Apologies on the hiatus for posting on THCB. As many of you know, I was running around getting Health 2.0 in order this past weekend. Today we are featuring a piece on understanding how machine learning can actually work in health care today-Matthew Holt
By LEONARD D’ AVOLIO, PhD
There’s plenty of coverage on what machine learning may do for healthcare and when. Painfully little has been written for non-technical healthcare leaders whose job it is to successfully execute in the real world with real returns. It’s time to address that gap for two reasons.
First, if you are responsible for improving care, operations, and/or the bottom line in a value-based environment, you will soon be forced to make decisions related to machine learning. Second, the way this stuff actually works is incredibly inconsistent with the way it’s being sold and the way we’re used to using data/information technology in healthcare.
I’ve been fortunate to have spent the past dozen years designing machine learning-powered solutions for healthcare across hundreds of academic medical centers, international public health projects, and health plans as a researcher, consultant, director, and CEO. Here’s a list of what I wish I had known years ago.
Earlier this month, the Centers for Medicare and Medicaid Services Administrator Seema Verma proposed bold changes to Medicare’s Accountable Care Organizations (ACOs), with the goal of accelerating America’s progress toward a value-based healthcare system—that is, one in which providers are paid for the quality and cost-effectiveness of care delivered, rather than volume delivered.
CMS has created a number of ACO programs over the last six years in an effort to improve care quality and reduce care costs across its Fee-for-Service Medicare population. In a Medicare ACO, hospital systems, physician practices and other voluntarily band together and assume responsibility for the quality and cost of care for beneficiaries assigned to them by Medicare. All ACOs meeting quality targets at the end of a given year receive a share of any savings generated relative to a predetermined cost benchmark; and depending on the type of ACO, some incur a financial penalty if they exceed the benchmark.
According to CMS’ recent analyses, ACOs that take on higher financial risk are more successful in improving quality and reducing costs over time. So one important objective of CMS’ proposed changes is to increase the rate at which ACOs assume financial risk for their beneficiaries’ care.
In July 2009, the family of Massachusetts teenager Yarushka Rivera went to their local Walgreens to pick up Topomax, an anti-seizure drug that had been keeping her epilepsy in check for years. Rivera had insurance coverage through MassHealth, the state’s Medicaid insurance program for low-income children, and never ran into obstacles obtaining this life-saving medication. But in July of 2009, she turned 19, and when, shortly after her birthday, her family went to pick up the medicine, the pharmacist told them they’d either have to shell out $399.99 to purchase Topomax out-of-pocket or obtain a so-called “prior authorization” in order to have the prescription filled.
Prior authorizations, or PAs as they are often referred to, are bureaucratic hoops that insurance companies require doctors to jump through before pharmacists can fulfill prescriptions for certain drugs. Basically, they boil down to yet another risky cost-cutting measure created by insurance companies, in keeping with their tried-and-true penny-pinching logic: The more hurdles the insurance companies places between patients and their care, the more people who will give up along the way, and the better the insurers’ bottom line.
PAs have been a fixture of our health care system for a while, but the number of drugs that require one seems to be escalating exponentially. Insurance companies claim that PAs are fast and easy. They say pharmacists can electronically forward physicians the necessary paperwork with the click of a mouse, and that doctors shouldn’t need more than 10 minutes to complete the approval process.
In most other human activities there are two speeds, fast and slow. Usually, one dominates. Think firefighting versus bridge design. Healthcare spans from one extreme to the other. Think Code Blue versus diabetes care.
Primary Care was once a place where you treated things like earaches and unexplained weight loss in appointments of different length with documentation of different complexity. By doing both in the same clinic over the lifespan of patients, an aggregate picture of each patient was created and curated.
A patient with an earache used to be in and out in less than five minutes. That doesn’t happen anymore. Not that doctors and clinics wouldn’t love to work that way, but we are severely penalized for providing quick access and focused care for our well-established patients.
