Files
Abstract
Colorectal cancer poses a significant public health challenge, particularly among the elderly population. The main treatment for this condition is surgical resection, but it comes with the risk of complications and potential readmissions. The occurrence of unplanned readmissions after colorectal surgery is a crucial indicator of treatment quality, linked to worse survival rates and substantial healthcare expenses. The Center for Medicare and Medicaid Services (CMS) Hospital Readmission Reduction Program (HRRP) links provider payments to readmission rates, and due to the paucity of existing research on factors contributing to readmissions for Medicare patients undergoing colon cancer surgery. This study examined relevant scholarly studies on colon cancer readmission rates, the risk factors, and the performance of various predictive models used in the prediction of 30-day hospital readmission after index admission discharge for colectomy. Findings show that several factors impact readmission, including patient demographics and clinical and administrative factors. Other factors contributing to a higher risk of readmission within 30 days include the location of the rectal tumor, prolonged hospital stays, comorbidities, emergency surgery, stoma creation, open surgical approach, male gender, and the hospital's geographical location. Similarly, machine learning-based predictive models (random forest and neural networks) were found to perform better than traditional statistical models (logistics regression) in predicting readmission events in Medicare patients with an optimal F1 score of 86.33% and 74.53%, respectively, against 70.94%.
In conclusion, patient demographics, such as age and gender, along with clinical factors, such as length of stay, comorbidities, surgical complications, and healthcare utilization metrics, have consistently emerged as significant predictors of readmissions. The superior performance machine-learning models can facilitate the development of reliable decision-support solutions for providers and payers for better care and outcomes. This study's findings align with existing studies on colon cancer surgery readmissions and machine learning models; therefore, they are not conclusive. More studies are required to fully understand the impacts of known and unknown variables on readmissions, especially non-clinical variables from social determinants of health data sources.