Decisional trees in renal transplant follow-up.

INTRODUCTION

The predictive potentialities of application of data mining algorithms to medical research are well known. In this article, we have applied to a transplant population classification trees to build predictive models of graft failure, evaluating the interactions between body mass index (BMI) and other risk factors. The decision trees have been widely used to represent classification rules in a population by a hierarchical sequential structure.

PATIENTS AND METHODS

We retrospectively studied 194 renal transplant patients with 5 years of follow-up (128 males, 66 females, mean age at time of transplant of 43.9 +/- 12.5 years). Exclusion criteria were: age < 18 years, multiorgan transplant, and retransplant. The BMI was calculated at the time of transplantation. In the classification algorithm, we considered the following parameters: age, sex, time on dialysis, donor type, donor age, HLA mismatches, delayed graft function (DGF), acute rejection episode (ARE), and chronic allograft nephropathy (CAN). The primary endpoint was graft loss within 5-years follow-up.

RESULTS

The classification algorithm produced a decision tree that allowed us to evaluate the interactions between ARE, DGF, CAN, and BMI on graft outcomes, producing a validation set with 88.2% sensitivity and 73.8% specificity. Our model was able to highlight that subjects at risk of graft loss experienced one or more events of ARE, developed DGF and CAN, or has a BMI > 24.8 kg/m(2) and CAN.

CONCLUSIONS

The use of decision trees in clinical practice may be a suitable alternative to the traditional statistical methods, since it may allow one to analyze interactions between various risk factors beyond the previous knowledge.