Biomedical Engineering Reference
In-Depth Information
Nanoscale and other biotechnological areas of research transport the engineer into uncomfortable
venues. Whether a risk is acceptable is determined by a process of making decisions and implementing
actions that flow from these decisions to reduce the adverse outcomes, or at least to lower the chance
that negative consequences will occur.
19
Risk managers can expect that whatever risk remains after their project is implemented; those
potentially affected will not necessarily be satisfied with that risk. It is difficult to think of any situation
where anyone would prefer a project with more risk than the one with less risk, all other things being
equal. But who decides what comprises “other things being equal?” It has been said that “acceptable
risk is the risk associated with the best of the available alternatives, not with the best of the alternatives
which we would hope to have available.”
20
Since risk involves chance, risk calculations are inherently constrained by three conditions:
1. The actual values of all important variables cannot be known completely and thus cannot be
projected into the future with complete certainty.
2. The physical science of the processes leading to the risk can never be fully understood, so the
physical, chemical, and biological algorithms written into predictive models will propagate
errors in the model.
3. Risk prediction using models depends on probabilistic and highly complex processes that
make it infeasible to predict many outcomes.
21
The “go or no go” decision for most biomedical and biosystem engineering designs or projects is based
upon a kind of “risk-reward” paradigm, which balances benefits and costs.
22
This creates the need to have
costs and risks significantly outweighed by some societal good. The adverb “significantly” reflects two
problems: the uncertainty resulting from the three constraints described above and the “margin” between
good
versus
bad. Significance is the province of statistics, i.e., it tells us just how certain we are that the
relationship between variables cannot be attributed to chance. But, when comparing benefits to costs,
we are not all that sure that any value we calculate is accurate. For example, a benefit-cost ratio of
1.3 with confidence levels at a range between 1.1 and 1.5 is very different from a benefit-cost ratio
of 1.3 with a confidence range between 0.9 and 1.7. The former does not include any values less
than 1, while the latter does (i.e., 0.9). This value means that even with all the uncertainties, our
calculation shows that the project could be unacceptable. This situation is compounded by the second
problem of not knowing the proper margin of safety; that is, we do not know the overall factor of safety
to ensure that the decision is prudent. Even a benefit-cost ratio that appears to be mathematically high,
i.e., well above 1, may not provide an ample margin of safety, given the risks involved.
The likelihood of unacceptable consequences can result from exposure processes, from effects pro-
cesses or from both processes acting together. So, four possible permutations can exist:
1. Probabilistic exposure with a subsequent probabilistic effect
2. Deterministic exposure with a subsequent probabilistic effect
3. Probabilistic exposure with a subsequent deterministic effect
4. Deterministic exposure with a subsequent deterministic effect.
23
A risk outcome is deterministic if the output is uniquely determined by the input. A risk outcome is
probabilistic if it is generated by a statistical method, e.g., randomly. Thus, the accuracy of a deterministic
model depends on choosing the correct conditions, i.e., those that will actually exist during a project's life
and correctly apply the principles of physics, chemistry, and biology. The accuracy of the probabilistic
