Statistical process control charts.

An outbreak of invasive aspergillosis among allogeneic bone marrow transplants: a case-control study. Infect Confvol1985;6:347-355. 3. Sherertz RJ, Belani A, Kramer BS. et al. Impact of air filtration on nosocomialA@ergil-lus infections. Unique risk of bone marrow transplant recipients. A discriminant scorecard for diagnosis of invasive pulmonary asper-gillosis in patients with acute leukemia. Statistical process control (SPC) is possibly the most enticing gadget in the industrial quality control toolbox. It promises much. While reading John Sellick's article,' an old aphorism came to mind: " There is no such thing as a free lunch. " The potentials of SPC are dual: A) that control charting of clinical variables will reveal " opportunities for improvement " by directing scrutiny to events that involve special causes of variation; and B) that a clinical process , once tuned to eliminate special cause variation, is as well-suited as it can be for alterations aimed at reducing common cause variation or producing more desirable mean values of a process variable. The A-B sequence is crucial to quality improvement (CQI). A feeds to CQI signals sorted from noise. B seems a safe approach to the hornet's nest inherent in improving clinical care because it limits opportunities for drawing erroneous cause-effect inferences after details of care are altered to improve outcome. Shewhart2 derived SPC from theoretical considerations that involve normal (ie, Gaussian) distributions, but it is a common misconception that SPC is hampered for processes whose inherent variation is other than normal. " Being in control " is not tantamount to " being in a normal (or Poisson or binomial) distribution " and vice versa. Dr. Sellick's discourse on SPC's origin hints that he may think otherwise. Wheeler and Chambers " have compared charting of normally distributed data and data from a variety of non-normal distributions (Burr, chi-square with two degrees of freedom, right triangle, uniform, and exponential) for hypothetical in-control processes. Shewhart 3-sigma charts give false alarms for a meager 1% to 2% of process data in this test. In these instances, SPC would have correctly advised managers with 98% to 99% accuracy to leave in-control processes unchanged. I am confused by the statement that " the number of sigma that defines the control limits will determine the number of times that an out-of-control signal will be erroneous. " This is non-sensical and should have been nailed by reviewers. What is meant by the word erroneous? A few …