Preoperative assessment system for hand-assisted laparoscopic donor nephrectomy by discriminant analysis

Abstract

We developed a preoperative assessment system to predict surgical workload in hand-assisted laparoscopic donor nephrectomy (HALDNx) using the normal-based linear discriminant rule (NLDR). A total of 128 cases of left HALDNx performed by a single operator were used as training data. Surgical workload was measured by operative time. The optimized model had 9 explanatory variables: age, total protein, total cholesterol, number of renal arteries (numberRA), 4 variables of perinephric fat (PNF), and thickness of subcutaneous fat. This model was validated using cross-validation and the .632 estimator to estimate discrimination rates with future test data. PNF and numberRA were the predominant factors affecting workload followed by the computed tomography value of PNF, body weight, and male sex. The estimated accuracy of the prediction system was 94.6%. The complication rate was 9.38% and did not correlate with surgical workload. We also made our program available online for constructing assessment functions from other cohort data. In conclusion, the surgical workload of HALDNx could be predicted with PNF and numberRA as the dominant risk factors.