Pre-operative nomogram to predict risk of peri-operative mortality following liver resections for malignancy

Abstract

Introduction: The majority of liver resections for malignancy are performed in older patient with major co-morbidities. There is currently no pre-operative, patient-specific method to determine the likely peri-operative mortality for each individual patient. The aim of this study was to develop a pre-operative nomogram based on the presence of co-morbidities to predict risk of peri-operative mortality following liver resections for malignancy.

Methods: The Nationwide Inpatient Sample database was queried to identify adult patients that underwent liver resection for malignancy. The pre-operative co-morbidities, identified as predictors were used and a nomogram was created with multivariate regression using Taylor expansion method in SAS software, surveylogistic procedure. Training set (years 2000-2004) was utilized to develop the model and validation set (year 2005) was utilized to validate this model.

Results: A total of 3,947 and 972 patients were included in training and validation sets, respectively. The overall actual-observed peri-operative mortality rates for training and validation sets were 4.1% and 3.2%, respectively. The decile-based calibration plots for the training set revealed good agreement between the observed probabilities and nomogram-predicted probabilities. Similarly, the quartile-based calibration plot for the validation set revealed good agreement between the observed and predicted probabilities. The accuracy of the nomogram was further reinforced by a good concordance index of 0.80 with a 95% confidence interval of 0.72 and 0.87.

Conclusions: This pre-operative nomogram may be utilized to predict the risk of peri-operative mortality following liver resection for malignancy.