Other Publications by HCSL Fellows
A. Design and evaluation of statistical quality control
1. Hatjimihail AT. A microcomputer program for the evaluation of alternative quality control rules and procedures. [Abstract]. Clin Chem 1990;36:1010.
We have developed a microcomputer program for the evaluation and comparison of alternative quality-control (QC) rules and multi-rule procedures. The program simulates the application of a QC rule or procedure during consecutive runs of an analytical procedure. It estimates the average run length, the probability for false rejection and the probability for error detection for various values of random and systematic error. Two power function graphs are plotted; one describing the performance for random error and the other for systematic error. When two alternative QC rules or procedures are compared the graphs of the difference of their probabilities for rejection for various values of random and systematic error are plotted. The QC rules and multi-rule procedures are predefined or user defined. The ratio of the between run error to within run error, the number and the levels of the controls are user defined. The QC rules are derived from six generic rules with up to eight parameters. The user defines the QC multi-rule procedures using Boolean algebra syntax as the program includes an interpreter. The program is written in Turbo Pascal, ver. 5.0, and runs in an IBM or compatible microcomputer.
2. Hatjimihail AT. A tool for the design and evaluation of alternative quality control procedures. Clin Chem 1992;38:204-10.
I have developed an interactive microcomputer simulation program for the design, comparison, and evaluation of alternative quality-control (QC) procedures. The program estimates the probabilities for rejection under different conditions of random and systematic error when these procedures are used and plots their power function graphs. It also estimates the probabilities for detection of critical errors, the defect rate, and the test yield. To allow a flexible definition of the QC procedures, it includes an interpreter. Various characteristics of the analytical process and the QC procedure can be user-defined. The program extends the concepts of the probability for error detection and of the power function to describe the results of the introduction of error between runs and within a run. The usefulness of this approach is illustrated with some examples.
This publication presents the first interactive statistical QC simulator in clinical laboratory medicine.
3. Hatjimihail AT. Computer aided selection of alternative quality control procedures [Abstract]. Abstracts of the 2nd Mediterranean Medical Meeting. Athens: Mediterranean Medical Society, 1992:144.
B. Genetic algorithms based optimization of statistical quality control
1. Hatjimihail AT. Optimization of alternative quality control procedures using genetic algorithms. [Abstract]. Clin Chem 1992;38:1019-20.
Genetic algorithms are optimization procedures based on the principles underlying the evolution of life. Their operators - reproduction, crossover, and mutation - are derived from the processes of the molecular biology of the gene and natural selection. To explore the application of genetic algorithms in statistical quality control (QC) I have developed an interactive microcomputer program that optimizes alternative QC procedures so as to detect the total allowable analytical error with a stated probability while their probability for false rejection is as low as possible. The program is written in Turbo Pascal, ver. 6.0, and runs in an IBM or compatible microcomputer. It includes the following units:
1. The Rules Unit. The QC rules are derived from 8 generic rules with up to 10 parameters. The user selects the rules to be used and the parameters to be optimized.
2. The Interpreter Unit. It permits the user to define the QC procedures with boolean algebra syntax.
3. The Simulator Unit. It simulates the analytical process and evaluates the QC procedures by estimating the probabilities for error detection and for false rejection. The parameters of the analytical process are user defined.
4. The Optimization Unit. The user defines the order of the optimization problem. The unit searches the user defined parameter space and finds an optimal or near optimal solution. The algorithms of the unit are messy-genetic algorithms based on thework of Goldberg et al.
The program is computationally intensive and can be useful as a research tool.