Abstract
BACKGROUND: Persistent pulmonary nodules are increasingly identified in patients recovering from coronavirus disease 2019 (COVID-19). However, factors associated with long-term persistence remain insufficiently understood. OBJECTIVE: To determine whether a predictive model integrating clinical and CT imaging features can estimate the risk of pulmonary nodule persistence at 6 months after COVID-19. METHODS: In this single-center retrospective cohort study, 419 patients with newly detected pulmonary nodules after confirmed COVID-19 infection who had ≥ 6 months of follow-up were included (January 2020-December 2024). Clinical and computed tomography (CT) features were collected. Predictors were selected using least absolute shrinkage and selection operator (LASSO) regression and incorporated into a multivariable logistic regression model. Model performance was assessed using receiver operating characteristic curves and calibration analysis. Internal validation was performed using 1,000 bootstrap resamples to estimate optimism-corrected performance. Decision curve analysis was also conducted. RESULTS: Among 419 patients, 210 (50.1%) had persistent nodules at 6 months. In age- and sex-adjusted analyses, ≥ 4 hospitalizations, prior tuberculosis, larger maximum nodule diameter (OR per mm increase: 1.121, 95% CI: 1.074-1.170), vascular convergence sign positivity, and ICU admission were associated with persistence. LASSO selected four key predictors, and multivariable analysis confirmed ≥ 4 hospitalizations, prior tuberculosis, larger nodule diameter, and vascular convergence sign as independent risk factors. The model achieved an AUC of 0.728, with bootstrap-corrected AUC of 0.717. Decision curve analysis demonstrated clinical net benefit within threshold probabilities of 50-83%. CONCLUSION: The proposed clinical-imaging model effectively identifies patients at higher risk of persistent pulmonary nodules after COVID-19 and may assist in optimizing individualized follow-up strategies.