Developing Models To Assess the Relationship Between Greenness and Asthma
A Talk by Hadiqa Tahir (NIHR Predoctoral Fellow | Research Assistant, University of Leicester)
About this Talk
Background:
The protective and detrimental effects of greenness on asthma’s multifaceted nature remain a compelling area of interest. On one hand, the presence of greenness might indicate higher pollen levels, posing risks to those suffering from asthma. On the other hand, areas of greenness might provide opportunities for recreational activities that can improve health. However, most statistics methods fail to consider how an individual’s movement beyond their residential area might influence their exposure to greenness. This oversight could skew our understanding of the true impact of greenness on asthma.
Methods:
Three models were developed with increasing complexity to explore the effects of the Normalised Difference Vegetation Index (NDVI) on asthma using spatial logistic regression models, which considered distance-weighted, pixel-level environmental exposures. These models were developed with increasing complexity by integrating interaction terms and capturing non-linear effects. A simulation study was conducted using 1,000 repetitions to assess the models’ ability to retrieve true parameter values under varying sample sizes and effect sizes. Model performance metrics: bias, empirical standard error, mean square error, and coverage were evaluated for each parameter estimate and method.
Results:
The baseline spatial logistic regression model, which used NDVI as the primary predictor, showed low bias, good accuracy, and precision in estimating true parameter values with a sample size of 500. Larger sample sizes and effect sizes lead to better model performance. Complex models, particularly those incorporating interaction and spline terms, presented significant challenges in terms of performance.
Conclusion:
Increasing model complexity led to poor performance measures and the model’s incapability of retrieving true parameter values, however further testing is required to explore, validate, and analyze model capabilities. Additionally, the scope of this project has been extended to apply these models to Demographic and Health Surveys (DHS) data, which may provide further insights into the effects of greenness on respiratory conditions across different populations.
