Kenny Rogers & Playing the Odds to Improve HEDIS

Craig Johnson HeadshotInterview with Craig Johnson, Chief Technology & Science Officer
By Andrew Whitman

I sat down with Craig Johnson, Decision Point’s Chief Technology & Science Officer, to talk about HEDIS and how “playing the odds” can help improve HEDIS and continuously break through the rate ceiling across all measures.

Playing the odds sounds like gambling. How can you apply gambling principles to HEDIS?

I’ll start out by saying that a lot of plans have developed a certain level of proficiency in handling HEDIS – and by HEDIS I’m referring to gap closure (such as breast cancer screening) and disease control (such as diabetes blood sugar control).

That said, what I’ve also noticed is that there’s a significant amount of inefficiency in handling HEDIS: for the most part, plans run HEDIS prospectively and remind all (or most members) with open gaps to fill their gaps in care. Also, I’ve noticed that many plans don’t really start working on HEDIS until mid-year when gaps for 1-year measures start to present.

Using the gambling analogy, that’s akin to betting on all possible odds.

There’s another way. If I look at my membership and predict their likely outcome (in terms of HEDIS), I can bet my resources most efficiently based on each member’s likely behavior. For example, if a member has a 10% likelihood of filling their gaps, I know that spending my resources to engage this member will yield very little in terms of gap closure. By contrast, if a member has a 90% chance of filling their gaps, I know that they are likely to fill their gaps, but outreaching to them might be a waste of resources because they may fill their gaps on their own, with little or no intervention. Then there are those in the middle – say 50% or 60% chance of gap closure – who may fill gaps on their own, but may also benefit from assistance and reminders to fill their gaps in care.

The key here is to place your bets (or in the case of health plans, resources) on what (and who) will produce the best results.

How to I place my bets and how does timing effect these bets?

If I think of my membership holistically, members generally fall into 5 prediction categories. First there are the tails that I talked about earlier: members who are either highly likely or highly unlikely to fill their gaps.

For members who are highly unlikely to fill their gaps, my best bet (since I know they’re likely not going to fill gaps on their own) is to get to them as early as possible. There’s no sense waiting until mid-year. If I get to them as early as possible and offer them assistance in closing gaps, I’m basically giving them the full year to comply. My expectation for this group is very low (in terms of gap closure) and if I’ve done this for a couple of years and can anticipate what kind of yield I can expect from these members. Plus, if I get even a small percentage of these member to engage long-term, I have now converted these members from very high risk to moderate risk – something that I can build upon for the future.

For members who are highly likely to fill their gaps my best bet is to wait and see if they fill their gaps on their own. I can then wait until the last possible period of time before I spend my resources on them for HEDIS. This is usually after mid-year – likely around July or August. I know my yield is going to be really high for these members, but now I’m working against time (i.e. December 31) to get them to fill their gaps. As above, if I’ve done this for a year or two, I can start anticipating what the likely result of what this outreach will be.

For the three groups in the middle risk tiers, I can spend most of my resources playing the odds for this group and change their gap closure odds from a 50/50 chance to a 70/30 or better.

What I’m doing here is placing my bets on populations that are likely to yield me the best (or most optimal) results.

What about members with multiple gaps? Should we put more emphasis on these members since their behavior impacts multiple measures?

Sure. That said, we’ve found that the HEDIS member-level predictions naturally (and actually, unintentionally) account for this. Higher risk members – members who are likely not going to fill their gaps – are ones with very few gaps (i.e. 1 or 2). While lower risk members are ones with multiple conditions and appear in the denominator for multiple measures. Of course that’s not always the case, and you definitely need to prioritize members based on their risk and their gaps. The absolute hardest group of members to move are high risk (of HEDIS gap closure) and ones who are healthy – these are what we call the “healthy and unengaged”.

What about other measures, like admissions, readmissions, and other resource-based metrics? Do the same principles apply?

I would say definitely not, and that’s what makes HEDIS so unique in a health plan world. To continue the betting analogy, my best bet here would be to target higher risk members. There’s no sense outreaching to low risk members (members that are likely not to have an admission or readmission, etc.) and offer them support. Here my focus exclusively is on the highest risk members who are actionable.

Other than HEDIS rates what metrics should I use when I’m deploying this method?

The most important metric is a metric that we’ve developed here at Decision Point: it’s yield per member contact compared to a control group. For example, in the HEDIS world, my yield per contact for very low risk members (likely going to fill their gaps) is really high, but if I compare that to a control group (members not getting the intervention), the rate for the control group is equally high since these members are likely going to get their screenings on their own. So I’m always trying to maximize my yield per member contact when I’m comparing my yield to a control group.

So putting this all together, it sounds like you’re recommending that that plans bet on different member groups depending on each group’s probability of getting their HEDIS screenings. And then choosing the risk groups based on the proximity to the end of the measurement year – essentially choosing “easier to move” members the closer you get to the end of the year. Is that correct?

Yes, you’ve have it. Kenny Rogers would be proud.

Stay up to date

Keep in touch.  Give us your name and email and we’ll keep you in the loop.