Probably the answers to these questions depend whether and how much a laboratory’s pre-analytic processing is automated, and for how long the local process has been in place. We always know the potential exists in the lab for a specimen load error to actually send results to a chart. But how many of us have actually looked?
In my (automated) lab, we turned everything upside-down in 2014 and have largely placed functions into new silos on a new track and middleware solution. People are now placed at “Inlet”, “middleware/result”, and “analyzer” stations instead of the “one tech per automated spoke” approach. We re-designed our index testing approach too. This might sound great, because reducing redundancies in work along an automated track should be a goal. One person loads specimens rather than 4 people. One person monitors and releases results rather than 4 people. Index testing occurs only on tests that are actually affected by the index. There is a beauty to this logic, but there are also extremes to any such re-vamp…of course there are.
How much change is too much? And, how do you really know when you’ve taken the refinement too far?
For me, the moment of realization that I had gone a little too lean with the track functions came when I found that an uncentrifuged plasma specimen had been inappropriately loaded to the track, tested as whole blood, and the results from this testing reported. The root-cause: the tube had been loaded to the wrong inlet lane (the “capped, spun” lane rather than the “capped, un-spun” lane). Here is where the dark side to silos of function on the new track became clear. The specimen had settled enough, between initial results and automatic repeat testing, that it looked to the middleware “person” like the repeat results were true. There was no question of specimen integrity, no one pulling the tube to inspect it. How did I know this? The audit log showed the hemolysis index raw result was over 10,000 mg/dL, and the lipemia index raw result was over 1,000 mg/dL, reflecting whole-blood hemoglobin, and likely the analysis of cell membrane fragments. All elements of the audit trail reflected automated function with no human intervention (aside from releasing the results, held because of the initial testing exceptions). My conclusion: re-design elements for the new track and middleware contributed to the problem.The new work approach was TOO siloed, the index testing TOO selective.
Realizing the issue, that day, I implemented hemolysis and lipemia indices on every track-based serum and plasma test, and updated the middleware rules to hold all 4+ hemolysis and lipemia results for verification. I verbally trained the staff at huddles to be on the lookout for the pattern, until I could get more definitive data. Then I pulled data – about 5,000 baseline hemolysis index results. I focused on the 4+ hemolysis index scores and pulled the raw data from audit trails, along with any lipemia data. A 4+ hemolysis score was seen in 0.3% of our specimens…in our lab, 3 to 5 tubes per day. The presence of whole blood (4+ hemolysis and 4+ lipemia, reliably) reflected a rate of about a tenth of this – 3 in 10,000. Rough sigma process calculation: 4.93. This, I knew within a day of the investigation.
I then monitored for about a week with the expansion of index testing (which created about 1,100 index results per day), and saw that the rates of hemolysis did not change (thankfully – my fear was that it would suffer a large up-tick), but did start to see the patterns of tests that reliably gave “exception” status, physiologically absurd results, critical or extremely abnormal results, and physiological results in whole blood. This pattern didn’t take long to find. A job aid to clue in the operators and formal training followed shortly, along with a daily report of any specimens with the “double 4+” index results for back-end analysis.
In the 2 months following the initial change and 6 weeks following formalization of the detection process, we have not had any results inappropriately released on a whole blood specimen, and the “double 4+”s don’t really make it to the daily report any more, except for occasional specimens from very ill inpatients that still fall into the range, but with significantly lower raw results (e.g., H of 500-1000 mg/dL; L of 150-250 mg/dL). We remain cautiously hopeful that this trend will continue.
So, have you looked? What’s your system? I’m sure that readers would be thankful for information about other approaches.