M.S. Applied Data Science - Capstone Chronicles 2025

2

studies that track change over time, NHANES offers a cross-sectional snapshot — capturing a moment in time across a diverse U.S. population. While this limits causal inference, it allows for robust population-level screening and risk profiling. By comparing models built with and without pharmaceutical data, this research explores how much predictive value lifestyle and behavioral indicators hold on their own. In doing so, it supports both individual health decision-making and broader public health strategies aimed at reducing metabolic risk. We hypothesize that behavioral and lifestyle features alone will demonstrate significant predictive ability, and that incorporating medication use will further improve classification performance.Addressing metabolic syndrome through predictive modeling is not only a public health priority but also a business and policy imperative; early identification of at-risk individuals can reduce long-term healthcare expenditures, improve workforce productivity, and inform targeted resource allocation in preventive health programs. 2 Background Obesity—a primary driver of metabolic syndrome—affects a substantial portion of the U.S. population, increasing cardiometabolic risk, healthcare costs, and reducing quality of life. Metabolic syndrome, characterized by abdominal obesity, hypertension, dyslipidemia, and impaired glucose regulation, has become increasingly prevalent in recent decades. Analysis of U.S. NHANES data from 2011–2018 shows that adult prevalence rose from 37.6% (95% CI: 34.0%–41.4%) to 41.8% (95% CI: 38.1%–45.7%), with particularly sharp increases among individuals with lower educational

attainment (Liang, 2023). These trends highlight the urgency of understanding not just clinical management, but also the modifiable behaviors that contribute to metabolic risk. These conditions are commonly managed through pharmaceutical interventions, particularly in clinical settings, where blood pressure medications, glucose regulators, and weight-loss drugs are widely prescribed. While medications may offer short-term benefits or be necessary in advanced stages of disease, they often do not address the root causes—many of which are behavioral and environmental. Given the increasing prevalence of lifestyle-related chronic conditions, understanding the independent contribution of modifiable behaviors to metabolic health status is essential. The National Health and Nutrition Examination Survey (NHANES) provides an opportunity to explore this relationship using a nationally representative, cross-sectional sample of U.S. adults. The dataset includes comprehensive demographic, nutritional, medical, and laboratory data, making it well-suited for predictive modeling. Specifically, variables such as fasting insulin and glucose, BMI, diet composition, physical activity levels, and medication use enable the identification of patterns associated with metabolic syndrome. This study uses NHANES to investigate whether individuals at risk for metabolic syndrome can be accurately identified based on lifestyle-related indicators alone—or whether the inclusion of pharmaceutical use adds meaningful predictive value. Although the cross-sectional nature of NHANES prevents longitudinal analysis, this study offers valuable insights into which present-day behaviors and clinical factors are most strongly associated with metabolic risk. Although prior research has examined

153