Biomedical Engineering Reference
What happens next? If you need to perform an intervention then there will be some form
of design change. Either something is modified to be stronger, or some form of detection is
included, or even a special instruction is put in the instructions for use . Whatever happens
there will be a design change that inherently means the FMEA will have changed. This is the
reason for the last five columns in an FMEA (as in Table 7.11 ).
In the first of the five columns one writes a paragraph that describes the change (or the evidence)
for the new RPN determined in the last four columns. This is not a word or a phrase - it needs to
be complete and succinct.
The second column is the same value of S from the first assessment, simply because the
severity of the failure cannot have changed, you only can mitigated against it. Consider a
single use item; the obvious danger is mistakenly reusing the device. While you can build in
indicators to stop someone from reusing it again, the severity of reuse is still the same.
You should now reevaluate your values of O and D to give new ratings O* and D*. One or
the other, or hopefully both, will have been reduced. If you have done your design correctly
D* will be 1 and O* will have been reduced. When you calculate your new value of RPN* it
should fall in one of the “safe” categories. Note that your design change must reduce O* (we
shall see the reason for this when we come to do the mandatory risk assessment later).
CASE STUDY 7.2
Consider the case of a failure mode of a personal insulin pump. Let us assume that the pump has a
display to show the time of the last infusion and the amount. Now let us imagine a nasty failure mode.
The worst I can think of is that the pump display is showing all is normal, but no insulin is being infused.
There is little doubt that this is severe; there is a chance of a patient dying (if they are remote from
help, for example when mountaineering, etc.); more likely they will collapse and require urgent medical
attention without which the failure could be fatal - hence, from Table 7.11 we have a severity rating of 4.
We now need to determine O. We have little data to go on, hence we resort to looking at the
FDA and MHRA websites to identify recalls of similar products. Equally, we can refer to any
postmarket surveillance data. We find that over the last year a similar product had four reported
failures, hence O = 3.
The detection was assumed to be D = 1 as the patient should feel “wet” as it is leaking under
The rating comes out as 12, which is in the “intervention suggested” zone. This means something needs
to be done. The infusion line was redesigned with fittings that are tamperproof and robust, and a special
band was designed to ensure the “needle” remains intact and in position. Hence the occurrence was
reduced to 2. The new table reveals an RPN of 8 which, although not perfect, is acceptable ( Table 7.13 ).
Another, highly acceptable way to analyze RPN is to use a graphical map. If you plot
S versus O on a graph you are able to establish contours for various levels of D.
Figure 7.10 illustrates an example RPN plot. As before this is not written in stone; you have to
decide your own critical RPN boundaries. This graph does, though, make the decision process