Future-proofing population health: Embrace predictive analytics, social determinants and patient-generated data now

Early success stories in hospitals show how these emerging data types are fated to become critical to population health.
By Tom Sullivan
06:00 AM
Share
Future-proofing population health

The need to simultaneously reduce healthcare costs and deliver optimal care are driving the expansion of population health management initiatives among hospitals today.

Add to that mix the rising number of patients with chronic conditions, the rapidly growing Silver Tsunami population with its longer lifespans, the shift to value-based care and it’s no wonder healthcare entities are under more pressure than ever before.

So it’s also not surprising that many software vendors are rushing into the market, from the dominant EHR makers to pure-play pop health specialists.

[Also: Checklist: These 7 steps will future-proof your population health program]

What can hospitals do now to prepare for the future of population health management?

Start by understanding what products will best meet your particular needs, know that social determinants of health and patient-generated data will only become more and more important and recognize that as pop health programs evolve so too, will hospitals’ expectations for the technologies that underpin them.

Tech vendors today

Software comprises the largest chunk of the population health management market and is expected to reach $33.73 billion by the end of 2025, according to Wahid Khan, lead analyst at BIS Research.

“Cloud-based population health management software is evolving as the most preferred mode of software delivery and is anticipated to witness a high growth rate of 21.2 percent during the forecast period," Khan said.

Some of the more prominent population health companies, according to KLAS, include IBM Watson Health, Philips Wellcentive and HealthEC, right alongside EHR vendors including Allscripts, athenahealth, Cerner and Epic.

What’s more, health IT research specialist Black Book predicted that Epic, Cerner and Allscripts will emerge as dominant population health vendors amid the already-underway wave of consolidation, mergers and acquisitions.

All of which means that picking a particular product among so many available options today is tricky. But hospital and IT executives should know that HIMSS Analytics Research Director Brendan Fitzgerald said earlier this year EHRs, portals and patient engagement tools are the most popular now for population health management initiatives. The firm’s research also found hospitals implementing data aggregation, health information exchange, business intelligence and analytics software to advance pop health.

In other words: There’s no one-size-fits-all pop health platform so hospitals should understand the range while also knowing that analysts are predicting the list of products and vendors will continue to get whittled down and, as such, become more manageable in the relative near-term.

Analytics like Facebook and Amazon

As population health management projects become more and more advanced, hospitals are going to need not only new technologies but new ways of thinking about their data. Yes, this means predictive and prescriptive analytics and ultimately artificial intelligence, cognitive computing and machine learning.

Leonard D’Avolio, CEO of machine learning startup Cyft pointed to data-savvy technology giants Amazon, Facebook, Google and Spotify as particularly apt at using their information by innovating in-house. Amazon, for instance, could not have simply bought a software platform and then plugged it into its network to suddenly start what D’Avolio called “population bookselling.” What’s more, the online retailer does not consider every person to be the same consumer, just like not every patient is the same.

“Amazon understands you based on the data,” D’Avolio said. “That’s the foundation for transformation.”

Amazon, Facebook and the other companies use all of their data, not just one type, to answer as many questions as possible, D’Avolio added. They’re also harnessing cutting-edge technologies to determine those answers.

"Predictions and dashboards are merely suggestions until you put them into play."

Leonard D’Avolio, Cyft

Translating that to healthcare, D’Avolio said it’s simple to know who the sickest patients are but it’s much more useful to recognize who the most actionable people are. “Predictions and dashboards,” he added, “are merely suggestions until you put them into play.”

Consider the case of Atrius Health, headquartered in Auburndale, Massachusetts.

Craig Monsen, MD, medical director of analytics and reporting wanted to equip frontline clinicians with relevant information to help them understand the risk of specific people being hospitalized within 6 months.

“Our goal is to get these patients seen within 90 minutes,” Monsen said. 

The tricky part is to first identify them. To make that happen, Atrius evaluated off-the-shelf analytics packages and found that even the best commercial options were 10 percent off, performance-wise, when compared against its own proprietary predictive analytics — and the information they house lags behind by 3 months since they’re based on claims data.

