Background: Forecasting future baseline morbidity rates is crucial to air pollution benefits assessment and disease burden analysis. Although numerous methods have been developed, no standard epidemiological approaches have been proposed. The purpose of this review is to provide an overview of available methods and examine sensitivity of forecasts to method selected. Methods: A literature search on morbidity forecasting was conducted using PubMed and Google Scholar. Sensitivity analyses were conducted using age-specific hospitalization rate data and population estimates. Results: Morbidity forecasting approaches can be grouped into standard regression models and dynamic microsimulation models. Regression models assume that predictor and dependent variables are uniformly correlated over time. However, this is not always the case. While the aging population is expected to increase morbidity, this may be partially offset by improved medical care and reduced exposures to risk factors like smoking. Dynamic microsimulation models have emerged as a reliable tool to address this evolution. The models simulate individuals' risks and health determinants, and consider alternative "what if" scenarios to accurately project health outcomes. Using a conventional forecasting method based on population age distribution and age-specific all-cause hospitalization rates in Canada, we found that the projected rate increased from 91 per 1,000 population in 2001-02 to 102 per 1,000 in 2015-16. Conversely, the observed rate in 2015-16 decreased to 84 per 1,000, indicating an overestimation of 20%. Results differ for specific causes of hospitalization. Several studies also found that neglecting demographic dynamics and using unadjusted age-expenditure profile could overestimate healthcare expenditure by between 10% and 20%. Conclusion: Demographic dynamics and other factors must be considered to ensure the accuracy of baseline morbidity projections in air pollution health impact assessment.
展开▼
