We’ve all seen cool maps of health data, such as these representations of diabetes prevalence by US county. But few people think about how thoroughly geospacial data is transforming public health and changing the allocation of resources at individual hospitals. I got a peek into this world at the Esri Health GIS Conference this week in Cambridge, Mass.
Bill Davenhall, Senior Health Advisor for Industry Solutions at Esri, described GIS as a glue for other types of data. I heard at the conference of an example in Louisiana, where epidemiological issues such as babies with low birth weights are correlated with location. This is an important issue for public health policy-makers because such babies cost Medicaid four times as much as babies with normal weights.
Ryan Bilbo, GIS manager at the Louisiana Department of Health, told me that the department didn’t just group incidences of low birth weight by an arbitary geographic marker, such as zip code or county. That would obscure big differences within each area–no matter how small you made the areas. So they let the Esri software find clusters among locations that were near each other and generate a amp based directly on the data. Typically, a cluster would be found in a place that was significant demographically, such as a housing project.
Note that this fine-grained targetting requires addresses, which are personally identifiable information (PII). Therefore, the health department can’t release the data to others for privacy reasons. They could fudge the addresses in order to create a deidentified data set, but it would be much less valuable. (See the O’Reilly book Anonymizing Health Data for a discussion of de-identification and data value.)
Privacy is only one reason data is hard to get. One speaker said his facility obtained Medicare data from the Center for Medicare & Medicaid Services (CMS), but that the researchers had to sign an agreement to use the data only for specific purposes. And Medicare data is the best available in this country, he said–although other data sets are coming up to speed.
Davenhall says that the public as well as individual health and social care providers can’t get information about clusters of low birth weight babies until long after the data has been electronically submitted (sometimes as long as 18 months after it is collected). Although the hospital staff enter the data right after birth, the data has to traverse a twisty path to the county and then to the state and then to a national center for statistical purposes. Some useful metadata is added along the way, but if a problem is developing in some area, pediatricians and policy-makers would sure like to hear about it before 18 months have passed. He thinks this is a public health information workflow that needs serious attention to improve its timely access and usefulness within the community.
What is the data good for?
In general, I saw three benefits from big data in health care:
- It can turn up issues. In Louisiana, for instance, plotting the instances of certain diseases produced a pattern over a particular waterway that they deduced to be contaminated.
- It can identify what’s working well. If certain regions have a lower rate of unnecessary X-ray tests than others, you can investigate what the good regions are doing and try to get the practice adopted elsewhere.
- You can allocate scarce resources more effectively. This criterion bothers me a bit because, idealistically, you would gave everybody the attention they need for perfect health. Given limitations in what we can spend, we’ll help more people by looking for those clusters of low birth weights.
I saw a presentation about a VA hospital showing how GIS data can be applied at the level of a single facility. In their case, they place RFIDs on all important equipment and made an inventory of everything in the hospital rooms, right down to the railings next to the toilets. (If someone reports a broken railing, they want to find and fix it.) They can also use the data for continuous quality improvement, such as tracking where the most slips and falls happen.
One wouldn’t perhaps expect a child welfare agency to do market segmentation and seek our information on how to reach the public more effectively, as businesses do. But GIS analyst Kathryn Kulbicki helps agencies do just that. Finding foster parents and adoptive parents is an urgent matter for them. So the agencies feed in information about the current foster and adoptive parents they know in each area, and look for ways to attract people with similar characteristics. In one case they learned that these families liked a certain restaurant chain, and persuaded that chain to do a promotional campaign for the agency.
GIS data has a lot more to offer child welfare agencies as well:
- To minimize disruption in the child’s life, the agency looks for foster parents near the birth parents, or at least try to keep the child in the same school district. They also try to place siblings near each other.
- Before sending a child to a new location, they look for a range of services that are close by, ranging from medical services to Head Start programs. This data is available from case workers, but needs to be updated as services come and go. (Of course, the same data could be approached from different angles. Head Start could conceivably use it to open new programs where they’re needed.)
- In case of disaster, the data helps them identify families who need help.
A global vision
Everything at the Esri conference was pretty impressive, but for an awe-inspiring take, nobody could beat Bill Davenhall. We talked about the promise of combining genetic data, personally collected data, conventional doctors’ records, and geographic environment data.
One of Davenhall’s favorite public information sites is the EPA’s Toxic release inventory. It has been collecting data on toxic waste since 1987, and Davenhall is mining it to produce a list of the 100 most carcinogenic sites in the US. When he goes to the doctor, “They never ask me about my place history,” he points out in his TED talk. But in genetic databases, 35% to 50% of the cause for disease is listed as “unknown” and therefore “environmental.” Esri has released an iTunes app called My Place History that individuals can use to match EPA data to their locations.
Several universities already offer courses on geomedicine in their medical schools, but Davenhall looks forward to the day when doctors can integrate the information into their practices. It’s a fine example of the merger we need to see between personal with public health, to improve the quality of life for all of us.