In Denver, Mining EHR Data for Public Health Monitoring, at a Very Local Level

Oct. 2, 2017
The Colorado Health Observation Regional Data Service (CHORDS) is a regional partnership between Colorado health providers, public health departments and the University of Colorado Denver to share health data in order to track population health trends.

Public health officials are finding that, in many cities, there are significant health disparities and health gaps among neighborhoods, with childhood obesity rates and tobacco use much higher in some neighborhoods compared to others, which can contribute to different life expectancy rates. In the Denver metropolitan area, a collaborative pilot project is leveraging electronic health record (EHR) data to monitor public health trends and evaluate interventions, and the data service is providing public health officials with data to identify and address health disparities.

The Colorado Health Observation Regional Data Service (CHORDS), based in Denver, is a regional partnership between Colorado health providers, public health departments and the University of Colorado Denver to share health data in order to track population health trends and develop effective interventions. CHORDS, which is a pilot project by Denver Public Health, a part of Denver Health and Hospital Authority, collects, analyzes and presents data from participating partner EHRs into one registry per topic that can be used to monitor population health and conduct research.

Currently, there are 12 partner organizations that contribute data to CHORDS, including Children’s Hospital Colorado, Kaiser Permanente of Colorado, Denver Health and the Colorado Alliance for Health Equity and Practice, and a number of federally qualified health center (FQHC) organizations. Technology partners for the project include the Denver Public Health Department and the Colorado Clinical Translational Sciences Institute.

Each registry collects and presents health information specific to its topic area. For example, to monitor tobacco use, CHORDS collects demographic characteristics and geographic information on the patient and their visit, in addition to whether or not the patient uses or is exposed to tobacco. This information is extracted and used to populate the tobacco registry.

Public health officials in the Denver area are tapping into data from CHORDS to see the prevalence of health issues, such as childhood obesity or tobacco use, in the region, and can even map the variance down to specific neighborhoods. Officials can then take the data to city council representatives and community forums to open up discussions about health inequalities and disparities, according to Art Davidson, M.D., director of informatics and epidemiology at Denver Public Health and CHORDS project director.

“Our goal,” Davidson says, “is to get politicians, policymakers, community-based organizations and community advocates interested in moving the agenda and moving the needle on public health issues.”

CHORDS, which refers to both the technology platform and a virtual organization of partners, currently has registries on obesity, tobacco use and exposure, diabetes, cardiovascular disease and depression. Examples of relevant data fields pulled from EHR data include, within the obesity registry, BMI, height and weight, and within diabetes, diabetes diagnosis codes and hemoglobin A1C lab test results. The 12 health entities in Colorado that are participating contribute data broken down by race, gender, age and location. Project leaders want to expand the data service to include registries on asthma, hepatitis C and congenital heart disease.

The use of EHR data for health monitoring is not unique, but applying these efforts at a local level is a relatively new concept, Davidson says. There are several national initiatives that have developed networks, including the Food and Drug Administration (FDA)-funded Sentinel Initiative and PCORI-funded PCORnet. Efforts to adapt these national models have been implemented locally through New York City’s Primary Care Information Project, and the MDPHnet program in Massachusetts.

According to Davidson, data sharing in the CHORDS health data network is powered by PopMedNet, a software application that enables the use of distributed data networks. The CHORDS instance of PopMedNet is hosted and supported by the University of Colorado Cancer Center’s Research Informatics Shared Resource with blended funding from National Institutes of Health (Colorado Clinical and Translational Sciences Institute) and a variety of grants from Colorado governmental and foundation funders.

Work to build CHORDS began about five years ago and project leaders first had to seek out healthcare provider partners to share their healthcare data and to make it available for public health agencies and researchers. The project leaders started with FQHC facilities and also reached out to other non-profit healthcare organizations. “It was somewhat challenging because we’re asking them to share health data for public use, so we had to make the case and build why this is important,” Davidson says. Each participating health care provider chooses which registry they want to participate in and contribute data to. As an example, a children's hospital would not provide data on adult indicators.

As CHORDS is a distributed database, data partners retain full control over their data and decide whether to answer a request for data. Each data-contributing partner, such as a hospital, stores data from their EHR in a virtual data warehouse (VDW). Each data-contributing partner requests and receives permission to download a PopMedNet “client” that connects their VDW datamart to the CHORDS network. CHORDS securely exchanges data using the client through a federated query, removing all personally identifiable information before data are shared.

Establishing robust, transparent governance policies and principles fostering data sharing was key to building CHORDS and continues to be critical to growing the health data network, Davidson says.

“You can have this technology down cold and you can fall on your face because you did not pay attention to the relational-issues and the need for a strong and transparent governance process that encourages and enforces trust,” he says.

CHORDS initially provided data to one local public health agency (Denver Public Health), but is expanding to include local public health users in the Denver metropolitan area, and other researchers across Colorado. The data service provides a way for public health officials and researchers to track population trends, across healthcare providers, and show the outcomes of policies and clinical- and community-based initiatives. Using data available in CHORDS, public health officials can monitor health indicators like community body mass index (BMI), community diagnosis and control of cholesterol and hypertension across populations.

As an example of how the information can be used, a public health agency can query for data on obesity rates and present to city council representatives in the form of a map showing obesity rates in a particular city council district and how it compares to other districts. “Public health officials can ask ‘Is there something you want to do as a policy maker to advocate for services in your neighborhood?’,” Davidson says. “And if you are the most obese neighborhood, they can look at, ‘Are you a food desert? Do you have bike lanes?’ Those are the kinds of social determinants of health that we try to bring out in this discussion. Community-based organizations can take that information and go to the politicians and policy makers and create a groundswell in the community with what they feel is evidence of health inequality and disparities.”

Davidson says researchers and public health officials are interested in the EHR data in the CHORDS health data network because EHR data can provide a level of accuracy, statistical power, and geographic detail unavailable through established health surveys, the traditional local public health information source, he says.

EHR data can answer many health questions that surveys or claims data typically cannot, he contends, pointing to the Centers for Disease Control and Prevention’s (CDC) Behavioral Risk Factor Surveillance System (BRFSS), a system of health-related telephone surveys that collect self-reported data about U.S. residents regarding their health-related risk behaviors, chronic health conditions, and use of preventive services. “If I want to know if you have hypertension, I can get that from the BRFSS, but if I want to know if your hypertension is controlled, I can’t get that from the health surveys, but I can get that from an actual measure of your blood pressure,” he says.

Davidson continues, “There is a lot of interest in the [EHR] data and how it can be used and how it could inform policy changes and system interventions. There are other people who look at it and are still skeptical, ‘Is it representative of the true population?’ This is still in its infancy, so we’re still learning about this. And that’s a major area of focus right now, doing some comparisons to the population-based surveys.”

In addition to contributing to public health monitoring efforts, Davidson says there are some other incentives for healthcare provider organizations to participate in the CHORDS project. According to Davidson, CHORDS population health monitoring is considered a specialized registry under Meaningful Use Stage 2, so eligible hospitals and providers in Denver County seeking MU Stage 2 incentive payments can fulfill the requirement to submit specialized registry data.