Patient dissatisfaction following total knee arthroplasty: external validation of a new prediction model.

Abstract

Tools designed to predict patient satisfaction following total knee arthroplasty (TKA) have the potential to guide patient selection. Our study aimed to validate a model that predicts patient satisfaction following TKA. Phone surveys were administered to 203 patients who underwent TKA between 2009 and 2016 at the University of Illinois. We utilized health records to document age, gender, body mass index (BMI), and comorbidities. First, we compared the descriptive variables between the satisfied and dissatisfied groups. We then performed multivariate linear regression and multiple logistic regression to assess the predictive value of the questions in the Van Onsem et al. model. The true satisfaction rate in our study was 65%. The Van Onsem et al. model predicted a satisfaction rate of 70%. The scatter plot of predicted satisfaction score versus observed satisfaction score showed poor agreement between actual satisfaction and predicted satisfaction. Comparing satisfied and dissatisfied groups, there was a significant difference with respect to pain prior to surgery and BMI. The validity of the Van Onsem et al. prediction tool was not supported. While the predicted satisfaction rate was near the measured satisfaction rate, the model misidentified which patients were likely to be satisfied. Preoperative variables including pain, anxiety/depression, and a patient's ability to control pain symptoms showed potential for inclusion in future prediction models. LEVEL OF EVIDENCE: Level III, developing a decision model.