In the last blog post about the problems of contract labs’ laboratory informatics data, I wrote, “Tomorrow, I’ll write about one strategy labs are taking to address speeding-up the building of methods in their ELNs.” Well, “tomorrow” (yesterday) came and went without this post so here it is, finally…
Method-Building: Speeding Up the Process
At IMACS, all the labs that had presented and who mentioned method-building gave one important strategy as their path to quick success in building a LOT of methods. They use the people who are the experts in those methods to build them. Even those labs that didn’t present seemed to be backing this strategy as the way they sped-up this part of the implementation process.
Often, we try to take people who are expert in the operation of the lab and in the methods, in general, and use them to build things like analyses and methods. Long-term, we have administrators (LIMS and ELN admins) build these records. However, the person who is building the record must spend some amount of time with the person who is the expert in it, anyway. And the true expert needs to try it out and give feedback. And there’s always rework, resulting in using more of the expert’s time to retry it.
If the tools are easy-enough to use, there’s no need to bring multiple people into the building of each method, unless more training on building them is needed.
Where we’re building the process, where we mean that we’ve got someone identifying that the process is “do this then this then this” (method execution) and where the tool to do this is easy to use, I would think using the person who actually runs the method is the best way to get the methods built. While we can all have some sympathy for the person that we’re now dumping yet another thing on their plate, it seems the most practical path to take. After all, in some companies, that person would be the one forced to write the SOP (Standard Operating Procedure) for the method, anyway, so this is a similar task.
However, where we start getting into complex calculations, especially where they need to be validated, this is the place where I’m skeptical that anyone can do this but a well-trained expert (during implementation, the consultants sent to the project or the people pulled from the lab to be trained to do this, or later on the administrator). Many of the tools used for this still aren’t that easy to use and even MS Excel has lots more options than someone who doesn’t know statistics might be overwhelmed by. Then, getting to the validation portion of this, it still takes someone who would know what it takes to get it validated to inspect and test it, probably.
My own comment is this: when you validate some of this code, if the person doing the validation can’t figure out what it’s doing, I sometimes wonder if it’s not easier to rewrite the code. When we hack some of these calculations together just to get them to work, documenting and testing is much more difficult than it would be if we took a little time to create calculations that are readable and understandable by the final viewer. Even where calculations don’t need validation, long-term maintenance of them brings up the same issues. Unless the same person who built them stays around forever (unless they never leave the company and never get hit by the proverbial bus), then the issue remains that the next person that needs to modify it is still left in something of a mess.