Development of a Prediction Model for Cranioplasty Implant Survival Following Craniectomy

Abstract

Cranioplasty after craniectomy can result in high rates of postoperative complications. Although determinants of postoperative outcomes have been identified, a prediction model for predicting cranioplasty implant survival does not exist. Thus, we sought to develop a prediction model for cranioplasty implant survival after craniectomy. We performed a retrospective cohort study of patients who underwent cranioplasty following craniectomy between 2014 and 2020. Missing data were imputed using multiple imputation. For model development, multivariable Cox proportional hazards regression analysis was performed. To test whether candidate determinants contributed to the model, we performed backward selection using the Akaike information criterion. We corrected for overfitting using bootstrapping techniques. The performance of the model was assessed using discrimination and calibration. A total of 182 patients were included (mean age, 43.0 ± 19.7 years). Independent determinants of cranioplasty implant survival included the indication for craniectomy (compared with trauma-vascular disease: hazard ratio [HR], 0.65 [95% confidence interval (CI), 0.36-1.17]; infection: HR, 0.76 [95% CI, 0.32-1.80]; tumor: HR, 1.40 [95% CI, 0.29-6.79]), cranial defect size (HR, 1.01 per cm2 [95% CI, 0.73-1.38]), use of an autologous bone flap (HR, 1.63 [95% CI, 0.82-3.24]), and skin closure using staples (HR, 1.42 [95% CI, 0.79-2.56]). The concordance index of the model was 0.60 (95% CI, 0.47-0.73). We have developed the first prediction model for cranioplasty implant survival after craniectomy. The findings from our study require external validation and deserve further exploration in future studies.