• Politics
  • Diversity, equity and inclusion
  • Financial Decision Making
  • Telehealth
  • Patient Experience
  • Leadership
  • Point of Care Tools
  • Product Solutions
  • Management
  • Technology
  • Healthcare Transformation
  • Data + Technology
  • Safer Hospitals
  • Business
  • Providers in Practice
  • Mergers and Acquisitions
  • AI & Data Analytics
  • Cybersecurity
  • Interoperability & EHRs
  • Medical Devices
  • Pop Health Tech
  • Precision Medicine
  • Virtual Care
  • Health equity

Are Predictive Analytics "a Bridge too Far" for Some Hospitals?


Marco Huesch speaks on “cozy duopolies and oligopolies” in regional healthcare markets, and what it may take for analytics innovation to thrive.

healthcare analytics,big data and analytics,healthcare analytics news,value-based care,marco huesch analytics

Earlier this week, Healthcare Analytics News posted frank insights from Marco D. Huesch, MBBS, PhD, Penn State University's Milton S. Hershey Medical Center. The interview stemmed from a “Case for Data Scientists Inside Healthcare” that Huesch had published with a colleague, and he provided specific, frontline thoughts on the state of analytics work in the American healthcare system.

In this follow-up, he speaks about pluses and minuses of utilizing outside analytics vendors (which he said academic medical centers should not), “cozy duopolies and oligopolies” in regional healthcare markets, and the types of markets in which innovation can thrive.

Do you think there’s advantages and disadvantages of using outside analytics firms?

I think there is definitely a tradeoff. What you get with external vendors is some reliability, some baseline sense of trust because they have done it before. But when you think of IBM Watson’s health division, they were thrown out of MD Anderson after a few years and $60 million in expenditures, though as far as I know they’ve succeeded at Memorial Sloan Kettering. Even the best vendors can fail.

It’s also very expensive, and if you’re a small player for just waiting and seeing what happens. Maybe Google Alphabet’s health division will come out with a free layer that just sits on top of everyone’s EMR and just does everything for free. Crazier things have happened in this space.

There is a tradeoff, but for academic medical centers there’s almost no justification for not doing it in house. It would be equivalent to just getting rid of your medical school and just being a provider. If you think of an academic medical center as having those three missions of education, research, and care, they all complement each other. You can’t outsource analytics, it’s like outsourcing education or a service line. It doesn’t seem coherent with being a center of excellence.

It seems like there would be a gold rush, almost, with MACRA and all of these other sources of obligation, but that gold rush isn’t happening everywhere.

You see this rollback on the federal MACRA incentive payments and obligations, you see health plans not really being as tight as they could be in terms of demanding certain things from their providers because of competitive issues and lack of desire to antagonize, if the provider groups are really powerful in certain markets. That could be it, there just might not be enough pressure. The other thing is, it just might not be doable, right?

We always talk about this stuff as feasible, but it’s incredibly hard. I’ve been doing this for 30 years, and I have an MBBS (which is an Australian MD degree) and a PhD and 10 years of experience in academia. A lot of these problems demand really deep thought, and sometimes they can’t be solved, or they can’t be solved in a way that’s timely or useful to managers. We have our fair share of failures, I’m happy to admit to that. If you’re coming from the outside without that situational knowledge, you may actually do worse than nothing, you might hurt what’s going on.

If you look at the job market for data scientists who would technically be able to do what we do, a lot of them are being recruited for very narrow use cases, like ‘here is a bunch of images, do some deep learning and classify what you feel is an abnormality or not.’ Very simple things, prediction models where it’s very clear what the target class is, what the features are, what the models are, the processes of going from A to Z.

What we do internally is much different from that. Someone might come to us and say ‘look, I have radiologists’ dictated speech, I want you to analyze the raw speech to see if you can detect fatigue in the raw voice. Is there disengagement, is there a tiredness that is affecting diagnostic areas.’ Those are incredibly deep, complex problems. We’re lucky we have a team of 6 now and we have a lot of resources in this academic medical center, but in a run-of-the-mill hospital that’s a science fiction scenario. I think it’s important to be fair to some hospitals where it may legitimately be a bridge too far.

Isn’t the ROI there? The argument is that the up-front is a big bullet to bite, but in the long run you’ll be saving x amount of dollars and x number of lives.

If you think of a hospital in their output market, they have a very steady duopoly or oligopoly with some other providers, and there’s say a cozy duopoly of commercial payers, and there is a legitimate competitive outcome where no one rocks the boat and no one is in a price war and no one is messing around with provider networks. If that’s the output world you’re facing, it becomes very hard to justify obtaining and analyzing data to inform strategy. The strategy is not going to change, it’s just going to be micro-level marketing, micro-level practice purchases, micro-level negotiations with payers. The whole value of doing that data gathering and analyzing is to prescribe innovative changes in practice and strategy, and there’s no absorptive capacity for those strategic thrusts in the local market, or you’d be insane to do them because you’d give up a lot of your existing market and your cozy financial outcomes.

I think it would be a very rational decision by a leader to say, ‘this is crazy, it’s like putting shock absorbers in a tractor.’ You drive a tractor, you’re going to go over some bumps, shock absorbers and rearview mirrors are stupid in that context, I can just twist around in my saddle and look. If you’re telling me it’s going to cost a million bucks a year to give me one of these features for my tractor, you’re smoking what comes out of the back of my car.

In that sense, it’s a realistic thing to say that ROI is a theoretical concept that implies that you can do innovative things to attract new customers with new value propositions at new price points in new ways of delivering care. Unless some or all of those things are true, you could write the most beautiful prescriptive analytics that will just stay as a folder on someone’s desk.

A good example [of the opposite situation] is in Los Angeles, they’ve got 82 hospitals in the county: there are no cozy duopolies. There’s a lot more aggressive, intensive competition, there’s a lot more innovation in terms of how you detail to specialists and physicians without trying to buy them all, because there’s too many. There’s a lot more clever marketing and outreach to patients to display competitive advantage, there’s more things to attract labor to work in your hospital.

Putting all this together, in general one would think a market such as the US’s would be more conducive to strategic innovation which healthcare analytics could drive. Unfortunately, because of some of the almost non-competitive arrangements one sees amongst providers and payers and amongst the two, some of that strategic innovation potential is thwarted.

Related Videos
Image credit: ©Shevchukandrey - stock.adobe.com
Image: Ron Southwick, Chief Healthcare Executive
Image credit: HIMSS
Related Content
© 2024 MJH Life Sciences

All rights reserved.