Development and validation of an algorithm for identifying patients undergoing dialysis from patients with advanced chronic kidney disease

• DOI: 10.1007/s10157-024-02614-3
• Publication Date: 2025-01-06T17:50:19Z
• Authors: Takahiro Imaizumi; Takashi Yokota; Kouta Funakoshi; Kazushi Yasuda; Akiko Hattori; Akemi Morohashi; Tatsumi Kusakabe; Masumi Shojima; Sayoko Nagamine; Toshiaki Nakano; Yong Huang; Hiroshi Morinaga; Miki Ohta; Satomi Nagashima; Ryusuke Inoue; Naoki Nakamura; Hideki Ota; Tatsuya Maruyama; Hideo Gobara; Akira Endoh; Masahiko Ando; Yoshimune Shiratori; Shoichi Maruyama
• Keywords: Nephrology

Abstract Background Identifying patients on dialysis among those with an estimated glomerular filtration rate (eGFR) < 15 mL/min/1.73 m 2 remains challenging. To facilitate clinical research in advanced chronic kidney disease (CKD) using electronic health records, we aimed to develop algorithms identify laboratory data obtained routine practice. Methods We collected of eGFR from six core hospitals across Japan: four for the derivation cohort and two validation cohort. The candidate factors classification models were identified logistic regression stepwise backward selection. ensure transplant not included non-dialysis population, excluded individuals code Z94.0. Results 1142 patients, 640 (56%) currently undergoing hemodialysis or peritoneal (PD), including 426 763 214 379 prescription PD solutions perfectly dialysis. After excluding prescribed solutions, seven parameters algorithm. areas under receiver operation characteristic curve 0.95 0.98 positive negative predictive values 90.9% 91.4% 96.2% 94.6% cohort, respectively. calibrations almost linear. Conclusions ml/min/1.73 . This study paves way database nephrology, especially non-dialysis-dependent CKD.