Multi-Dimensional County-Level Risk Factors Associated with Lung Cancer Incidence in the United States: Generalized Linear Models and Machine Learning

Overview

Numerous studies examined how individuals’ health behaviors affected lung cancer incidence and showed racial and regional inequalities in the United States. My study drew county-level risk factors to predict age -adjusted lung cancer incidence rates. In a n ational sample of 2,469 counties, I used risk -adjusted Generalized Linear Models (GLMs) and Bootstrapping Machine Learning methods to test the county -level associations between healthcare access, social vulnerability, smoking rates, environmental quality, and lung cancer incidence rates. I had the following findings: (1) A county with the highest percentage of primary care physicians was associated with a 6% lower rate of incident lung cancer (Odds Ratio=0.94, p<0.05); (2) A county with the highest smoking rate was associated with a 51% higher rate of incident lung cancer (Odds Ratio=1.51, p<0.05); (3) A county with the poorest environmental quality had a 12% higher rate of lung cancer incidence (Odds Ratio=1.12, p=0.05); (4) A county with the higher percent age of black residents (Quartile 2 and 3 vs Quartile 1) was associated with at least 5% higher rate of lung cancer (Odds Ratio=1.05, p<0.01). My study demonstrates the statistically significant effects of multidimensional county -level risk factors on lung cancer incidence, advancing existing literature that focused predominantly on individual factors. This study suggests that more resources should be distributed to counties with unfavorable healthcare access and environmental quality as well as those with h igh levels of social vulnerability, especially in rural areas. Reducing smoking rates will also help decrease the occurrence of lung cancer. Missouri