Even the most trivial of emergency-room trips can quickly add up. Going in for an upper respiratory infection averages more than $1,000. A urinary tract infection can set patients back thousands of dollars. But before Obamacare came on the scene, New Jersey physician Jeffrey Brenner was already working on innovative ways to slash health-care costs. He scoured health-care billing data at local hospitals and discovered that a small number of “super utilizers” clustered in certain geographic areas were responsible for the bulk of health-care costs in Camden, N.J. He brought together a team of social workers and medical professionals, who made regular house calls to those patients, accompanied them to doctor’s appointments and conducted long interviews with them to obtain health histories—all to help the city cut medical costs and provide better care to these neediest patients. That was some six years ago. His work, called health-care hot spotting, helped net him a MacArthur “genius” award in 2013.
Now he works full-time on this issue and oversees a team of about 20 nurses, social workers, community health-care workers, Americorps volunteers and a psychologist who attack this problem around Camden. More than 50 similar operations have popped up around the country, and Brenner assists half of them. The latest such health hot spotting project Brenner works with is Sutter Health, a huge system consisting of some 30 hospitals in northern California. Brenner, the executive director of the nonprofit Camden Coalition of Healthcare Providers, spoke with Scientific American about how to predict who will cost the health-care system the most, his plans for his “genius” prize winnings, and his latest efforts to study health hot spotting with a randomized controlled trial.
[An edited transcript of the conversation follows]:
What made you think to start mapping out “super utilizers” of health care?
I was a frontline family doctor in Camden, N.J., for 12 years. I accepted Medicaid patients and found that they had the most complex health problems to tease apart. In a typical primary-care model, we don’t serve those patients very well. It was a big, audacious, hairy problem where the tools we have been given are inadequate to solve it.
How can communities identify these complex, chronic patients—these so-called super utilizers that cost hospitals the most—while respecting patient privacy? Wouldn’t tapping such billing data run up against HIPAA protections?
It turns out that HIPAA allows you to work with large data sets for billing purposes, if you are improving quality or if it’s a valid research project. In our case, we originally got approval because it was a large research project. But we also have a business agreement as part of the health information exchange. That exchange under HIPAA says you are allowed to have data sharing agreements as long as patients are given forms to explain what their data is being used for. Not many patients opt out.
Your early hot-spotting efforts saved community hospitals millions of dollars, I’ve read. How much did you actually save?
We have no idea. Statistically, savings are actually really hard to calculate. I have not talked about dollar figures in the last few years because the only way we will know savings for sure is by doing a randomized controlled trial. That’s what we are doing now. We certainly believe our interventions save money.
Why is it hard to determine the savings?
There is a patient in Trenton, N.J., who went 450 times to the local hospitals in a single year. She was chronically homeless and alcoholic, and she had a lot of physical and sexual abuse in her history. Through a collaboration with the local hospitals and social agencies, she was able to get into a special housing unit and worked with a multidisciplinary team like ours that got her down to 18 visits a year.
We have a policy premised on why the intervention would make a difference, but it turns out that if you took 200 overutilizers like her and watch them over a year, they drop in utilization some 20 percent to 30 percent—even if you do nothing—because statistically, when you are dealing with outliers, outlier data tends to regress toward the mean. These people are quite sick, and it’s hard to get to the hospital 450 times each year. Our randomized controlled trial will get us some real answers.
Why are you doing a randomized controlled trial now?
There is a lot of research on pills and devices, but there has been very little high-quality research on how to deliver better care at lower cost. If you look at our funding for our nonprofit, there are about 28 sources of funding cobbled together to keep our team in the field and to keep the structure in place so we can do this randomized controlled trial. That’s why it’s taken so long to launch a trial. We’ve now partnered with the Abdul Latif Jameel Poverty Action Lab, which does randomized trials around the world on social interventions. The lead is up at the Massachusetts Institute of Technology \, and it’s the researcher that did the well-known Oregon health research that randomized people into access to Medicaid. They’ve been helping us set this up.
You were awarded a MacArthur “genius” grant in 2013. What did you do with the $625,000?
It’s an interesting grant. It’s not a grant to the Camden Coalition of Healthcare Providers organization. It was granted to me individually. I had a private Medicaid practice in Camden, and my payment rates kept getting cut. I actually went out of business. By the time I closed my office, I was getting $19 a visit because of cuts happening at the state level that were trickling down through the Medicaid HMOs.
