MH: For organizations that are new to data analytics, what do you see as the first steps?
Kauffman: A low-cost way to begin—and most organizations have the capability to do this in-house—is to start with a simple creation of disease registries and look for those historically high-utilizing, high-cost patients, or those patients who have problems but perhaps haven't seen their primary-care physician in the year to date, and engage those patients.
Then there's the opportunity to move on to hindsight. Again, it's the easiest information to wrap your hands around. That is the historically high-cost utilization. This presumes that past behavior is the best indication of future behavior. And then as your organization matures, endeavor to identify your rising-risk patients.
MH: It sounds as if transparency really played a role in achieving the outcomes you were seeking. How did you approach communication with physicians?
Kauffman: Years ago, when we first started down this path and multiple payers were approaching us with a list of 24 and 36 and 47 different quality measures they wanted us to track, we decided that we needed to pick a finite number of measures, develop workflows and processes and point-of-care reminder tools within the EHR for a subset of those measures and focus on those measures. And now, instead of a health plan approaching us with a myriad of measures, we approach them and say, “These are the measures we are prepared to knock out of the ballpark. Let's focus our work around these, and then we'll add a couple more next year and a couple more the year after that.”
The dashboard is fully transparent. Any of our providers can go on to our Internet site and look at their aggregate stars on these Healthcare Effectiveness Data and Information Set measures. Similarly, on a quarterly basis we send out a side-by-side report that shows each individual primary-care physician the number of patients they have in a particular contract, their medical-expense ratio, their risk score and their key performance indicators.
The newest addition to that stable of transparency tools relates to the distribution of bonus or pay-as-you-go performance dollars. Any time any of those dollars are distributed, every provider sees the same report with the provider name right there and the amount and the reason for which they were receiving incentive dollars.
Jason Dinger: Mission Point Health Partners started with 10,000 members in 2012. We're now managing the needs of over 250,000 members, and we've clinically integrated with more than 7,200 providers. We've learned a lot by being in different geographies and seeing quite a bit of variation.
The first step is getting a historical view of your population. As you know, a small percentage of people generate most of the cost, and that's one of the big challenges for ACOs. We have found, as we go on our IT journey, is that being able to stratify patients to make sure we're allocating the right amount of time to each person and engaging them in the right setting to really help them and their families is so important.
We are starting to do a lot of work around predictive modeling and machine learning, putting more and more data into kind of our data repository and letting that data get smarter and smarter about which interventions are working and which ones aren't. For example, recently we were looking at some data and found that our second call with the member is by far the most predictive of improving outcomes and lowering costs.
MH: Could you give us an example of how data stratification allows you to allocate resources efficiently and in an appropriate setting?
Dinger: The one that comes to mind is depression. Historically, we would have looked at folks with depression as asthma patients or active cancer patients and relate to them as such. But we now know that unless we can really help them through their depression, all the other conversations are just not going to have the same impact. By doing some scoring directly with members and working through our providers, we can get a little bit closer to the root cause.
MH: What would you recommend as first steps for organizations that are new to data analytics?
Dinger: I would find a partner to just help clean and standardize your data. There are a number of lower-cost solutions now on the market, and finding a partner can take a whole bunch of things off your plate as well as kind of reduce the number of potential errors. Then I'd listen and watch that data and really kind of soak yourself in what it can tell you about the people you're serving. And then I'd customize and slowly add to that data set and just let it get richer and richer for you and your partnering providers.