Making Sense of Results
Whether reading a body fluid smear, a culture plate, or chemistry panel, we are paid to make sense of results. Decades of bench work teaches techs to know what is expected and to focus on the unexpected. Many of the techs working today make what is difficult look effortless because of their experience.
Doctors and nurses will miss that when it's gone. And new techs coming into the field will naturally create an experience gap. So, what do we do about it?
If experience makes experts, then expert systems will be made to replace experience. That's one of the promises of the information age – from HAL the shipboard computer in 2001 to modern video games – but in a practical sense the manipulation of large amounts of data makes applied heuristics a reality. Computers think, in other words, if we tell them how.
Consider a univariate approach to quality control, either testing a known (commercial control) or unknown (patient sample) for random variation. This approach is limited by the application of one rule – in or out? – that makes it easy to detect a single system failure but harder to detect subtle flaws.
A multivariate approach, by contrast, applies multiple rules that consider more than one analyte, different limits for different settings, and other situations. Rather than looking for a point of failure it looks for patterns that seem improbable.
Cheap computing power makes this possible. Delta checks and decision-making software are two examples. But the broader field of physiological patterns can be harvested beyond "normal" ranges. An anion gap, for example, can be compared to an albumin or creatinine given your patient population to increase its sensitivity. The AST to ALT ratio is another example. Your information system can be programmed to look for unexpected statistical variation within or across runs.
If a shortage of experience begets expert systems, I wonder what this will teach new techs who enter the field. We all need these new tools to make more sense of results.