The MacArthur grant is not one lump payment. It’s broke out over five years and about half of it is paid out in taxes. It comes as quarterly payment, and the first few years will go to paying off the debts from my practice.
How does that experience affect your current efforts with hot spotting?
Primary care is dying while hospitals are expanding, which underscores why reforms are needed. You get what you pay for. If you underspend on primary care, then you won’t get enough of it. We need to move some of that money spent in hospitals back to primary care providers and save the health-care system costs.
What does your randomized controlled trial look like?
It will have a total of 800 patients. Four hundred patients will receive our intervention, and 400 will be controls receiving normal routine care where they are discharged from the hospital and make their own appointments. We recruit patients into the trial from four hospitals in New Jersey where we have set up real-time data systems that allow us to know when these patients have been admitted: two admissions in six months signals to us that a patient may be the worst of the worst and that she or he is likely a $20,000 patient. We then explain our project to the patients and ask them to consent to participate. If they consent, we leave the room, hit the “random button” on our computer and the patient is randomized into intervention or control. We need to do that 800 times.
We then follow them in our data system. At the end of the study, we will also look through Medicaid records to make sure we catch if they received care elsewhere.
When do you expect all the data to be in?
We’ve been collecting good data now for six months. We have 80 patients in each arm of the study now, and so if we can ramp up and accelerate enrollment, then we’ll probably have data next December.
What do patients receiving the intervention get?
For 90 days we go to patients’ appointments with them, make home visits, and if they are homeless, we help them get housing. We also help them apply for other social services. It’s a multidisciplinary approach with social workers, community health-care workers and nurses, and we are also inside local primary health-care offices for training.
What happens after the 90 days of intervention?
We try to graduate them and plug them into a stable, well-run system of care. Sometimes, since many primary care providers have closed, we have trouble finding a practice that accepts Medicaid patients. We have been using some of our funds to augment Medicaid payments to primary care providers. We pay them $150 if they get one of these patients in for a visit five to seven days after when they were in the hospital. And we pay the patient with a $20 gift card and a cab voucher to go see the doctor. We have found in our data that the first week to two weeks after hospitalization is a critical time, and if we can engage them quickly, it makes a world of difference. And for those medical practices, it’s a lot of money. We’re giving out a couple thousand dollars to practices that are struggling.
Do you think this model of hot spotting is a good fit in both rural and urban areas?
Yes. We have worked with groups in Eureka, Calif., which is incredibly isolated, and found the same patterns hold up. We have also worked with a group in rural Maine, another in rural Michigan and also in rural Pennsylvania.
What we’re finding over and over with our partners across the country is that the number-one determinant of being a high utilizer of health care is the amount of adverse childhood experiences you had, like physical and sexual abuse. There is interesting literature to back that up. In short, those traumatic experiences in early childhood lead to lifelong health costs and can help predict health-care utilization rates.
Is it early life trauma specifically, or might other factors be at work there, such as socioeconomic status, economic and health access issues or childhood stability?
In a lot of studies we say that some bad outcome is due to socioeconomic status, but there has been very little work to look at the causality. There are higher levels of early life trauma in underserved communities; therefore, the true variable is probably the early life trauma and probably trauma and early life conditions. The social determinants of health and all the underlying pieces of it have not been fully explored, and we don’t understand the ethnography of it all.
Has the Affordable Care Act (Obamacare) impacted your work? The law sent a huge market signal out to the health-care industry that the game needs to change and become more efficient and accountable. Under the Affordable Care Act, there was also a $10 billion fund put together to support innovation over a decade. The Centers for Medicare and Medicaid Services have been putting grants out. We got a $2.7 million three-year innovation grant that is helping to pay for the research team in the field. It’s one of our 28 sources of funding.
Your approach has been likened to a “weather map” for health. Is that an appropriate analogy?
Hot spotting is not just making maps. It’s the strategic use of data to find outliers and to improve their care. Mapping is one example of how you segment data. There are other strategies you can use as well, like hospital claims data.
A lot of our work has been simplified down to terms like hot spotting and super utilizers, but it’s a multidimensional intervention. We are trying to get the cost curve to drop by focusing on the poorest patients. We are using data in real time to target outliers who are the canary in the coal mine to understand how the system is failing.