Geostatistical predictive modeling for asthma and chronic obstructive pulmonary disease using socioeconomic and environmental determinants.

The spatial distribution of the prevalence of asthma and chronic obstructive pulmonary disease (COPD) remains under the influence of a wide array of environmental, climatic, and socioeconomic determinants. However, a large proportion of these influences remain unexplained. In completion, this study examined the spatial associations between asthma/COPD morbidity and their determinants using ordinary least squares (OLS) and geographically weighted regressions (GWR). Inpatient records collected from the secondary and tertiary care hospitals in Kandy from 2010 to 2014 were considered as the dependent variable. Potential risk factors (explanatory variables) were identified in four distinguished classes: 1) meteorological factors, (2) direct and indirect factors of air pollution, (3) socioeconomic factors, and (4) characteristics of the physical environment. All possible combinations of candidate explanatory variables were evaluated through an exploratory regression. A comparison between the regression models was also explored. The best OLS regression models revealed about 55% of asthma variation and 62% of COPD variation while GWR models yielded 78% and 74% of the variation of asthma and COPD occurrences respectively. Relative humidity, proximity to roads (0-200 m), road density, use of firewood as a source of fuel, and elevation play a vital role in predicting morbidity from asthma and COPD. Both local and global regression models are important in assessing spatial relationships of asthma and COPD. However, the local models exhibit a better prediction capability for assessing non-stationary relationships of asthma and COPD than global models. The geostatistical aspects used in this study may also provide insights for evaluating heterogeneous environmental risk factors in other epidemiological studies across different spatial settings.