Healthcare Scene’s recent Boston conference, EXPO.health, walked the audience through numerous innovations in health IT, especially where analytics give a boost to patient care. Still, I detected at the show, next to the crisp digital images of modern technologies, a sepia-toned photograph of older practices. The two stood out against each other in some ways, but could also be entwined. Sometimes one would encounter the hoped-for future on the conference stage and then walk into the tired reality of today’s health care on the showcase floor. This article will look at both sides and the way their interaction shapes health care.
Good Omens for Value-Based Care
I have expressed worries before about the measurement of quality in health care. At EXPO.health, I talked to Robin Roberts and Joy Rios, who consult with health care institutions about how to apply for quality-based payments, while also running the “HIT Like a Girl” podcast. Robin said that “we are still going at a crawl” in creating and using quality measurements properly. For instance, she confirmed the complaints by critical access hospitals that quality measures tend to underestimate the morbidities of low-income, vulnerable populations, and therefore to also underestimate the costs of treating them, thereby underpaying the hospitals.
On the other hand, a recent study by Blue Cross Blue Shield of Massachusetts suggests that value-based care pays off. BCBS of Massachusetts was one of the first insurers in the US to institute a value-based care program, stepping forward years ago around the time that Massachusetts also instituted universal health insurance coverage. The study showed that, indeed, providers are reducing costs and receiving bonuses for doing so. However, the article cited here reveals only a very limited change in culture. It mentions only two measures the health care providers have taken to reduce costs, and only one of those involves a longitudinal, behavioral program.
So it’s nice to get more evidence of success, this time from Shore Quality Partners of New Jersey, a modest-sized clinical provider with 240 doctors on staff or in its network. Although a group of that size does not generate a very large data set, simple analytics can still produce value. I noticed, during the talk by Cliff Frank, that Shore tends to use analytics to confirm what management already knows and then to render their plan of action compelling and concrete so that their doctors will come along. The success of Shore confirms the claim often made that health professions respond to data.
Frank started with the basic weakness of ACOs: their best efforts are constantly in danger of being undermined by outside competitors. ACO patients can go to anyone they choose, constrained only by their insurance plan. You can institute quality measures within the ACO to prevent costs from flying out of control, but competitors using fee-for-service can push unneeded procedures on patients, ruining your bonuses while padding their own pockets. A patient feeling pain, for instance, may go to an expensive orthopedist and undergo an unnecessary procedure without consulting either the patient’s insurer or her PCP at the ACO.
Here are three examples where Shore drives change through analytics. In each case, managers reminded doctors how their own personal reimbursements would reflect their efforts to lower costs:
- In order to reduce radiology costs, Shore set up its own radiology clinic. But a large, expensive, and lavishly advertised network of radiologists in their area was dominating referrals. At first, Shore’s doctors hardly ever referred patients to their own clinic. Showing them the difference in costs led to a small improvement, but the real breakthrough came when the analysts used outcomes to prove that the expensive outside company produced outcomes that were no better than the ACO’s clinic. Thus, they reduced costs while saving a lot of patients from painful and ineffective procedures.
- The orthopedics situation was even worse. Not only was a large, external orthopedics firm more expensive, but it recommended surgery 50% more often. Again, analysts proved that staying in-house was cheaper and had equally good outcomes. They persuaded their doctors to reduce referrals to the expensive firm, but they couldn’t reach the patients who would go there first when they felt pain. So the managers used predictive analytics to determine people who were at high risk of developing pain, and started intervening with effective early responses before the pain struck. The experiment is in too early a stage to judge its success, but it potentially will exemplify the gold standard for preventative, value-based medicine.
- Many labs set up stations within Shore, paying rent (which upon analysis looks more like bribery) to be right next to Shore’s doctors. Of course, such in-house labs are convenient for both patient and staff, but they’re very expensive. Showing doctors the cost savings persuaded them to use cheaper labs outside the hospitals.
Shore analyzes data using the Group Practice Reporting Option (GPRO) within the Physician Quality Reporting System (PQRS). They will move up to Level C of the Medicare Shared Savings Program this coming January, which means they are willing to take on the risk of being penalized for poor financial performance in exchange for higher bonuses when they do well. Still, Frank says, “we’re nibbling at the margins” of what can be done with analytics. He offered two pieces of basic advice: keep the analytics simple, and tackle one thing at a time.
Keynoter Lygeia Ricciardi reported statistics showing that value-based care is on the rise, with 25% of doctors expecting more than half their reimbursements to come through value-based programs in 2021. She tied value-based care to patient engagement, which we’ll turn to next.
