Usually in healthcare, there's little or no discussion of the risks involved in tampering with a system or under-controlling a system in response to a perceived quality improvement issue.
It's one of the reasons why, I'll share, that some systems don't get too far with respect to quality improvement! (Other classics include: inability to collect data, unwillingness / lack of knowledge regarding how to listen to good data, or something else blocking an organization from changing its mind in response to meaningful data...but more on that in another entry!)
We've previously discussed how data-driven process improvement projects are incredibly useful in healthcare. Of course they include some well-known items like:
- making sure individuals are NOT singled out
- being able to follow the process over time to make sure it doesn't back-slide
- guarding against the risks of tampering with a system that doesn't need adjusting (a type 1 error) and
- under-controlling a system that needs something done fast (type 2 error).
For more on type 1 and type 2 errors in healthcare process improvement, with examples, click here.
I share this entry to specifically highlight some useful questions to ask when choosing the risk of a type 1 (alpha) error you're willing to accept in your project. You likely remember from some statistics course long ago that you get to choose that alpha level you'll accept...but how?
In the entry beneath, Robert Elliot II (a Lean Six Sigma Master Black Belt) shares some useful questions about how to select the risk of tampering with a system that you or, more importantly, the process owner is willing to accept!
Click the link beneath for Robert's useful thoughts on how to select the alpha that makes sense for the team!
And that is the magic of how to actually CHOOSE a meaningful Alpha. Ask the owner of the results, not the person conducting the test, these three questions: 1. What is the action you will take if there is a significant difference? 2. What is the cost of that action? 3. How willing, as a percentage, are you to be wrong and spend that much? The answer to the last question is what your Alpha risk needs to be! Yes – I know there is more to it than that such as sample size, data collection costs, and Power. The important point here is make a conscious decision about the Alpha risk, or P Level, based on the impact of being wrong.