Application of patient data exponentially weighted moving average method in the establishment of internal quality control model for valproic acid therapeutic drug monitoring project by LC-MS
10.3760/cma.j.cn114452-20241108-00614
- VernacularTitle:患者数据指数加权移动平均法在LC-MS法丙戊酸室内质量控制模型建立中的应用
- Author:
Qi GUO
1
;
Yungang PU
;
Jing HE
;
Sihai LING
;
Canjun RUAN
;
Chunyan ZHOU
;
Xiangyi LIU
;
Chengeng LIU
Author Information
1. 首都医科大学附属北京安定医院 国家精神疾病医学中心 国家精神心理疾病临床医学研究中心 精神疾病诊断与治疗北京市重点实验室，北京 100088
- Publication Type:Journal Article
- Keywords:
Quality control;
Internal quality control;
Exponentially weighted moving average;
Liquid chromatography tandem mass spectrometry;
Therapeutic drug monitor
- From:
Chinese Journal of Laboratory Medicine
2025;48(5):656-661
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To establish a practical patient-based internal quality control method for valproic acid drug concentration monitoring.Methods:Observational Study. A PBRTQC model based on the exponentially weighted moving average (EWMA) method was established using Python. All results of a total of 28, 757 valproic acid concentration data from February 1, 2023 to January 31, 2024 were collected and split into training set and validation set at a ratio of 80% and 20% respectively. The truncation limit (TL) was optimized by using the winsorized mean method and the trimmed mean method. Different weighting coefficients λ were set. Different TL and different λ were combined with the EWMA algorithm into different patient-based real-time quality control (PBRTQC) models. The optimized models were verified by introducing simulated constant errors (CE) and proportional errors (PE) respectively. The false positive alarm rate (FAR) was used to evaluate specificity, and the average number of patients before error detection (ANPed) was used to evaluate sensitivity. According to the daily test volume and quality target requirements, we comprehensively judged whether the performance evaluation indicators of FAR and ANPed meet the laboratory requirements. Bias detection curve was used for determination of the best model.Results:The parameters of the best PBRTQC model for valproic acid drug concentration monitoring are: trimmed mean method with 1.5 standard deviations (i.e., truncating data outside 1.5 standard deviations of the data mean), λ=0.01. The performance verification result shows that ANPed of CE and PE of this model are both less than 100. The comparison between the EQA results and the EWMA results show that the EWMA method results are comparable to the EQA results.Conclusion:A PBRTQC model for the valproic acid drug concentration monitoring project based on the EWMA method has been successfully established. It is comparable with both IQC and EQA results, which means PBRTQC may be used as a supplement to the quality control of daily quality control products.
