A logistic regression model for predicting health-related quality of life in kidney transplant recipients

BACKGROUND: To develop a logistic regression model capable of predicting health-related quality of life (HRQOL) among kidney transplant recipients and determine its accuracy. METHODS: Three groups of patients were selected: 70 healthy controls, 136 kidney transplant patients as a derivation set, and another 110 kidney transplant patients as a validation set. SF-36 score was used for HRQOL measurement. A cutoff point to define poor versus good HRQOL was calculated using the SF-36 scores of healthy controls. A logistic regression model was used to derive predictive parameters from the derivation set. The derived model was then tested among the validation set. HRQOL predictions made by the model for the patients in the validation set and the SF-36 scores were compared. We calculated sensitivity, specificity, positive and negative predictive values, and model accuracy. RESULTS: SF-36 scores below 58.8 were defined as an indication of poor HRQOL. The regression model suggested that poor HRQOL was positively associated with lower education (below high school diploma), being single or widowed, and diabetes/hypertension as etiology. It was negatively associated with younger age (<45 years) at the time of transplantation. Optimal sensitivity and specificity were achieved at a cutoff value of 0.74 for the estimated probability of poor HRQOL. Sensitivity, specificity, positive and negative predictive values, and accuracy of the model were 73%, 70%, 80%, 60%, and 72%, respectively. CONCLUSION: The suggested model can be used to predict poor posttransplant HRQOL among renal graft recipients using simple variables with acceptable accuracy. This modal can be of use in decision making in the recipients for whom achieving good HRQOL is the main aim of transplantation, to select high-risk patients and to start interventional programs to prevent a poor HRQOL.