Development and internal validation of a risk score for incident obesity among Peruvian formal workers

Adult obesity is a growing public health problem, and workplace settings offer opportunities for early identification of individuals at increased risk. Longitudinal evidence to predict the transition from normal weight or overweight to obesity among workers remains limited in Latin America. This study aimed to develop and internally validate an interpretable score to estimate incident obesity risk among Peruvian formal workers with a baseline body mass index of 18.5 to <30 kg/m².

This was a retrospective cohort study using occupational health records and Cox survival modeling to develop and internally validate a risk score predicting incident obesity over time among Peruvian workers without obesity at baseline. The outcome was the first subsequent assessment with BMI ≥30 kg/m². Candidate Cox models were fitted using baseline clinical, behavioral, and occupational predictors. Performance was assessed using the C-index, Brier score, 36-month calibration, internal bootstrap validation with 500 successful replicates, and decision curve analysis. The score was derived from the coefficients of the final model.

The primary cohort included 9390 workers; 987 developed incident obesity. The final model included age, sex, baseline BMI, waist circumference, occupation, night work, smoking, alcohol use, and high baseline blood pressure. Among 7979 participants with complete data, the model had an apparent C-index of 0.857, an optimism-corrected C-index of 0.855 after 500 successful bootstrap replicates, and a 36-month IPCW Brier score of 0.091. The observed 36-month risk was 0.91%, 5.62%, and 37.05% in the low-, intermediate-, and high-risk categories, respectively.

The score showed high discrimination and acceptable calibration for stratifying occupational risk of incident obesity. External validation and local recalibration in other occupational populations are required before routine implementation outside the source setting.