Posts Tagged ‘geocoding’

Healthcare Hotspotting: Camden Coalition Drives Interventions with Data

July 11th, 2013 by Patricia Donovan

Look out, big data: a coalition of providers in one of the most dangerous cities in the country is tapping widely available hospital claims data to defragment the healthcare system and close gaps in the care continuum.

The process is called “hotspotting,” explains Ken Gross, director of research and evaluation for the Camden Coalition of Healthcare Providers. In a recent interview with the Healthcare Intelligence Network, Gross described the coalition’s plans to evangelize the use of hotspotting to develop key care interventions via its Healthcare Hottspotting Toolkit. In the first of two blog posts on hotspotting, Gross provided a primer on healthcare hotspotting.

HIN: What is the Camden Coalition of Healthcare Providers?

Ken Gross: It’s a membership organization made up of representatives from three Camden hospitals. The city of Camden is nine square miles. It’s one of the most dangerous cities in the country, and it has one of the lowest average income per person in the country as well. Members come from a number of organizations, including hospitals, clinics and private primary care providers.

We are the spoke in a fragmented healthcare system; we bring everyone together to see what kind of programs are needed to fill in the gaps — between hospitals and clinics, between social services and mental or behavioral health. We are a nonprofit and we’ve been in existence for about 10 years now.

HIN: There’s a lot of emphasis today on drilling down to local healthcare data in order to identify high-risk patients and better coordinate their care. Where does hotspotting fit into this trend?

Ken Gross: We believe health interventions should be data driven. But first, you need to quantify the problem and understand the population health issues — whether that population is in the geography of Camden or the population of patients coming to a particular hospital.

To be data driven, you can wait for the data to come down from the state or government level, but it’s rarely released in small enough geographies to look at the health needs of the community in order to plan. Planning means looking at how many utilizers do you have? How many high utilizers are ED-only high utilizers? How many are inpatient high utilizers? What are the top diagnoses and how much did the most expensive patient in aggregate cost the system?

To do this, you need local solutions, local data. Most organizations and technology focus on the electronic medical record (EMR). Well, there are a lot of EMR companies out there, and many different data formats. It’s hard to aggregate reporting from EMRs.

But there’s something sitting off to the side that is a lot easier to make use of, and that’s claims data. People ignore claims data as a means of understanding the population problem because they think it’s to get bills paid and that’s it. But claims data is useful for both clinical and economic purposes. It includes basic demographics: the patient’s age, insurer, top 20 diagnoses, procedure codes, charges and receipts. That data can help.

It’s also more readily available. It’s easy to activate to quantify the problem and start to plan programs that are data driven to say, “We have x amount. The problem is this big.” Every hospital keeps that data in the same format, so you don’t run into that problem of different formats across different hospitals like you do with EMRs. It’s powerful information and it’s easier to get started than any other data source in healthcare.

HIN: Hotspotting advocates the use of hospital claims to pinpoint heavy users. Don’t you also need payor and other provider data to get the total picture?

Ken Gross: We would love to have more data on outpatient visits and primary care. It’s harder to get all that data and link it together. It is correct that with hospital data, we will get only a part of the picture, such as ED and inpatient visits. Often, that’s pretty powerful to paint the picture for a community. If a lot of people are going to the ED for say, asthma-related conditions, that’s a proxy for the health of the community and barriers to the primary care system. It speaks to utilization and why people are going to the ED: that there aren’t enough primary care physicians or the hours aren’t good, or there’s a behavior change needed to educate people and link them to primary care.

Another value of getting claims data from the hospitals themselves is that it gives you an idea of the level of charity care, which is a proxy for the uninsured. Payor data would miss that, because there isn’t a payor in these cases.

What’s unique about getting it from the hospital is that we can ‘hot spot’ to see geographically where people are receiving charity care and do outreach efforts to enroll them in ACA coverage. You wouldn’t get that information from any other data source.

HIN: What are some examples of hotspotters?

Ken Gross: We look by neighborhood or particular address. In our community, we saw there were high-rises and assisted living nursing homes that had lots of utilization. We didn’t know exactly what to do in those areas, but it narrowed things down for separate discussions. It led to discussions with the people who managed one high-rise: what can we do in that building to provide better access to care? And a clinic opened up.

In other hotspotting examples, we focus on what claims data shows from utilization patterns: how many ED visits and inpatient visits come from a particular location? Hotspotting could also identify people with the primary diagnosis of diabetes in certain geographic areas. You would then look at whether you have enough resources. We have diabetes education programs as part of our coalition.

We can then overlay this on a map: where there’s a high instance of diabetes in the community and where we offer our programs, to see if we need programs in other places.

Hotspotting isn’t just geographic. It really is segmentation of the population to get a better understanding of different utilization patterns. And then the next step as we’re doing that is spatial, but it doesn’t have to be just spatial.

HIN: How do you create a map, and what you would do with that map?

Ken Gross: The first step is allowing for the hospital to provide address level information from claims data. Often the data needs to be cleaned up a bit to make use of it. There are a number of software packages that allow you to do “geocoding:” — taking those addresses and putting them as dots on the map. Once that dot is on the map, you have all the underlying data, so you can query the dots that are just diabetes or just ED high utilizers. And then we aggregate it to specific buildings.

There are ways to display the data in aggregate in small geographies. The census block group is a geography we often use. Then we can make maps to identify different utilization patterns by neighborhood.

It’s important to do it both at the address level and the neighborhood level. For example, if there’s a high rise within a block group with high utilization, high variance, and we only looked at it at the neighborhood level, you would think this is a neighborhood problem. But then if you notice they’re all coming from one address, you realize it’s a building-related problem to address.

There’s two levels — mapping point level or address level and the aggregate by the address.

HIN: If collecting local data is not that difficult, why have organizations been slow to do this?

Ken Gross: There are three common reasons. First, no one is looking in that direction because they don’t see the value of the claims data. They forget that there’s useful information there.

Second, when people do see it there, they have legal concerns. They think, I can’t share that with anyone because of HIPAA. We actually recommend and will have on our toolkit site information about our business associates agreement, which allows the sharing of data in specific circumstances.

The third barrier is the cost data in the claims. In a competitive hospital environment, they don’t want that data out there. There’s concern that one hospital is doing this many procedures or seeing this many would be market or business intelligence versus competitive market. In agreements with our three hospitals, we agree we’re never going to report that Hospital A is doing this, or has this many, Hospital B is doing that compared to C. We say the only way to report on the health of the population is to get aggregate data from all three — for example, the residents of Camden go to these three hospitals, and we’ve seen these aggregate number of visits, this aggregate number of costs.

That seems to ease some concerns. Start with the ‘ why’ and get everyone on board to understand why we want this data, then start the legal conversation and the contract discussions.

HIN: We’re starting to see some of this cost information, with CMS’s recent release of cost data by facility for Medicare.

Ken Gross: That was a great first step. And that was claims data. For CMS to have taken that step shows transparency. And there’s been lots of discussions since about what to do about that data.

Similarly, hotspotting creates a context for people to start talking about either trends they didn’t know about or knew anecdotally, but didn’t know the total sum of the problem.

Editor’s note: In a future post, Gross describes the coalition’s Hotspotting Toolkit, developed with a grant from the Commonwealth Foundation.