Q&A: Prescriptive Analytics to Play Big Role in Transforming Healthcare (Part 2 of 2)

As healthcare becomes more efficient and effective, prescriptive analytics will play a crucial role, according to Meg Aranow, and expert in healthcare and BI.

In the second half of a two-part interview with TDWI, Meg Aranow continues her discussion with BI This Week on how healthcare companies are using prescriptive analytics to increase efficiency and effectiveness and to manage risk. (Editor's note: You can read part 1 here.)

Aranow is senior director of research and insights at The Advisory Board Company, a membership-based global research and consulting firm focused on healthcare. She has 25 years of experience in healthcare, with expertise in business intelligence and analytics, IT strategy and planning, clinical information systems, electronic medical records, and vendor assessments. Prior to joining The Advisory Board Company, she served as VP and CIO of Boston Medical Center, and held leadership positions at Partners Healthcare and Brigham and Women's Hospital. In 2009, she was appointed by the Massachusetts governor to the Healthcare IT Council for the Commonwealth of Massachusetts, which oversees the state's development of a regional extension center and health information exchange.

BI This Week: Can you share some specifics on how healthcare companies are using prescriptive analytics?

Meg Aranow: Here's an example I presented in our recent Webinar with TDWI and my firm, The Advisory Board Company. The Steadman Hawkins Clinic of the Carolinas partnered with River Logic, a company that specializes in analytics across industries, not just in healthcare. The Steadman Hawkins Clinic of the Carolinas wanted to optimize its physician and ancillary mix -- in addition to seeing the physician at this clinic, probably an orthopedist, you might have radiology studies done, and pharmaceutical intervention, and some rehab and physical therapy facilities -- all considered ancillary interventions. The clinic wanted to optimize the mix of services at each location for better capacity and scheduling. Essentially, as in any industry, they wanted to consider their fixed and variable assets and maximize the value and return they could bring.

Using prescriptive analytics helped them think through optimization and other issues. The idea was to take in all the data points to understand the optimal mix of physicians to technicians, for example, or to operating rooms. How many operating rooms did they need to maximize their physician throughput? They took in all of those constraints and data points to figure out the optimal way to design a facility and staff it. By doing that, they were able to get $20 million a year net profitability improvement in their facilities. They were also able to increase the predictability and reliability of their financial predictions from a 30 percent variable to a 2 percent variable. They were able to improve their current performance and their forecasting. That was one great success story.

A second case study I offered during the Webinar, from an organization that did not yet want to be named, studied 100-plus homes treating more than 500 patients with disabilities. In their case, they wanted to use prescriptive analytics to optimize their patient placement to make sure, as in the previous example, that they were taking advantage of their staff and the locations of the group homes to maximize the number of patients they were able to serve, as well as the profit margin.

Some of the things they considered for their analytics models were, for example, what are the care requirements of the patients we are serving, where are our group homes located, what does our staffing look like, what different types of training do they have supporting what different types of intervention, what are the staff constraints around the scheduling, how many days a week can they work, how many miles can they travel, and how long does it take to travel from group home to group home?

The organization put all of that into an optimization model, and while they have yet to implement some of the recommendations, I believe they have identified a 10 percent cost reduction if they change the way they are servicing patients. That's based on information they are seeing in the model, and it means changing the way they are doing their patient placement in order to be provide more efficient and effective care.

One side note that they brought to the table, which I thought was interesting, was that it is a highly unionized organization. The study had the unintended consequence of easing union negotiations, because they had such strong data to use in the negotiations regarding staff issues -- expected travel times and so forth.

Those both sound like cases where analytics was hugely helpful. Given that, what holds healthcare providers back in pursuing prescriptive analytics? What are some of the biggest roadblocks that you see your member organization confronting?

I mentioned that prescriptive analytics is one of the more complex models, certainly more so than descriptive and predictive. Because healthcare is a bit late to the game, we are still collecting and organizing the data that will be necessary for these models. We have done a lot with electronic medical records and other data collection methods, but we are behind other industries in terms of our implementation. We are doing that now, but if you compare us to industries such as retail and banking, which have been at it for decades, we are still relatively new to the process.

Also, many would argue that the data we are collecting is very complex and variable, probably much more so than some of the non-healthcare industry leaders in analytics. We have very complex data that we need to organize in ways that are useful for these analytic models. It's just a question of maturity; we're now in a bit of catch-up mode.

How are prescriptive analytics being used to help healthcare organizations manage risk? What kinds of risk are we discussing here?