In May Philip Alston, the United Nation’s Special Rapporteur on Extreme Poverty, and John Norton Pomeroy Professor of Law at New York University Law School released his, “Report of the Special Rapporteur On Extreme Poverty and Human Rights on His Mission to the United States.” The 20-page report was based, in part, on Alston’s visits this past December to California, Georgia, Puerto Rico, West Virginia and Washington, D.C. After reading the report and the response to it, one is again forced to question how legitimate is our concern for the health and well being of the poor, or those disproportionately burdened with disease.
The UN report found over 40 million Americans live in poverty, or upwards of 14% of the population. Those living in extreme poverty number 18.5 million and 5.3 million live in 3rd World absolute poverty. Among other related statistics, Alston cites the fact the US has the highest comparable infant mortality rate, 50% higher than the OECD mean, due in part to an African American mortality rate that is 2.3 times higher than that of whites. The US has the highest youth poverty rate in the OECD. In 2016, 18% of children were living in poverty comprising 33% of all people in living poverty and 21% of those were homeless. These facts are explained in part by the report noting between 1995 and 2012 there was a 750% increase in the number of children of single mothers experiencing annual $2-a-day poverty. US poverty, the report explains is due in part to the continuing growth in income and wealth inequality. The report found in 2016 the top 1% possessed 39% of the nation’s wealth while the bottom 90% lost 25% of its share of wealth and income. Since 1980 annual income for the top 1% has risen 205% and for the top .1% by 636% while annual wages for the bottom 50% have stagnated. The report reminds us the US has approximately 5% of the world’s population but 25% of its billionaires. The US in sum ranks 18th out of 21 wealthy countries in labor markets, poverty rates, safety nets, wealth inequality and economic mobility.
As healthcare gradually tilts from volume to value, physicians and hospitals fear the instability of straddling “two canoes.” Value-based contracts demand very different business practices and clinical habits from those which maximize fee-for-service revenue, but with most income still anchored on volume, providers often cannot afford a wholesale pivot towards cost-conscious care. That financial pressure shapes investment and procurement budgets, creating a downstream version of the two-canoe problem for digital health products geared toward outcomes or efficiency. Value-based care is still the much smaller canoe, so buyers de-prioritize these tools, or expect slim returns on such investment. That, in turn, creates an odd disconnect. Frustrated clinicians struggle to implement new care models while wrestling with outdated technology and processes built to capture codes and boost fee-for-service revenue. Meanwhile, products focused on cost-effectiveness and quality face unexpectedly weak demand and protracted sales cycles. That can short-circuit further investment and ultimately slow the transition to value.
To skirt these shoals, most successful innovators have clustered around three primary strategies. Each aims to establish a foothold in a predominantly fee-for-service ecosystem, while building technology and services suited for value-based care, as the latter expands. A better understanding of these models – and how they address different payment incentives – could help clinicians shape implementation priorities within their organizations, and guide new ventures trying to craft a viable commercial strategy.
A mere two decades ago, the headlines were filled with stories about the “HMO backlash.” HMOs (which in the popular media meant most insurance companies) were the subject of cartoons, the butt of jokes by comedians, and the target of numerous critical stories in the media. They were even the bad guys in some movies and novels. Some defenders of the insurance industry claimed the cause of the backlash was the negative publicity and doctors whispering falsehoods about managed care into the ears of their patients. That was nonsense. The industry had itself to blame.
The primary cause of the backlash was the heavy-handed use of utilization review in all its forms –prior, concurrent, and retrospective. There were other irritants, including limitations on choice of doctor and hospital, the occasional killing or injuring of patients by forcing them to seek treatment from in-network hospitals, and attempts by insurance companies to get doctors not to tell patients about all available treatments. But utilization review was far and away the most visible irritant.