“That’s because we’re training the models for our patients, our outcomes,” Monsen said.

Predictive analytics have helped Atrius identity specific opportunities at macro levels like which patients might need help, such as advanced kidney failure or cystic fibrosis populations. 

“You need to be more thoughtful than to just say ‘lets focus on high-risk patients,” Monsen explained.

To wit, Monsen said Atrius baked what it dubbed a PIE methodology: Population Intervention and Evaluation. 

“Our PIE framework suggests you define population and intervention together,” he said.

The evaluation slice is to determine whether or not the population and intervention are working, and how those could be improved.

“It takes clean data, data science competency — though not as much as you might think — and informatics expertise to figure out how to get it in front of clinicians and patients,” Monsen said. “And it involves empathy to understand pain points of clinicians and frontline staff so predictive analytics can help and not just add to those.”

It’s not just Atrius. The science of predictive analytics now enables such output to be accurate, said Niteesh Choudry, MD, executive director of the Center for Healthcare Delivery Sciences at Brigham and Women’s Hospital.

“Predictive analytics are being deployed at scale and quality,” Choudry said. “But predictive analytics are not magic yet – and I hope that will be soon.”

And when innovating and finding new ways to use your data, there are a few things to remember to avoid misfires.

“Sometimes simple is better,” said Ken McCardle, senior director of clinical operations at Mount Sinai Health System in New York City. “Finding practical, useful, actionable ways that data can be used is important to our analytics.”

So is providing the right tools. McCardle said an IT shop might think developing a slick new analytics self-service query tool would be really valuable, but if it’s not what clinicians actually need or want, they may never use it.

That last point is critical and it applies to patients, too.

“The more you want to harvest value from investments in IT, the greater need to create a data culture that optimizes communication internally and externally,” said Parsa Mirhaji, Director of Clinical Research Informatics at Albert Einstein College of Medicine and Montefiore Medical Center’s Institute for Clinical Translational Research. “If you really want to mature, your analytics infrastructures will have to link to outside sources of data.” 

"If you really want to mature, your analytics infrastructures will have to link to outside sources of data."

Parsa Mirhaji, Albert Einstein College of Medicine

That also means relentlessly understanding patients who are interacting with social networks, putting data out there, using devices like Fitbit or Apple Watch, doing more and more genetic tests.

Patients and clinicians, analytics and the population health programs they support are of course undergoing continuous change and that will not end anytime soon. 

“The holy grail,” Mirhaji said, “we haven’t found it yet.

Case Study: Pop health and the Quadruple Aim

Population health management is closely linked to the Triple Aim of improving patient care while lowering costs and, of course, making the population itself healthier. Not to mention the Quadruple Aim, which adds increasing physician satisfaction to the recipe.

Take Shore Quality Partners, in Somers Point, New Jersey, and Staten Island Performing Provider System as examples.

When Shore Quality wanted to reduce how much money it spent treating diabetic patients, the system turned to risk adjustment and predictive modeling technologies, as well as chronic disease support tools such as home health monitoring, according to Executive Director Cliff Frank.

Frank said it was spending $2,800 per patient every month in 2015. By harnessing the technologies and getting creative about ways to incent people to make in-person appointments, such as an Uber fund doctors could tap to send a car to bring patients in, Shore Quality cut that per-member per-month cost to $1,300 in just about a year.

“We picked up a couple million bucks,” Frank said. And he attributed that success to deploying technologies as well as the “rapid diffusion of medical advances in terms of best practice, standards of care, protocol development and implementation, and qualitative feedback to providers.”

The Staten Island Performing Provider System, for its part, has brought substantive change with population health tools.

It uses what Executive Director Joe Conte described as a robust collection of technology.

The full tech stack includes Microsoft Azure to host the data warehouse, Alteryx for data preparation and data blending, Tableau for data visualization dashboards, Python and R for machine learning and building algorithms for patient risk profiling.

