Simulation and modeling for optimizing quality control and improving analytical quality assurance.

Computer simulation is becoming a commonly used technique for assessing factors that affect the estimation, control, and regulation of the quality of analytical testing processes. For example, in this issue of Clinical Chemist,y, Hatjimihail (1) describes a microcomputer simulation program for the evaluation and design of internal quality-control (QC) procedures. A few issues ago, Parvin (2) discussed important considerations in the design of simulation programs to properly characterize the detection of persistent analytical errors. Ehrmeyer et al. (3) have provided an ongoing series of computer simulation studies for assessing the performance of external proficiency testing schemes, and Bennett et al. (4) have recently expanded the application to an alternative proficiency testing approach. For readers who are encountering papers on simulation and modeling for the first time, it may be useful to have some historical perspective on applications in quality control. The use of simulation in clinical chemistry was described in this journal in the early 1970s in a landmark paper by Aronsson et al. (5) from Uppsala University in Sweden. They used computer simulation to study how the analytical performance of an automated analyzer would be affected by the design of the analytical run (number and location of calibration samples, length of analytical run, etc.). At the University of Wisconsin, Hunt and 1(6) used manually simulated sets of method-comparison data to assess the usefulness of different statistics for estimating analytical errors, and Cembrowski et al. (7) used computer-simulated control data to study the optimization and interpretation of trend-analysis techniques in QC. In 1976, the Uppsala and Wisconsin groups began collaborative work to study more thoroughly the factors affecting the performance of internal QC procedures in clinical chemistry (8). An interactive computer QC simulation program was developed by Groth et al. (9) to provide a tool for investigating the many factors affecting the performance of QC procedures. With the availability of a user-friendly simulation program, the performance of QC procedures could be readily studied, resulting in a standardized assessment by use of power function graphs to describe the probability for rejecting analytical runs having different sizes of random or systematic errors (10). Studies on a wide variety of QC procedures showed that multiple decision rules could provide advantages over traditional Levey-Jennings single-nile control charts, leading to the recommendation of a multirule Shewhart chart for application in clinical chemistry (11). This procedure, now commonly known as “Westgard rules,” found widespread application in the U.S. and is often available in the software of instrument and laboratory computer systems. Cembrowski et al. (12) extended the applications of computer simulation to study the use of patients’ data algorithms, including techniques for assessment of “mean of normals,” anion gaps (13, 14), and Bull’s algorithm for erythrocyte indices (15) and even a multirule form of Bull’s algorithm (16). On the basis of extensive simulation studies, Cembrowski and Carey (17) developed broad guidelines for practical applications of internal QC. Others also began using computer simulation for selecting and designing internal QC procedures. Blum (18) formulated more complex multirule QC procedures and compared their performance with that of Westgard rules and of single-rule procedures. Although implementation of more complex multirule procedures has been limited, there is increasing use of the 12 singlerule procedure recommended by Blum. Smith and Cossift (19) developed a microcomputer simulation program to assess QC performance and evaluate alternative QC designs. Wood (20) used computer simulation to assess the effect of skewness (nonnormal distributions of error) on power curves of multirule QC procedures. Parvin (2) considered the proper simulation approach to assess the detection of persistent analytical errors. Ehrmeyer and Laessig (21) developed a simulation program for studying external QC or proficiency testing procedures. They investigated the effects of different criteria for judging acceptability (22-24) and also predicted the performance of proposed governmental proficiency testing regulations (3, 25). These studies are of particular interest now that proficiency testing has taken on a more important role as the primary approach for U.S. governmental regulation of the quality of laboratory testing. This brief siamtnrny is not meant to be a complete review of applications, but to illustrate that extensive applications exist, which have developed over a period of nearly 20 years. During this time, much has changed, including some of the assumptions about the nature of analytical errors.