FMEA and industry
Regarding Tom DuBois’ response to the March 2012 letter to the editor on failure mode and effects analysis (FMEA) ("Inbox," May 2012), I think the point of confusion is that FMEA is viewed differently in the medical device industry than it is in the industries where FMEA was first developed and used, including aerospace, automotive and semiconductor.
Having worked in each of these industries, I have used FMEA in each of its incarnations. Neither is wrong; they just have a different focus. This is not a case of either/or. In fact, in my current position, I successfully use FMEA in many different situations.
The standard DuBois refers to, ISO 14971, was written for the medical device industry and is focused on reducing harm to the patient. So the method determines a severity of the harm and a probability of occurrence of the harm. The ranking scales used reflect this.
The older publications on FMEA focus on defects and their severity of occurrence to the customer, as well as the probability of occurrence of the defect—not the harm. In this incarnation, improved quality and reliability are the goals, and the ranking scales reflect that.
I would like to respond to the comments of Tom DuBois and Edwin Bills ("Inbox," March 2012) regarding FMEA.
Risk analysis is iterative and consists of risk identification (Is there a potential source of harm?), risk assessment (Do I really care?) and risk mitigation (Can I fix the targeted risk without increasing risk elsewhere?). Author Terje Aven wrote, "Reliability analysis may be viewed as a special case of risk analysis or as an analysis which provides input to risk analysis."1
FMEA is identified in the QS9000 automotive standard as a reliability tool that, as I mentioned earlier, is a special type of risk analysis tool. In addition, in annex G of ISO 14971:2007, nowhere does it state that FMEA is sufficient for risk analysis.
Based on that, just performing an FMEA is not risk analysis; it’s only a partial, incomplete analysis. To do a complete and correct analysis, you need to consider a top-down and bottom-up approach.
If you have quantitative, historical failure rate data, use fault tree analysis, event tree analysis or failure modes, effects and criticality analysis. If you’re giving it your best guess, use root cause analysis, FMEA, hazard and operability studies, or hazard analysis and critical control points.
In addition, remember that design FMEA (DFMEA) is different from manufacturing FMEA. You don’t use detectability in DFMEA.2 Using detectability in a DFMEA erroneously underestimates the real risk—the combination of probability of occurrence and severity of harm. It’s critical that the correct procedure be used for the correct application, especially when you’re dealing with safety in medical devices.
Reference and note
- Terje Aven, Foundations of Risk Analysis, Wiley, 2003, p. 24.
- For further information, see Mike W. Schmidt, The Use and Misuse of FMEA in Risk Analysis, Medical Device and Diagnostic Industry, 2004.