Using “art” and “interpreting healthcare data” in the same sentence may seem as sensible as comparing an octopus to a bicycle. Isn’t data supposed to be objective? So how and where does the “art” part come in?

Years ago, I cut my healthcare data teeth using the revolutionary-for-the-time Market Planner and Physician Practice Planner software developed by the original Sachs Group. It combined various data sets and use rates to create precise projections of future inpatient, outpatient and, physician services need five years into the future.

Despite the software’s mathematically driven underpinnings, you needed a light hand in interpreting and applying the data. I followed these five principles:

1. Recognize data and methodology limitations – Although we used the best available data, it still had some degree of error, especially when it came to the market physician supply numbers.  Also, since the model employed patient use rates from all of Southeast Michigan, it assumed behavior in my part of Southeast Michigan would mirror that of the whole region.  That’s probably a good assumption, but it does introduce some fuzz.

2. Rather than talk in terms of absolute need, indicate the direction of the need.  Some medical specialties indicated a current surplus of physicians while others indicated the need for more.  In other words, is there an excess or a deficit?

3. Coupled with this, emphasize the magnitude of the need.  Is the need slight or significant?

4. Next, look at the context of the overall market.  Can physicians from other specialties provide the same services for a specialty with an indicated deficit?

5. Finally, in presenting results, rather than treating results with mathematical certainty, use terms like “the model suggests a market need” or “there does not appear to be a need for additional cardiologists.”  This shows an appropriate recognition that, although the results are reliable, they are not bullet-proof.  I have seen more than one belligerent physician or hospital executive make it their personal mission to undercut the results of a modeling projection they didn’t like.  Using more “humble” language helps soften some of the criticism.

So here’s an example of how to apply all this. If an analysis of a particular market shows the need for 4.2 pediatricians, I would explain that the initial analysis indicates a need for pediatricians, and apparently it is pretty strong. However, before I send the physician recruiters to get 4 more pediatricians, I would crosscheck the supply of family physicians and internists.  If those specialties have excess capacity, it is likely that some of the need for pediatricians already is and will continue to be filled by the complementary specialties. And 4 might be overly aggressive.  I would be more comfortable recommending 1 or 2. Besides, the reality is that finding 4 more pediatricians would be very tough.

So the art of interpreting data involves a degree of humility over the limitations of the data and modeling approaches plus a nimble hand when finalizing recommendations. You can’t eliminate all criticism from people who don’t like the message, but don’t give them additional ammunition with which to shoot the messenger.