Introduction: To establish an assessment table of acute rejection for kidney recipients from deceased donors (DD) based on random forest algorithm.

Methods: Recipients in our center from January 2015 to October 2020 were included in this study. Length of follow-up was at least half of the year. All the recipients were randomly divided into two cohorts in a 7:3 ratio, including training cohorts and validation cohorts. In the training cohorts, a random forest classification model was established to screen variables that affect acute rejection. Finally, an assessment table was constructed based on the results of random forest and clinical practice. The assessment table was then validated in the validation cohorts.

Results: Totally 1206 recipients were included. 111 recipients developed acute immune rejection within half of the year. The training and test datasets comprised 844 and 362 patients, respectively. In the training set, donor BMI, age gap between donor and recipient, donor age, recipient age, and cold ischemia time were the most important factors for acute rejection. In the training datasets, patient stratification by different risks had different incidence of acute rejection. The rate was 7.2%, 10.5%, and 22.6%, respectively (p<0.05). In the test datasets, the acute rejection was 8.0%, 13.5%, and 26.7%, respectively (p<0.05), and the difference was statistically significant. Intermediate-risk and high-risk recipients had a higher rate of DGF and 1-year mortality compared with low-risk recipients (p<0.05). There were no differences in infection rate between the three groups.

Conclusion: 1. Our assessment table was an effective tool to assess an individuals risk of developing acute rejection in kidney recipients in China; 2. For intermediate-risk and high-risk recipients, more immunosuppressive agents should be considered under dynamic infection monitoring.