With the increased adoption of EHRs, there has been an interest in better analysis of the voluminous amounts of patient data generated by these systems. At Children’s National Health System in Washington, D.C., researchers have added a new wrinkle to this analysis through the use of geographic information systems (GIS) technology.
Compiled by Brian Albright, Contributing Writer
By combining geospatial and patient data, Children’s National Health System is developing new ways to improve outcomes and reduce readmissions.
With the increased adoption of EHRs, there has been an interest in better analysis of the voluminous amounts of patient data generated by these systems. At Children’s National Health System in Washington, D.C., researchers have added a new wrinkle to this analysis through the use of geographic information systems (GIS) technology.
Children’s National Health System is one of the top pediatric hospitals in the country and operates a community-based pediatric network, seven regional outpatient centers, an ambulatory surgery center, two emergency rooms, an acute care hospital, and works with a network of 240 independent physicians.
Brian Jacobs, M.D., CMIO and CIO at Children’s National, has helped spearhead the health system’s GIS initiative which gives physicians a unique geographic view of patient population trends. Using this technology, Children’s can generate a map showing which zip codes and neighborhoods have the highest incidence of diabetes, childhood obesity, or other conditions. They can then use that data to craft targeted programs to improve outcomes. In this Q&A, Dr. Jacobs details the origins, applications, and resulting outcomes of this powerful technology integration.
Q. What prompted Children’s National to incorporate geospatial data into population health management efforts?
A: Our geospatial work goes back many years to a time when we had many children coming into the emergency room with thermal burns from hot water heaters turned up too high. One of our trauma surgeons thought this through and noticed these patients were coming from certain parts of the city. He thought it would be valuable to use some geospatial tools to map the origins of the patients using their zip codes and, lo and behold, they were coming from two areas in the D.C. Metro region. By rolling out an education program targeted at these specific neighborhoods, we were able to largely eliminate the problem.
After that, we started to use our Esri ArcGIS geospatial analytic software in a more aggressive manner to look at a lot of different healthcare conditions. We’ve looked at sickle cell disease with pain crises, patients who are readmitted to the hospitals, a disorder called bronchiolitis, asthma, and immunization practices. We’ve used the geospatial tools to look at other environmental conditions that may have a relationship with these patients by mapping patient location data and correlating that with the incidence of specific conditions. We can overlay other types of demographics on those maps to see what other factors may be involved.
The area we’ve done the most work in is childhood obesity. We’ve looked at patient addresses in relationship to other environmental factors such as socioeconomic status, recreational opportunities in the area, and the number of fast food restaurants that potentially contribute to that particular health condition. This has provided us with more detailed insight into the situation and better prepares our staff to customize a preventative intervention plan.
Q. Why did you select Esri as your GIS data partner? And how does ArcGIS integrate with your EHR and other HIS platforms?
A: Esri offers a lot of functionality, such as the ability to make different map layers and lay one map over another. You can see the spatial relationship between the different map layers and conditions of interest. That’s not readily available in other platforms.
When you look at the EHRs, you now have a digitized environment where all of that health data is easily extractible. You can look at things in much greater detail than before. For the typical patient, there may be 40,000 essential health data elements that are available for aggregation and analysis. Things you would never get in census data or publicly available environmental data. You combine that health record data with the publicly available data to create these layers.
We started with a database extraction and a cleaning algorithm, and a de-identification algorithm. We ran the data through a process that allowed us to extract, clean, and de-identify it — a process that will be more centralized going forward. We’ll extract and normalize the data, place it into an enterprisewide data warehouse, and apply the geospatial tools. We’re building the data warehouse with Cerner which will be the single source of truth for the region.
It will be data-agnostic. Our inpatient emergency department and OR are on Cerner, and we’re rolling that out to our specialty physicians and clinics. In primary care, we have an inner city group of providers on eClinicalWorks, and a suburban set of practices on Greenway.
Q. How have your population health management efforts benefitted, and what specific actions did you take based on the added insight GIS provides?
A: We’ve used the childhood obesity data in a number of ways. First, we’ve better characterized the obesity problem in the metropolitan region, and we use that information in lectures and public forums. That forms a substrate of data for grant applications or lobbying efforts. It also helps define the condition and the underlying problems.
If I’m sending home a patient with sickle cell disease, and I know they are at high risk for readmission, we alert the discharging team. Once you know that risk stratification, they may be eligible for care coordination or a home patient navigator, or more aggressive followup calls to ensure they are following the medication regime, going to appointments, or following diet and exercise recommendations.
Q. What benefits has Children’s and its patient population realized as a result of this GIS effort?
A: With sickle cell disease, we have a lot of children admitted with sickle cell pain crises. One thing we’ve been able to do using that data is multivariant analysis. We’ve come up with an algorithm that predicts the likelihood of being readmitted to the hospital in 30 days, with an accuracy around 80 percent by looking at where they live and the types of environmental or economic conditions in those areas. We can apply different resources to avoid readmission.
We haven’t seen a decline in readmissions yet based on these other efforts, but that’s because we haven’t applied the resources to readmission prevention that we would like to. That’s something we plan to do over the next year. We have a great understanding of the issues, but tackling those issues with the right resources is a challenge.
Q. What advice would you give providers considering GIS to enhance their PHM (population health management) efforts?
A: It all starts with having the cleanest data possible. EHR data is not like banking data in that the data is not entered consistently — you have to clean it. Beyond the primary data, you need to have a data quality individual or team that goes through the data and says, “We’re going to accept 80 percent of this data because it’s clean, and we’re going to omit the data that is suspect.” When you have a population of thousands of patients, losing a few records here or there is not a big deal. You have to make sure it doesn’t exceed a certain percentage or you lose confidence in the data. The data also needs to be converted to common units, something that can pose up-front challenges.
When you do this work, it’s important to have the right technical infrastructure and server-based applications that allow you to process large volumes of data in a timely manner. We’ve made the move to the ArcGIS server environment and have the appropriate backups and disaster recovery to ensure fast, reliable processing of data.
One other thing I would add is a lot of the data you would want to map in a geospatial platform is your own or publicly available, but there is also valuable data from proprietary sources. You need to have the resources to purchase that data if you need it.
Q. What plans do you have for leveraging geospatial data in your PHM efforts?
A: We’re integrating an analytics platform into the data warehouse, so we can integrate it into the clinical workflow. That will allow clinicians to do work around risk stratification, which is important.
From a research point of view, we have folks doing things from molecular biology to public health research, and this data is valuable to them too. It’s important that they have access, but we have to be aware of privacy. We supply data to researchers in a private manner so they can come to a conclusion and look at opportunities for public health improvements without actually causing any privacy violations for the patients who may be involved.