Objective The chemiluminescence immunoassay (CLIA) reagents are prone to systematic deviation of test results when changing batch numbers, which affects the continuity and comparability of biochemical test items. This study aimed to establish a reagent batch-to-batch difference compensation model suitable for clinical laboratories, and to evaluate the measurement uncertainty of the compensated results, so as to provide an operational quality control scheme for reagent batch-to-batch switching in laboratory departments of primary hospitals. Methods This study was conducted on the currently used CLIA platform in the Department of Clinical Laboratory of Xuchang Hospital. Three biochemical-related immune projects, namely creatine kinase-MB (CK-MB), cystatin C (Cys-C) and procalcitonin (PCT), were selected as the research objects, and the detection data of three consecutive reagent batches (Lot A, Lot B and Lot C) were collected. Each batch was simultaneously measured for 31 clinical remaining serum samples and 2 horizontal laboratory quality controls. Taking Lot A as the reference batch number, the compensation transfer equation of Lot B and Lot C to Lot A was established by Deming regression. In addition, the measurement uncertainty was synthesized from components such as repeatability, intrabatch difference, inter-batch difference, and calibrator assignment, and the extended uncertainty U (k＝2) was given. Results The bias ranged from ?4.8%?+8.2% for Lot B and Lot C compared to Lot A before compensation, which exceeded 1/3 of the total error allowable (TEa) of Clinical Laboratory Improvement Amendment ""88 (CLIA"" 88). After Deming regression compensation, the bias decreased to ?1.3%? +1.5%, all of which fell within the acceptable interval. The uncertainty assessment showed that the extended uncertainty of CKMB (Lot B), Cys-C (Lot B), Cys-C (Lot C) and PCT (Lot C) after compensation was 4.3%, 4.1%, 5.2% and 6.2%, respectively (k＝2). The main contribution components were the uncertainty of calibrator assignment and the regression residual. In the practical application of the established compensation model in the following 6 months, the qualified rate of return of external quality assessment (EQA) increased from 92.3% before compensation to 100%. Conclusion The Deming regression-based compensation model can effectively correct the system shift caused by the change of batch number of CLIA reagents. Combined with the uncertainty evaluation of ISO/TS 20914 framework, the reliability of the compensated results can be quantified, and it is suitable for popularization in laboratory departments of primary hospitals.