Markovian modeling and simulations for the cost-public health return analysis of prevention campaigns

??Smoking and sedentary lifestyle make individuals more susceptible to certain diseases, thus negatively affecting the quality of life and life expectancy of the population, and ultimately resulting in a higher healthcare expenditure. The ultimate aim of this study is to quantify in the short/medium term the effects of a prevention policy that reduces exposure to such risk factors of the Italian population. We consider the cost of each prevention policy, and associate an economic cost to every year of life lost due to the disease (YLL) and every year lived with disability (YLD). We then compare a baseline scenario with each prevention scenario (in which a prevention policy is implemented), and estimate the net benefits achieved by each prevention policy.\\ We model the evolution of individuals by independent Markov chains whose state spaces describe the exposition to risk factors and the health of the individuals. We focus on five tracer diseases (lung cancer, stroke, myocardial infarction, chronic obstructive pulmonary disease and diabetes) which are responsible for a large fraction of YLL and YLD attributable to smoking and sedentary lifestyle. To calibrate the model, we use data from the Global Burden of Disease Study and Istat data and surveys on the Italian population.\\ We present and discuss the results obtained by the model. In particular, the model predicts in the baseline scenario the decrease of the size of the population and the increase of the average age over 30 years of simulations. We validate these outcomes by comparing our results with Istat forecasting and with a simplified model appropriately defined to capture only demographical aspects. Finally, we conduct an analytical sensitivity analysis to identify what parameters the model is more sensitive to, distinguishing between parameters that affect the baseline, and parameters that affect the difference between baseline and prevention scenarios. Such an analysis can be used as a tool to estimate the error of model estimations due to the uncertainty of the parameters.