1990 Medicare/CLIA final rules for proficiency testing: minimum intralaboratory performance characteristics (CV and bias) needed to pass.

On March 14, 1990, the Centers for Disease Control and the Health Care Financing Administration published criteria for defining minimum performance in proficiency testing (PT). Using our previously described computer modeling technique, we determined the likelihood of passing PT under the new rules. The model relates combinations of intralaboratory CV and bias to PT performance criteria. For example, a laboratory with a bias of zero and an internal CV of 5% will pass a 10% fixed-limit PT criterion (i.e., the criterion for glucose analyses) 98% of the time when five samples are used. The model provides similar analyses for all PT criteria and all relevant combinations of CV and bias. The probability of passing PT decreases as the number of analytes tested increases, i.e., from 98% to 37% as the number of analytes increases from 1 to 20. A laboratory's internal CV has a greater effect on the outcome of PT than do the corresponding bias values. We conclude that a laboratory that operates with methods that have internal CVs less than or equal to 33% and biases less than or equal to 20% of the PT criteria will have a greater than 99% chance of passing PT.

Predicting effects of quality-control practices on the cost-effective operation of a stable, multitest analytical system.

The cost-effective operation of an analytical system depends on quality-control (QC) practices such as the QC procedure itself (control rules, number of control measurements); the batch size or the run length; and the use of bracketed, nonbracketed, or pre-control modes of operation. Predictive value models that predict the defect rate and test yield of each test, as well as of the system as a whole, have been used to study these practices and to develop strategies for improving the quality and productivity of a multitest analyzer. Quality was optimized for most tests by achieving high error detection and low false rejection by the QC procedures. For a few tests where ideal QC performance could not be achieved, predictive models indicate that high quality is achieved, predictive models indicate that high quality is achieved as long as the observed stabilities (low frequencies of errors) of the measurement procedures are maintained. In our laboratories, productivity gains of 2.9% ($17,400/year) were achieved by changing QC procedures. Predictive models indicate that further gains are possible by increasing batch size and changing from bracketed to nonbracketed control operation. In general, the common practice of bracketed control on stable analytical systems may need to be re-examined owing to its effect on the cost of operation.

Selection of medically useful quality-control procedures for individual tests done in a multitest analytical system.

Quality-control (QC) procedures (i.e., decision rules used, numbers of control measurements collected per run) have been selected for individual tests of a multitest analyzer, to see that clinical or "medical usefulness" requirements for quality are met. The approach for designing appropriate QC procedures includes the following steps: (a) defining requirements for quality in the form of the "total allowable analytical error" for each test, (b) determining the imprecision of each measurement procedure, (c) calculating the medically important systematic and random errors for each test, and (d) assessing the probabilities for error detection and false rejection for candidate control procedures. In applying this approach to the Hitachi 737 analyzer, a design objective of 90% (or greater) detection of systematic errors was met for most tests (sodium, potassium, glucose, urea nitrogen, creatinine, phosphorus, uric acid, cholesterol, total protein, total bilirubin, gamma-glutamyltransferase, alkaline phosphatase, aspartate aminotransferase, lactate dehydrogenase) by use of 3.5s control limits with two control measurements per run (N). For the remaining tests (albumin, chloride, total CO2, calcium), requirements for QC procedures were more stringent, and 2.5s limits (with N = 2) were selected.