Future projections of elderly obesity in the United States using time series models

Abstract

This study aims to forecast the prevalence of obesity among the elderly population (aged 65 and over) in the United States through 2035 using time series forecasting techniques. Obesity data from 2013 to 2022 were analysed using six models: Autoregressive Integrated Moving Average (ARIMA), Long-Short Term Memory (LSTM), Gated Recurrent Units GRU, Random Forest (RF), Vector autoregression model (VAR), and eXtreme Gradient Boosting (XGBoost). The primary goal is to inform future public health strategies and optimize healthcare resource allocation for the aging population. The results indicate a consistent rise in obesity rates. ARIMA predicted an increase from 30.6 % in 2022 to 35.0 % in 2035, while VAR estimated 37.9 %. Machine learning models forecasted sharper growth: RF projected 40.6 %, LSTM 41.3 %, and GRU 39.8 %. XGBoost anticipated the highest rate, reaching 44.3 % in 2035. Model performances were evaluated using coefficient of determination (R²), mean square error, root mean square error, and sum of squares error. VAR and XGBoost achieved the best results (R² = 0.9995 and 0.9993, respectively), while LSTM (R² = 0.9004) and GRU (R² = 0.8648) showed moderate predictive power. ARIMA also performed well with R² = 0.9420. The findings reveal that ensemble and multivariate models, particularly XGBoost and VAR, offer higher forecasting accuracy. This study fills a gap in the literature by focusing on elderly obesity projections and offers valuable insights for developing targeted intervention policies and health programme.