Medical Director Cole Stanley continues his journey though the Circles of Healthcare Data Hell with Circle Six. You can find the other circles here.

A descent to our next level reveals a vast and treacherous sea where too many have gone overboard. It’s never been easier to access vast amounts of data, with the ability to draw on new sources and link sets together. Now with the advent of widespread generative AI, the amount of information available to us is expanding faster than ever before. Here we risk heeding the siren song of newer fancier data sources, analysis techniques, and systems, and can start to think that these will be the solution to our problems. But too often, we may be unable to separate the truly useful from distraction and noise, and end up lost at sea. I often get the feeling I’m in something akin to this circle of hell when I pick up my phone, presumably wanting to do something useful. Fast forward 15 minutes and I have flitted between various messaging apps, scrolled a bit, read a few brief anxiety-provoking headlines, and promptly forgotten why I picked up the phone. It was all interesting, surely bathing my brain with some pleasing dopamine as I scrolled  but it fully distracted from what I was otherwise trying to do.

In healthcare, a similar phenomenon can happen. EMR queries can be run, administrative data sets can be linked, visualization tools make it easier than ever to display vast data sets, and entire departments are dedicated to Decision Support. In short, it’s now relatively easy to gather and display more data, and we are flooded by it as a result. I’ve seen this in clinical operations reports in the past, where monthly graphs of many process measures are prepared, but when asked “what decisions do we make with this data” there isn’t any clear answer. A little like ordering a lab test when we know the result won’t change our management, we end up having data and measures looking for a problem.

When faced with a hard problem, we often steer towards an easier problem nearby and solve that instead (stay tuned to a future circle where we are struggling to get our troop of monkeys to dance on their pedestals).

Easier: get more data and create more measures

Harder: use these measures to inform decisions and detect improvements

It’s relatively easy for us to brainstorm more and more measures,  put them in reports, and give them to our frontline providers.  I start to wonder though, are we doing this at the expense of focusing on the hard problem, which in this case could be getting embedded QI in our daily frontline work. Though I agree that “limited access to data” can be a barrier for some QI projects, has it has become an excuse to drag our heels on getting started? After all, in QI we don’t need “the perfect measure” and I believe it’s usually better to get started with the understanding, learning, and improving sooner rather than later.

So how do we decide which data to look at and which to ignore? Well, instead of starting with the data set itself and unstructured exploration, it might be more useful to first think about what clinical problems are most pressing; where our focused efforts may have the most impact. In short, starting with “what’s important to our clinical care?” and then seeing if there are useful data sources relevant to these topics may be the better approach – a sort of Quality Planning process. Here we prevent ourselves from being led astray by data that is easy to see, instead focusing on the most important problems to us.

At CDSC, we have debated how much guidance we should give HDC Discover users when they are considering using our measures to do QI. Some advocate for a more “paint by numbers” approach, where the topic is pre-selected, while others worry that we may be distracting users from working on what is most important for them. The compromise may be that our included measures set has been informed by practicing primary care clinicians on the CDSC, and that we attempt to prioritize the addition of measures relevant to problems common to primary care. We have started selecting some common problems to develop more structured QI guidance, while also advocating that users choose problems that are important and potentially impactful for their practice. When learning QI, the most impactful thing to do may be to first choose a small-scale project with this structured guidance.

Aside from this, we’ve recently been paying greater attention to the value of measures, focusing more on building out those tied to real QI projects (e.g. UBC CPD Mental Health modules, planetary health,   heart failure QI Collaborative, opioid use disorder measures). Here we are trying to get closer to solving the hard problem of having our measures used in embedded QI, instead of solely making more measures available.

CDSC exaltations

  • Consider starting with a discussion on what important problems could require focus in your clinical context, then only after this see what data is available
  • Beware of analysis paralysis, getting lost in the vast amount of data available
  • If you are new to QI, starting with a structured small-scale QI project with some measurable clinical topic may be the most impactful given the learning potential (consider PSP support)
  • HDC is especially interested in developing measures that will be actively used in QI work

Return to Circle of Healthcare Data Hell Table of Contents