“The RHIO is also an important technology for us as are other data-sharing platforms,” Conte said. Those include hot spotting and geo-mapping to glean insights from multiple data sources.

These platforms are essentially vehicles that allow for data integration across partners, facilitating low-cost transfer and economy of scale for the ETL process.

The payoff? Staten Island PPS linked EHR and state Medicaid data to create a 360-degree view of the local population that enabled it to pinpoint and engage more than 500 unique patients with multiple chronic conditions and reduce instances of unintentional drug overdoses.

Progress, and then a speed bump

While population health management has been a big part of the healthcare discourse since early 2015, the concept is hardly new. The difference this time: In addition to the EHR and pop health vendors bringing new products to market, emerging open architectures and more agile implementation, not mention common technologies such as phones, mobile devices and sensors can help doctors and hospitals address, if not curb, chronic diseases like diabetes.

Health Catalyst, in fact, published research in August showing that 68 percent of executives ranked PHM as “very important” to their evolving care delivery strategy for the next 2 years and fewer than 3 percent ranked it as not important.

Indeed, data from Signify Research found that hospitals continued increasing their investments quarter-over-quarter for two years straight.

But then the first part of 2017 saw a blip. Population health vendors reported a 7.1 percent revenue decline when compared to the fourth quarter of 2016, according to Signify.

Yes, it was all the political wrangling in Washington, D.C., and the cloud of uncertainty cast over the future of the Affordable Care Act, said Signify Principal Analyst Alex Green.

“This has had some short-term impact on the progress hospitals have made with PHM initiatives,” Green said. “Some purchasing decisions were pushed back and so inevitably it has had repercussions on PHM initiatives.”  

That said, Signify is still forecasting 12 percent year-over-year growth for 2017 when it comes to hospital buying pop health software and 16 percent year-over-year through 2021.

While hospitals continue moving their PHM plans forward, CIOs and IT shops would be smart to consider that many of the analyst firms describe the market as fragmented. Signify, for instance, pointed out that the top five vendors own 35 percent of market share.

Green suggested that more mergers and acquisitions are in store for the future, particularly given the recent and recent consolidation, notably Allscripts buying McKesson, Optum acquiring Advisory Board, NextGen grabbing EagleDream Health, and Medecision buying AxisPoint Health.

And there’s more in store for the future of population health management than pure technology.

On the horizon: Social determinants, patient-generated data

Social determinants are drawing more and more attention today and it’s safe to presume that will only continue into the future. SDOH takes into account factors such as poverty, access to healthy foods, whether the patient’s neighborhood is safe, education levels and more, with a focus on things that people experience outside the hospital and including data not typically collected.

Both Staten Island PPS and Shore Quality are working to build on their success with extensive plans for more population health management work in the future. Such plans involve new data types, notably social determinants of health and patient-generated data.

These insights can fuel positive changes for a system’s most at-risk patients.

Conte said that partnering with top utilizers of service for high-touch care coordination is an urgent priority. And Staten Island PPS is working to establish new programs to support housing, employment, correctional health services and substance abuse.

On the IT front, the next step is rolling out a uniform assessment tool across all partner EHRs for SDOH variable data capture.

“This will rapidly grow our analytic breadth with these vital elements to better target new clinical and social programs,” Conte said.

Staten Island PPS is also piloting data visualization tools to locate hard-to-reach patients who need care coordination or mental health services, for instance, by obtaining data sources that have up-to-the-minute knowledge of their locations.

Shore Quality, meanwhile, is homing in on better measurement of cost-quality outcomes and proclivity, Frank said.

The system’s goal for future pop health: “Group decision-making,” Frank said. “That includes greater collaboration among primary care practitioners, as well as specialty referrals, network utilization, clinical criteria and care models.”

Healthcare IT News Editor-at-Large Bernie Monegain contributed to this story.

Twitter: SullyHIT
Email the writer: tom.sullivan@himssmedia.com

Future-proofing population health

Find out how forward-looking hospitals are preparing.