The geriatric nutritional risk index (GNRI) is a commonly used method to assess nutritional risk for predicting potential surgical site infections (SSI) in cancer patients. This study aims to create and verify a simple nomogram and a dynamic web-based calculator for predicting the risk of SSI among gynecologic oncology patients.
A retrospective evaluation was conducted on patients who were admitted into a tertiary hospital in China with confirmed diagnosis of gynecologic cancer between 01 August 2017 and 30 November 2021. A two-piecewise linear regression model with a smoothing function was used to investigate the non-linear association between GNRI and SSI to determine the ideal cut-off point. Three models were developed on the basis of different variables to predict SSI in gynecologic oncology patients. Through a nomogram the concordance index (C-index), the Akaike information criterion (AIC), and the integrated discrimination index (IDI) were used to determine the final model. Finally, the performance of the nomogram was validated using the 1,000-bootstrap resamples method and analyzed using C-index, GiViTI calibration belts, and decision curve. Also, a user-friendly dynamic web-based calculator was developed.
A total of 1,221 patients were included in the analysis. A non-linear association could be observed between GNRI and SSI risk with a GNRI cut-off value of 101.7. After adding GNRI to Model 2 (which comprised Morse Fall Scale score, preoperative length of stay, operation time, and estimated blood loss), the AIC value decreased, the C-index value increased and IDI increased significantly. The nomogram C-index in the development cohort and internal validation cohort demonstrates a moderate-high degree of discrimination. The GiViTI calibrated belt showed a good agreement between the observed and predicted probabilities of SSI. The decision curve validates the clinical feasibility of the nomogram with a threshold value between 0 and 49%.
The GNRI cut-off value of 101.7 allowed for appropriate stratification of patients into distinct SSI risk groups. This study found that including GNRI in the above nomogram (Model 2) would enhance its potential to predict SSI in gynecologic oncology patients.