Longitudinal Trends in Noncommunicable Disease Risk Factors and Premature Mortality in Saudi Arabia: A 33-Year Ecological Time-Series Study with Machine Learning Prediction

Abstract

Background/Objectives: In Saudi Arabia, noncommunicable diseases (NCDs) are an increasing public health concern, with almost 70% of deaths related to chronic diseases. The study aimed to analyze 33-year trends in NCD risk factors and apply machine learning (ML) models to identify ecological associates of premature NCD-related mortality, sex-specific analyses and project trajectories to 2030. Methods: A longitudinal ecological time-series design which used WHO Global Health Observatory (GHO) NCD Indicators (1990–2022; select lipid indicators from 1980). Five supervised regression ML models—OLS, LASSO, Ridge, Random Forest, and Gradient Boosting—were trained with TimeSeriesSplit cross-validation (five folds) to preserve temporal order and prevent data leakage. A formal PELT changepoint algorithm confirmed trend breakpoints. Linear projections to 2030 were estimated with 95% prediction intervals. Results: Adult obesity increased by +20.6 percentage points (pp) over 33 years. Under a no-policy-change scenario, female obesity is projected at 50.3% by 2030 (95% PI: 50.0–50.5%). Premature NCD mortality declined by −5.9 pp. Under TimeSeriesSplit CV, all models yielded negative R², confirming LOOCV R² = 0.98 reflected shared time-trend artefacts; the ML component is reframed as descriptive feature-importance analysis. The obesity sex gap (female minus male) was the strongest ecological associate of premature NCD mortality. Diabetes treatment coverage showed a strong inverse ecological association (r = −0.913). Conclusions: NCD risk factors in Saudi Arabia are evolving in complex ways. Targeted interventions addressing sex-specific disparities and healthcare system performance are urgently needed to meet national and global NCD targets.