Here are our Top 5 Health Conditions… Number 2 will shock you!

Okay I’m not really going to tell you that. But I have been working in healthcare data long enough to see multiple reports that rack-and-stack the most common health conditions in a population, e.g. from health insurance claims data and other grouper tools. Often these are compared against some general average, and sometimes there are curious remarks, e.g. “I didn’t realize how many X we had.” Most of these get chalked up under the hype-laden term, “actionable insight.” It’s really not an insight, though.

Don’t settle for “most common health condition” reports. Here’s why.

The problem with this “top X diseases” approach is that it doesn’t really tell you that much. Lots of people have certain diseases… but are those numbers higher or lower than we’d expect in our population? Also, is it accumulating faster than we would expect?

These basic statistics are called incidence (how fast disease accumulates) and prevalence (how common a disease is) in a population. They’re a basic aspect of epidemiology. And because two of the biggest predictors of disease prevalence are age and gender, it’s worth looking at these graphs with a comparison group in mind — i.e. what would we *expect* the prevalence of each condition to be, for a group of the same age & gender mix? You can go further and incorporate other factors, but age and gender tend to account for a lot.

That said, I haven’t seen anyone do this well yet. It’s tricky to do — you have to be big enough or otherwise have an age/gender stratified benchmarking data set available.

But what you get from this age-gender matched comparison approach is different, and a good way of sizing the opportunity for disease prevention and management:

Instead of: “20% of our population has Disease A”

We get “20% of our population has Disease A, versus 40% expected.”

See how the first statement is even misleading and frightening? If you’ve ever dealt with consultants or vendors in the health insurance and/or benefits world, these numbers will often come with a thumb-in-the-wind “oh your population is younger/older, so that’s not surprising.”

Really? What would be surprising?

Similarly, you might be tempted to say, as a benefits manager or healthcare population health strategist, “Oh, look how common Disease A is… we should really focus on Disease A.” But if it turned out that Disease A was half as prevalent as you’d expect, then does it really deserve as high a priority as, say, looking at another disease with a higher than expected prevalence?

Likewise you can look at year-over-year cohorts — or the same group of individuals (rather than a cross-section of the population at the time). This cohort approach means you can follow the same group to see how many new cases of a disease there are. The same pattern applies — it’s tempting to oooh and ahhh at the biggest numbers. But are they bigger than you’d expect them to be? Or smaller?

The British Medical Journal has a nice handy-dandy reference to teach folks about comparing disease rates.