Tıbbi Laboratuvarlarda Hasta-Bazlı Gerçek-Zamanlı Kalite Kontrol (PBRTQC)

Abstract

Quality control in medical laboratories is divided into two categories: Internal Quality Control and External Quality Control. Within the scope of Internal Quality Control methods; "traditional quality control" approach, which is widely used and performed by periodically measuring control samples, and "patient-based quality control" approach, which has more advantages in terms of performance and cost, although its applications are limited, are used. For patient-based quality control, the Average of Normals method was first used, and then Moving Average method started to be used in later studies. In recent years, various methods using different statistics have also been proposed, and this approach is now referred to in the literature as Patient-Based Real-Time Quality Control (PBRTQC).

In the thesis study, the performances of Moving Average (MA), Moving Median (MM) and Exponentially Weighted Moving Average (EWMA) control charts as PBRTQC methods were evaluated using different parameters, and studies were conducted to recommend the method and parameters that give the best results. Due to the fact that analytes show different distributions, the need to determine separate methods and parameters for each analyte, and the difficulties experienced in the design of the PBRTQC methods in this context, PBRTQC is not widely used in medical laboratories despite its many advantages. In this context, the sensitivities of MA and MM control charts, designed with the assumption of a Standard Normal distribution, to the normality assumption were measured using observations generated from very different distributions and for parameters such as block size, truncation limit usage/non-usage and its width, in order to determine a general design for different analytes. It was concluded that the MA control chart is robust when truncation limits are not used and especially when the block size is 20 observations. It was found that MM control charts are not effective for these studies and it would be better not to use them. The results obtained with the tests performed with empirical distributions of real patient data confirmed our conclusions. A regression model was also created using descriptive statistics such as skewness and kurtosis of the distributions of the analytes to allow designers to estimate the Average Run Length (ARL) performances of the MA control chart.

Furthermore, the zero-state and steady-state performances used in quality control studies for production and service sectors were calculated for MA and EWMA control charts with truncation limits and different parameters. In this context, methods and parameters that can detect the shift in the process as soon as possible were investigated. The change in the steady-state performance of a control chart designed under zero-state conditions, which is one of the situations that can be encountered in real life, was also examined.

These studies provided control chart method and design recommendations for practitioners in medical laboratories and in this context aimed to facilitate the design of PBRTQC methods as well as expanding their implementations. Work steps for the setting and implementation of PBRTQC methods in medical laboratories and further future studies were recommended.