BACKGROUND: A sedentary lifestyle, and obesity, are primary factors forcing the ongoing chronic disease health crisis in the United States. The aim of the current study is to assess if an ecological framework can predict United States physical inactivity and obesity prevalence using an artificial intelligence model.
METHODS: The current study utilized several United States county-level datasets representing 12 predictive variables of the ecologic framework. A non-linear artificial intelligence statistical approach was used to assess the ability of these variables (i.e., features) to predict United States county-level physical inactivity and obesity.
RESULTS: The R虏 values demonstrated that the performance of Extra trees models was different across the two outcomes. While models for both physical inactivity and obesity prediction were significant, physical inactivity always exhibited the higher R虏 for each feature number (6-12) compared to obesity. These models' performance was also influenced by the number of features. An increase in the number of features led generally to improved model performance. For physical inactivity, the highest R虏 and lowest AIC was achieved using all 12 features, hence, the 12-feature model was identified as the optimal model for physical inactivity prediction. For obesity, the highest R虏 and lowest AIC was achieved using 10 features.
CONCLUSION: These results further support validity of the proposed ecological framework, including culture, politics, policy, and social, physical and economic environment factors in explaining variability in United States physical inactivity and obesity prevalence.
