Published in:
01-03-2021 | Colorectal Cancer | Original Article
A simple prediction score for prolonged length of stay following elective colorectal cancer surgery
Authors:
Dedrick Kok Hong Chan, Jia Jun Ang
Published in:
Langenbeck's Archives of Surgery
|
Issue 2/2021
Login to get access
Abstract
Background
Current enhanced recovery after surgery (ERAS) protocols are designed for all patients without tailored programmes for at-risk groups. A risk score to determine elective colorectal cancer patients at risk for prolonged length of stay (LOS) would help to identify this group for preoperative intervention.
Methods
Multivariate analysis of demographic and preoperative variables was performed to identify independent risk factors for prolonged LOS, defined as 7 days or more. A stepwise variable selection approach using logistic regression was then used to build a risk prediction model.
Results
Among 172 patients in our population, 41.9% of patients had prolonged LOS. Five variables were included in our risk prediction model. These were age ≥ 65 years (OR 13.9 5.09–38.0; p < 0.0001), neoadjuvant therapy (OR 7.60 2.51–23.0; p < 0.0001), open approach (OR 3.96 1.68–15.9); p = 0.008), history of smoking (OR 5.18 1.68–15.9; p = 0.004) and white blood cell (WBC) count (OR 0.83/unit 0.69–0.99; p = 0.040). These variables were combined to produce a score, for which the area under the receiving operator curve was 0.82 (95% CI 0.76–0.88), and Hosmer-Lemeshow test showed a χ2 statistic of 9.14 and p = 0.519. Using 0.9 as a cut-off, the score has sensitivity of 81.9% and specificity of 65.0%.
Conclusion
A simple, clinical score can be used to predict for prolonged LOS based on preoperative variables, allowing for intervention before surgery.
Article summary
Age, neoadjuvant therapy, smoking status, open approach and WBC count are independent risk factors for prolonged length of stay following elective colorectal cancer surgery. A risk score comprising the above independent variables was developed with area under the receiving operator curve of 0.82 (95% CI 0.76–0.88), and a Hosmer-Lemeshow test showing a χ2 statistic of 9.14 and p = 0.519.