A preoperative nomogram and web-based clinical decision support system for predicting early renal function after living donor kidney transplantation: a retrospective multicenter cohort study

Background: Early post-transplant renal function is a key determinant for predicting long-term graft survival and overall patient prognosis after kidney transplantation. Existing predictive models primarily focus on deceased-donor kidney transplantation (DDKT) recipients or rely on post-transplant variables, making it difficult to predict early renal function recovery in the living-donor kidney transplantation (LDKT) setting. Materials and Methods: This study conducted a retrospective cohort analysis of 3,335 living kidney transplant recipients from Institution A, Institution B, and Institution C. Data from Institution A and Institution B were used as the training set and internal set, while Institution C data were used for external validation. The Comprehensive Model was developed using pre-transplant characteristics of both the donor and recipient, while The Basic-Parameter Model was developed using only recipient age, sex, body surface area (BSA), donor age, and donor whole kidney volume. Model performance was evaluated using Mean Absolute Error (MAE) and Root Mean Square Error (RMSE), and The Basic-Parameter Model was implemented as a Nomogram and a mobile calculator app. Results: The Comprehensive Model demonstrated high predictive accuracy with an MAE of 0.15 and RMSE of 0.36 in the internal set, and maintained generalizability with an MAE of 0.14 and RMSE of 0.18 in the external set. The Basic-Parameter Model performed similarly to The Comprehensive Model, with comparable predictive accuracy. Additionally, a Nomogram and mobile app were developed, demonstrating their potential utility as clinical decision-making tools. Conclusion: This study developed a practical model for predicting early renal function recovery using basic pre-transplant variables from both recipients and donors. The model demonstrated high generalizability through external validation and can be easily implemented in clinical practice as a predictive tool. The Nomogram and mobile app provide real-time decision support, enhancing their applicability in clinical settings.