Background: Few studies have used a multi-item approach to measure diet when examining its relationship with physical activity (PA) in youth. Design and Methods: Data came from the 2019 Youth Risk Behavior Survey (YRBS). Item response theory (IRT) was applied to develop the Poor Diet Scale (PDS). Ten items were dichotomized to indicate poor diet: fruit juice, fruit, green salad, potatoes, carrots, other vegetables, milk, water, breakfast, and soda. Six variables were dichotomized to indicate PA risk: active for 5+ days (PA5d), active all 7 days (PA7d), 3+ days of muscle strengthening exercise (MSE3d), physical education all 5 days (PE5d), participation in 1+ sport teams (ST1), and combined PA.MSE. Logistic regression was used to model PA using the IRT-derived PDS scores (theta) while controlling for age, sex, race, BMI percentile, and sedentary time. Results: All items significantly fit (ps < .001) the IRT model except for soda, which was dropped from the scale. In fully adjusted regression models, odds of PA risk increased significantly (ps < .05) with every 1 z-score (theta) increase in poor diet for PA5d (OR = 1.74), PA7d (OR = 1.53), MSE3d (OR = 1.85), PE5d (OR = 1.15), ST1 (OR = 1.73), and PA.MSE (OR = 1.81). Additionally, students not trying to lose weight had significantly (p = .004) greater odds as theta increased for PA7d risk (OR = 1.68), as compared to their counterparts (OR = 1.34). Conclusions: Results highlight the utility of using multiple items to measure a complex health risk behavior.

nutrition, diet, physical activity, item response theory (IRT), adolescent health