Postoperative complications following cancer surgeries are still hard to predict despite the historical efforts towards the creation of standard clinical risk scores. The differences among score calculators, contribute for the creation of highly specialized tools, with poor reusability in foreign contexts, resulting in larger prediction errors in clinical practice.
This work aims to predict postoperative complications risk for cancer patients, offering two major contributions. First, to develop and evaluate a machine learning-based risk score, specific for the Portuguese population using a retrospective cohort of 847 cancer patients undergoing surgery between 2016 and 2018, predicting 4 outcomes of interest: i) existence of postoperative complications, ii) severity level of complications, iii) number of days in the Intermediate Care Unit (ICU), and iv) postoperative mortality within 1 year. An additional cohort of 137 cancer patients was used to validate the models. Second, to support the study with relevant findings and improve the interpretability of predictive models.
In order to achieve these objectives, a robust methodology for the learning of risk predictors is proposed, offering new perspectives and insights into the clinical decision process. For postoperative complication's the mean Receiver Operating Characteristic Curve (AUC) was 0.69, for complications severity mean AUC was 0.65, for the days in the ICU the Mean Absolute Error (MAE) was 1.07 days, and for one-year postoperative mortality the mean AUC was 0.74, calculated on the development cohort.
In this study, risk predictive models which may help guide physicians at estimating cancer patient's risk of developing surgical complications were developed. Additionally, a web-based decision support system is further provided to this end.
On the predictability of postoperative complications for cancer patients: a Portuguese cohort study
