the simpler task of predicting the results of public health policies is difficult, devising those that make the best usage of available resources is an arduous challenge for decision-makers. METHODS: This paper proposes a fine-grained epidemiological simulation model based on differential equations, to effectively predict the costs and effectiveness of CC public health policies that include vaccination and screening. The model represents population dynamics, HPV transmission within the population, likelihood of infection clearance, virus-induced appearance of precancerous lesions and eventually CC, as well as immunity gained with vaccination and early detection with screening. RESULTS: We offer a compartmentalized modeling approach that separates population, epidemics, and intervention concerns. We instantiate models with actual data from a Colombian case study and analyze their results to show how our modeling approach can support CEA studies. Moreover, we implement models in an open-source software tool to simultaneously define and evaluate multiple policies. With the support of the tool, we analyze 54 policies within a 30-year time horizon and use as a comparator the CC policy that has been used until recently. We identify 8 dominant policies, the best one with an ICER of 6.3 million COP (Colombian Pesos) per averted DALY. We also validate the modeling approach against the available population and HPV epidemic data. The effects of uncertainty in the values of key parameters (discount rate, sensitivity of screening tests) is evaluated through one-way sensitivity analysis. CONCLUSIONS: Our modeling approach can provide valuable support for healthcare decision-makers. The implementation into an automated tool allows customizing the analysis with country-specific data, flexibly defining public health policies to be evaluated, and conducting disaggregate analyses of their cost and effectiveness.