The Study Of Analytical Power Of Control Rules For Interpretating Proficiency Test Data

Objective In order to select appropriate rules for interpretating Proficiency Test (PT) data to revel the causes of analytical errors, we studied the performance characteristics of each rule quantified in terms of its probability for false rejection (Pfr) and its probability for error detection (Ped).Method We used computer simulation techniques to study the performance of rules with limits based on the SD index (SDI), and percentages of allowable error (EA). Firstly, determing the parameters used in simulated conditions for analytes: (1) the group mean ((?)) and group SD (Sg) were obtained from the 2006 CAP Proficiency Survey CN3-A, (2) the internal laboratory SD (Si) was based on our IQC data over 6 months. Secondly, assessing the analytic quality of the method according to the indicator Sigma, which is calculated from the formula: Sigma=(TEa-Bias)/CV. Seven analytes for simulation were selected to span the range of quality from marginal to Six Sigma. Thirdly, simulating PT events. Analytical errors, including systematic and random error, were simulated in terms of multiples of Si. Systematic error was simulated by adding shifts ranging from 0 to 3.0 Si in increments of 0.5 Si, giving a total of 7 error levels. Increased random error was simulated by increasing the standard deviation from 1.0 Si to 3.0 Si in increments of 0.5 Si, giving a total of 5 error levels. For each analyte, 1000 trail analyses of groups of 5 PT samples were simulated at each error level. Then, investigating the simulated data. We calculated the SDIs of the simulated values according to the formula: SDI= (simulated value-(?))/Sg. Every SDI value was checked against the limit of each SDI rule. On the other hand, we assessed the simulated data's deviation from the group mean as a percentage, which was checked against the EA rules' limit. Rule violations were tabulated. We calculated Pfr as the proportion of violation signals when only the inherent error of the method existed in 1000 simulated events. Ped was calculated as the proportion of violation signals when each stated amount of systematic error or random error was included in 1000 simulated events. Finally, drawing the Power Function Graphs, graphs of probability of error detected vs magnitude of error.Results A series of power function graphs are gained in the unit of the ratio of Sg/Si, or the value of Sigma. Graphs of SDI rules, based on Sg/Si, include 0.5, 0.7, 0.9, 1.2, 1.4, and 1.6 six grades; graphs of EA rules are based on the range of Sigma, including 2～3, 3～4, 4～5, 5～6, and＞6 five qulity levels.Conclutions Selecting appropriate control rules for interpretating PT data through Power Function Graph according to the analytical quality and the ratio of Sg/Si, laboratories can achieve expected Ped while maintaining lower Pfr, and then improve the validity of quality control.