Date: Tuesday, June 4, 2019
Session Time: 6:00pm-7:00pm
Presentation Time: 6:00pm-7:00pm
Location: Hall C & D
*Purpose: Hospital readmission rates after kidney transplant (KTx) are high and readmission is associated with graft failure and mortality. Predictive models of post-transplant readmission have low accuracy. Develop more accurate predictive models of hospital readmission may require a better understanding of the risk factors of readmission and how their importance varies over time. We sought to compare the importance of various risk factors in predicting post-transplant readmission based on the timing of readmission.
*Methods: Adult KTx patients in the U.S. Renal Data System database (2005-2014) with primary MEDICARE coverage before Tx were included and followed through 12/31/2015. We built predictive models of 30day, 90day and 1 year hospital readmission using Random Forest and assessed the predictive accuracy by the area under the ROC curve. Predictors were categorized into 3 groups: Recipient factors (demographic, medical, socio-economic status), Donor factors (demographic, donor medical), and Transplant factors. The weight of the predictors were summed by group.
*Results: Among 40,810 transplant recipients, the predictive accuracy of our models were 0.61 [0.60-0.63], 0.63 [0.62-0.65] and 0.64 [0.62-0.65] at 30day, 90day and 1year respectively. Transplant factors remained the main predictive group for early and late readmission but decreased with time from 60% of the prediction at 30 days to 29% at 1-year. Although both recipients’ demographics and socio-economic factors only account for 6% and 12% of the prediction at 30 days, respectively, their contribution to the prediction of later readmission increases to 13% and 29%. Donor characteristics remained relatively poor predictors at all times.
*Conclusions: The ability to predict readmission remains low but improves over time. This study emphasizes the need to thoughtfully include risk factors in predictive models based on the time to the event of interest. Researchers aiming at predicting early readmission should focus on collecting detailed data on transplant characteristics, and patients’ comorbidities and socio-economic characteristics when predicting late readmission. These results may inform the development of future predictive models of hospital readmission that could be used in clinical practice to identify KTx recipients at high risk for post-transplant hospitalization and to design effective interventions to prevent readmission.
To cite this abstract in AMA style:Hogan J, Arenson M, Adhikary S, Li K, Zhang X, Zhang R, Valdez J, Sun J, Adams AB, Patzer RE. Timing Matters: Improving Prediction of Hospital Readmission Post Kidney Transplantation [abstract]. Am J Transplant. 2019; 19 (suppl 3). https://atcmeetingabstracts.com/abstract/timing-matters-improving-prediction-of-hospital-readmission-post-kidney-transplantation/. Accessed May 25, 2020.
« Back to 2019 American Transplant Congress