*Purpose: Predictive models capable of identifying patients at risk for readmission are in development. However, it is unknown how accurate clinicians are at predicting early post-transplant readmission in kidney and liver transplant recipients.

*Methods: In this single center prospective study, we electronically surveyed providers within 24 hours of any kidney or liver transplant discharge from January 25th to September 31st, 2019. Providers surveyed included primary surgeons, nephrologists and transplant pharmacists for kidney recipients and primary surgeons, hepatologists and nurses for liver recipients. All were asked to predict whether the patient would be readmitted within 30 days (yes/no) and the suspected cause of readmission. Readmission data were extracted from our Transplant DataMart and reported as cumulative incidence of readmission. The association between provider’s prediction and patient 30-day readmission was studied using logistic regression. Prediction accuracy by provider type was assessed by estimating the area under the ROC curve (confidence intervals estimated by bootstrapping) as well as the sensitivity, specificity, recall and F-score. Agreement between the different providers was measured using Kappa Scores.

*Results: Overall, 42 kidney transplant recipients over 140 (30%) and 21 liver transplant recipients over 65 (32.3%) were readmitted. Kidney and liver transplant surgeon response rate was 85.7% and 87.7%, respectively. Surgeons’ prediction of readmission was significantly associated with 30 day readmission in kidney (OR 3.50 [1.57-7.79]) but not liver recipients (OR 1.80 [0.57-5.70]). However, the prediction accuracy was low, c-statistic=0.64, for kidney transplant recipients and 0.57, for liver transplant recipients. Many readmissions were missed (sensitivity: 54% in kidney recipients, 40% in liver) and many patients predicted to be readmitted were not readmitted (low recall: 50% in kidney, 44% in Liver) (Table 1). Agreement between providers was low to moderate and no significant difference in accuracy was found by provider type in either kidney or liver transplant recipients (Table 2).

*Conclusions: Assessing the risk of post-transplant readmission is challenging for clinicians. The development of accurate predictive models of post-transplant readmission are needed to help clinicians identify patients at risk for readmission and target preventive interventions.

« Back to 2020 American Transplant Congress