The average and the individual II – Quality and Quest Diagnostics

April 21, 2009

crimeI have argued for some time that individual results as well as averages are important in evaluating diagnostic assays. EP21, a CLSI standard on total error and EP27, a new standard (not yet available) on error grids are useful in providing analysis of individual as well as average results.

Apparently, Quest Diagnostics does not agree but has paid the price – 302 million dollars and pleading guilty to felony misbranding. The press releases are here and here.

I have previously blogged about Quest Diagnostics about fake news (the source for that is here).

This is from the Quest mission statement:


The patient comes first in everything we do. We strive to provide every patient and every customer with services and products of uncompromising quality – error free, on time, every time. We do that by dedicating ourselves to the relentless pursuit of excellence in the services we provide.

This is excerpted from Quest’s statement about their commitment to quality and Six Sigma:

Six Sigma is a rigorous, data-centered approach to process improvement with the goal of delivering “virtual perfection” — meaning consistently high quality. Motorola first introduced Six Sigma in 1979 and companies such as General Electric and Allied Signal use it to dramatically reduce errors by trying to prevent them from happening in the first place. Quest Diagnostics began its Six Sigma journey in 2000 and is committed to being a Six Sigma Company in the eyes of its employees, customers and shareholders.

Looking at the 2008 proxy statement, there in no one on the executive committee with a quality title.

A Comment after reading the latest version of EP23 (Laboratory Quality Control Based on Risk Management)

April 2, 2009

emptyEP23 is much improved from its initial state. Skimming it prompted the following comment.


A university comes up with a new marker. It is picked up by a manufacturer, who starts development to commercialize it. This is an opportunity for FMEA. There are no results reported and no observed errors, only the probability of errors.


Towards the end of development, the manufacturer tests the system both in-house and in clinical laboratories (clinical trials). There are still no reported results because the assay is not yet released, although errors can now be “observed” through suitable classification and FRACAS can be used.


The assay is approved and sold. A clinical laboratory buys the assay and before reporting results, can perform FMEA because no results have yet been reported.


Note that the above is for a single assay.


The clinical laboratory starts using the assay and reporting results. The clinical laboratory can track observed errors through FRACAS, and typically, FRACAS is performed for the entire laboratory not just one assay.


Clinical laboratories need to use both FMEA and FRACAS, with FRACAS as probably the major tool used. There is virtually no mention of FRACAS in EP23. Reducing the rate of observed errors reduces risk.