CEO = Chief Experience Officer
From the moment health care professionals started to bandy about the term “patient engagement”, I reacted with cynicism. The term seemed to parry the clamor among patients for “empowerment,” offering them an emasculated and anodyne substitute. However, Ricciardi reports a dynamic approach to engagement at Carium, which appointed her to be their Chief Transformation Officer. The key to engagement there is to use high tech to enable low-tech, time-tested basics: they put patients in touch with coaches who guide them through salutary choices in food and other everyday activities, guided by real time data.
Ricciardi used her keynote to lay out the aspects of patient engagement, To me, the central point is holism. This means adopting a public health viewpoint, which nowadays gets the fancy term “social determinants of health”. The most educational example I’ve heard of, where providers address SDoH, took place a few years ago in Oregon. A man kept coming into the ER with difficulties breathing, and they were treating him with medications. Eventually one of the staff thought of visiting his home. Noting mold everywhere, they put in new rugs and bought him an air conditioner. The ER visits ended.
One of Ricciardi’s examples was the illustrious provider Geisinger, which prescribes food for their patients and makes sure they get healthy options. Dignity Health is a faith-based non-profit that provides grants for SDoH.
Ricciardi recommended appointing a Chief Experience Officer, who would be responsible for driving culture change. She said that when a culture of caring about the patient is in place, the other necessary things will follow. She also mentioned the potential for telemedicine to save the patient time and trouble.
One of the legacies that health providers are saddled with is their old EHR data. Despite the importance of such data for analytics (let alone correct diagnoses), George Florentine reported in his session, “How to Harness the Power of Legacy EHR Data”, that most providers throw out old data quickly–often after 18 months.
Why do they discard data? Mostly because they’re afraid a lawyer will demand it as part of a lawsuit. The providers’ assumption that their records will compromise them legally represents an interesting self-assessment of their quality of care. But they probably are ignoring old data when it could offer valuable perspectives on their patients–better, then, in their minds, that nobody knows.
The reason for Florentine to bring up these lapses was to promote the solution provided by his company, Flatirons Digital Solutions. Here is where old and new technology meet: Flatirons uses a machine learning engine from OpenText to bring old EHR data into a single model with 22 categories. Exposing the data in XML, it allows the data to be exported easily and frees the user from worries about lock-in.
Flatirons can also graph data in several ways to reveal outliers: for instance, a patient who doesn’t have many medical conditions but takes a lot of medications. Florentine reported that the data they gather is particularly popular in ERs, where staff need a lot of background information in a hurry and don’t know the patient.
Other examples of adapting to old technology at the Expo:
- The Fax Guys simplify the sending of fax documents from computers. They integrate with the major EHRs.
- Foxit Software turns documents of many formats into PDFs, stored compactly. Editing, signing, and indexing are among their services.
Some Interesting Companies in Analytics
Among the companies I met on the Expo show floor were:
- VLink, which creates a number of health apps and analytics solutions, including one that predicts patients prone to sepsis. This improves decision-making and patient care by reducing the time spent on manual sepsis surveillance.
- Verto, which claims to generate valuable insights for improving care and reducing costs from the relatively small amounts of data generated within a single institution (as we also saw with Shore Quality Partners). I assumed that because each patient is unique–and is also treated in a unique manner by each doctor, a situation Verto is trying to adjust–one would need Medicare-sized databases to derive useful analytics. But according to representative Michael Millar, the wide variety of variables (features or dimensions in machine learning lingo) is a benefit rather than a drawback. They can find correlations in unusual places.
- ClosedLoop.ai offers a comprehensive analytics package through Software as a Service. Clients can run entire toolchains from a ClosedLoop catalog for common tasks, or can write their own toolchains. Onboarding, feature engineering, and deployment are all automated. Their representative at the show, Andrew Eye, claimed that data scientists are overly concerned about data cleaning, because machine learning can derive impressive insights even from very inconsistent input.
- SurgeryLink creates workflows for surgery, which requires complicated preparations and intake procedures. According to representative Marc Montalto, even though hospitals and clinics do these procedures thousands and thousands of times, staff and patients have to perform many of the steps (such as faxing requests for equipment and services) manually every time. So someone might forget to order a part needed for the surgery, or to remind the patient not to drink beforehand, and the whole procedure has to be cancelled. SurgeryLink suggests digital technologies that can automate more of the workflows, by following and recording what the staff does.
The benefits cited by these and other vendors shows that health IT and machine learning have a lot to offer. We’ve got to pump value-based care up to 100%, and rendering analytics both more relevant and more accessible will enable that goal.