When you think about healthcare as a business, we need to think about risk. Often, when we are talking about healthcare, in particular new models of healthcare and how we wish to transform healthcare as it is delivered in the U.S., we're talking about managing risk.

We're talking about three kinds of risk: quality risk -- meaning the quality of the care intervention that's being delivered; performance risk -- meaning the efficiency with which we are doing our jobs; and utilization risk, meaning how often people are coming to our facilities. Are they coming only when truly necessary, or are there things they could be doing at home or in clinics, away from our expensive facilities?

We think that by managing quality, performance, and utilization risks, we can have a more efficient healthcare system overall in this country -- one that is creating higher-quality health among our citizens, at a lower cost to our citizens. We spend a lot of time thinking and talking about how to manage those risks. We do that in many ways, but BI and analytics certainly has a role to play as we think about how we would manage risk, just as it does in any industry trying to manage costs and efficiency.

We might use prescriptive analytics to look at our quality risk -- things such as identifying variation in practice. What are best practices for specific interventions such as total hip replacement surgery? We are looking for all variations possible, because we know that unwarranted variation is the antithesis of quality. You drive good quality by doing things the same way over and over, so we might use prescriptive analytics in that way.

Looking at performance risk, we might look at such things as the evidence-based guidelines I talked about. We might use descriptive analytics there -- if we know this is the best way to practice, to do a hip replacement, let's measure all the physicians who are doing that procedure. Are they all using best practices? If they aren't, why not, and what can we learn from that? Was the variation warranted? We want to measure and find out.

If we look at something a little more challenging, predictive analytics, we might look at things such as patient compliance. We know that when patients aren't compliant with the medication regimens we've assigned to them, if they don't take their medications at the times we tell them to and for the duration we tell them to, that they are more likely to have complications, particularly for chronic conditions. We might want to know who among our population is most likely to be non-compliant? Who should we especially watch to make sure they are being compliant? We can use predictive analytics to create that focus.

Finally, going back to that readmission model I mentioned earlier, we might use prescriptive analytics to help us with patient discharge planning -- to help us understand which patients are at risk for an unplanned readmission, and how we might better intervene to avoid a readmission. On the business and planning side of things, we might use prescriptive analytics to understand, where might I optimally place a new facility? If I want to create a new clinic to handle outpatient services, in which community might I do that? Which services would be most in demand? How would I optimally use the new building space? What physicians should be there, serving what populations? What should their hours of operation be? How big should the parking lot be? How many will drive versus taking public transportation? All of these planning steps can be aided by using prescriptive analytics.

If prescriptive analytics is the most difficult of the three levels of analytics we've discussed, and also the most rewarding, what are some best practices you can share for getting started with prescriptive analytics?

As with all analytics, the first place to start is to identify your targets. You might start by identifying which of the three risks we talked about that you want to manage or focus on first. You can't do everything all at once. What is the performance metric that you want to improve? These aren't IT questions -- you're always starting with the business and clinical leadership within healthcare, understanding what the institutional priorities are, and what specific tactics would best support the broad, visionary goals of the institutions. From there, you back in to what kind of analytics would best support the effort.

We keep using this example of unplanned readmissions. If that is a priority for a healthcare organization, then you back in. How can analytics generally -- not just prescriptive analytics -- support my organization in trying to make improvements to its metric of unplanned readmissions? It wouldn't surprise me at all if optimization would then be included on a list as one of the mechanisms that would help that effort.

Generally, what we advise is that you start with a narrow focus. You're not necessarily trying to do something large with prescriptive analytics; you want to do one thing. You want to make a contribution to one specific goal with a narrowly focused prescriptive analytics initiative. That's going to help the organization understand the type of data that's required, the type of analytic skills that they need on the team, and how hard it is to then implement them.

Analytics is just one component of the work we have to do. Analytics simply presents an intervention recommendation based on the data. Getting the cultural change that you need to actually make that part of the work process within your organization -- and this is true in any organization -- that's the hardest part of the work. It's very hard to get people to change their habits or the way that they've always done things.

That all has to come together in order for analytics to have an impact on the organization, up to and including the implementation challenges that will have to be faced, not usually by the analytics team but by the operational leadership -- the leadership of the organization.

Also, starting with a narrow focus gives you a demonstration project, if you will, so that you can build more awareness of the power of analytics and the power of prescriptive analytics specifically. You can then build a broader base of both interest and competency within the organization, in a way to make meaningful change happen.

TDWI Membership

TDWI Members Get FREE anytime access to all reports and publications

Individual, Student, & Team memberships available.