Development and validation of a risk prediction model for major postoperative adverse events in elderly colon cancer using firth regression: a large-scale retrospective study

Background: Elderly patients with colon cancer are at increased risk of major postoperative adverse events, but existing risk assessment tools often require detailed clinical or laboratory information. We developed and internally validated a rapid prediction nomogram using routinely available administrative variables. Methods: This multicenter retrospective study included 4, 942 elderly patients with colon cancer from five hospitals. Major postoperative adverse events were identified using a predefined ICD-10 code-based algorithm adapted from the Healthgrades patient-safety framework. Given the low event rate, Firth penalized likelihood regression was used to develop the prediction model. Internal validation was performed using 1, 000 bootstrap resamples. Model performance was assessed by discrimination, calibration, Brier score, and decision curve analysis. Results: Independent predictors included payment method (medical insurance: OR = 2.630), admission type (outpatient: OR = 0.501), surgical approach (minimally invasive: OR = 0.473), and comorbidity index (OR = 1.041). The full model showed moderate discrimination, with an apparent AUC of 0.724 and a bootstrap-corrected AUC of 0.709. The apparent and bootstrap-corrected Brier scores were 0.0244 and 0.0247, respectively. Decision curve analysis suggested potential net benefit across threshold probabilities of 1%-51%. Conclusion: This nomogram may serve as a rapid, low-cost preliminary screening tool for estimating the risk of major postoperative adverse events in elderly patients with colon cancer. External validation and further refinement using more detailed clinical, nutritional, oncological, and functional variables are needed before broader clinical application.