Prediction models and nomograms for incident risk of chronic obstructive pulmonary disease in adults in Shanghai

Chronic obstructive pulmonary disease (COPD) is a major public health burden in China, but remains underdiagnosed, with most cases identified only after significant lung function decline. Effective, low-cost tools are therefore needed to identify high-risk individuals for targeted spirometry. This study developed risk prediction models and nomograms to guide personalized COPD prevention.

A community-based cohort of 32,327 adults aged 40-74 years without COPD was enrolled in 2016 and followed until March 2024. COPD diagnosis was obtained from Shanghai Songjiang District Healthcare Information Platform. Stepwise Cox regression analyses were conducted to select predictors, including demographics, lifestyle, disease history, and medical resources use, for overall and sex-specific prediction models. Discrimination and calibration were assessed by C-index, AUC, and Brier score, with 10-fold cross-validation and sensitivity analysis for validation.

Over 7.0 years of follow-up, COPD incidence was 7.98 (95%CI: 7.61∼8.23) per 1,000 person-years, higher in men (10.60) than women (6.23). Sex, age, BMI, cooking habits, physical activity, smoking, occupational exposure, family history of respiratory diseases, other respiratory diseases, outpatient visits/year, and antibiotic prescriptions/year were significant predictors in the overall model. Male predictors additionally included central obesity, while female predictors additionally included fruit intake, history of tuberculosis and menopause. C-index was 0.780/0.770/0.781 (overall/male/female); Brier score was 0.035/0.044/0.028; Time-dependent AUCs at 1, 3, and 5 years ranged from 0.783 to 0.813, indicating good predictive performance over short, medium, and long term follow-up periods. Nomograms were created for overall, male and female populations.

The study developed overall and sex-specific COPD risk prediction models with C-indices exceeding 70%. Corresponding nomograms provide practical tools for supporting routine COPD risk prediction to facilitate individualized preventive care.