Association and Predictive Value of Preoperative High-Sensitivity C-Reactive Protein for Postoperative Atrial Fibrillation After Video-Assisted Thoracoscopic Lobectomy: A Cohort Study Using the INSPIRE Database

Background Postoperative atrial fibrillation (POAF) is a common and serious complication following video-assisted thoracoscopic surgery (VATS), which affects patients' medical outcomes and quality of life. The predictive value of high-sensitivity C-reactive protein (hs-CRP) as an inflammatory marker for POAF in non-cardiac surgeries remains unclear. This study aims to investigate the association and predictive capability of preoperative hs-CRP levels with postoperative new-onset atrial fibrillation. Methods This study employed a retrospective cohort design, analyzing data from the INSPIRE database, which included 3,219 patients who underwent thoracoscopic lobectomy. Patients were classified into low-risk, moderate-risk, and high-risk groups based on preoperative hs-CRP levels. Multivariable logistic regression models were used to assess the independent association of hs-CRP with the occurrence of POAF in different risk groups. Additionally, the incremental predictive utility of hs-CRP for POAF was evaluated using C-statistics, continuous net reclassification improvement (NRI), and integrated discrimination improvement (IDI), in order to quantify the improvement in predictive accuracy when including hs-CRP in existing risk models. Feature importance calculations and predictive models were generated using the Boruta algorithm and machine learning methods. Results The results indicated that the incidence of POAF increased progressively with rising preoperative hs-CRP levels (2.9% in the low-risk group, 6.0% in the moderate-risk group, and 9.3% in the high-risk group, P < 0.001). After multivariable adjustment, the risk of POAF in the high-risk group remained significantly higher than in the low-risk group (OR 2.03, 95% CI 1.14–3.64, P = 0.017). The inclusion of hs-CRP in the traditional risk model significantly improved the C-statistic from 0.753 to 0.796 (P = 0.003) and exhibited significant NRI (0.371, P < 0.001) and IDI (0.015, P = 0.017). Analysis using machine learning models revealed that hs-CRP was one of the important predictive features among many variables, with the CatBoost model achieving the highest AUC of 0.90. Conclusion Preoperative high-sensitivity C-reactive protein is associated with an increased risk of POAF in patients undergoing thoracoscopic surgery, and its inclusion in clinical risk models can significantly enhance predictive capability.