The insurance industry understood this and, in the early 2000s, with the encouragement of the health policy establishment, rolled out an ostensibly kinder and gentler version of managed care, a version I and a few others call Managed Care 2.0. What distinguished Managed Care 2.0 from Managed Care 1.0 was less reliance on utilization review and greater reliance on methods of controlling doctors and hospitals that patients and reporters couldn’t see. “Pay for performance” was the first of these methods out of the chute. By 2004 the phrase had become so ubiquitous in the health policy literature it had its own acronym – P4P. By the late 2000s, the invisible “accountable care organization” and “medical home” had replaced the HMO as the entities that were expected to achieve what HMOs had failed to achieve, and “value-based payment” had supplanted “managed care” as the managed care movement’s favorite label for MC 2.0.
As physicians and healthcare leaders, we are already well aware that the majority of patients do not have the information they need to make a medical decision or access to appropriate resources, so we didn’t need to hear more bad news. But that is precisely what new research once again told us this spring when a new study showed that almost half of the time, patients have no idea why they are referred to a GI specialist.
While the study probably speaks to many of the communications shortcomings we providers have, across the board our patients often don’t know what care they need, or how to find high-value care. Last year, my organization commissioned some original research that found that while a growing number of patients are turning to social websites such as YELP, Vitals, and Healthgrades to help them find a “high quality” specialist, the top-ranked physicians on these sites – including GI docs – are seldom the best when we look at real performance data. Only 2 percent of physicians who showed up as top 10 ranked on the favorite websites also showed up as top performers when examining actual quality metrics. (The results shouldn’t surprise you as bedside manner has little to no correlation with performance metrics such as readmission rates).
As providers and health care leaders, we lament that our patients are not better informed or more engaged and yet across the board, we have not given them the tools or resources they need to navigate our complex system. But now for some good news: all hope is not lost, and patients can become better consumers, albeit slowly, if we all do our part.
The focus on the CMS rules on information blocking continues on THCB. We’ve heard from Adrian Gropper & Deborah Peel at Patients Privacy Rights, and from e-Patient Dave at SPM and Michael Millenson. Now Adrian Gropper summarizes — and in an linked article –notates on the American Hospital Association’s somewhat opposite perspective–Matthew Holt
It’s “all hands on deck” for hospitals as CMS ponders the definition and remedies for 21st Century Cures Act information blocking.
This annotated excerpt from the recent public comments on CMS–1694–P, Medicare Program; Hospital Inpatient Prospective Payment Systems… analyzes the hospital strategy and exposes a campaign of FUD to derail HHS efforts toward a more patient-centered health records infrastructure.
Simply put, patient-directed health records sharing threatens the strategic manipulation of interoperability. When records are shared without patient consent under the HIPAA Treatment, Payment and Operations the hospital has almost total control.Continue reading…
A ‘single-payer’ plan is a target on the back of its supporters. But what about a ‘Medicare Public-Private Partnership’?
MOUNT VERNON — In February 2017, President Trump famously said: “Nobody knew health care could be so complicated.” Nobody other than about 99.9 percent of the almost 300 million people in the U.S. with insurance, that is. Yesterday, I received a copy of “Get to know your benefits,” the 236-page “booklet” for my new health plan. Like most people, I’ll never read the book, but its weight alone says “complicated.”
And it’s safe to guess that Trump also will never read his Federal Employee Health Plan information, even though one Aetna choice available to him has a “brochure” of only 184 pages. Thinking about the amount of information available to health insurance plan consumers, I began to wonder what Health and Human Services Secretary Alex Azar meant, also last February, when he said, “Americans need more choices in health insurance so they can find coverage that meets their needs.”
Presumably, were we to have more choices, we could study the hundreds of pages of information about each available plan and make better choices. According to the federal Office of Personnel Management, federal employees who live at 1600 Pennsylvania Ave., Washington, D.C. 20500, have a choice of 35 monthly plans. Too bad the president doesn’t live in Maine, where he’d have only 20 plans to study!