Almost daily, it seems, we are seeing new examples of ways that aggregated EHR data can add value to healthcare quality initiatives. Just yesterday I wrote a news brief about a fascinating effort at UC San Francisco to use EHR data to track down a source of a common hospital-acquired infection by tracing the movements of more than 85,000 patients over a three-year period.
Last week, I saw a great presentation at the NIH Collaboratory by Michael Klompas, M.D., M.P.H, a professor of population medicine at Harvard University, about his team’s work to use EHR data to automate reporting to public health agencies on notifiable diseases and chronic conditions.
In describing the need for such a system, Klompas noted that there are some limitations involving the tools that public health agencies have traditionally relied on to gain insight into chronic disease prevalence, the Behavioral Risk Factor Surveillance System (BRFSS) and the National Health and Nutrition Examination Survey (NHANES).
“The BRFSS relies on patient self-report and in an area of shrinking budgets, many states are finding they can’t fund really large samples of patients,” he said. “In Massachusetts, for example, we used to do telephone interviews on 20,000 people, but we are now down to about 5,000.” The smaller a sample size gets, the less capacity you have to be able to detect rare conditions or to describe disease prevalence in rare populations or in small communities.
NHANES includes detailed, in-person examinations. It is clinically very rich and detailed information, Klompas said, but limited to only 10,000-person sample size for the entire country. With both NHANES and BRFSS, there is a substantial delay between data acquisition and when the data is available to be shared. “Often that delay can stretch to years,” he said. “And of course if you are trying to do a public health prevention program, and trying to get real-time feedback on whether it is working or not, a lag of two to three years is not helpful.”
To address some of these limitations, the Therapeutics Research and Infectious Disease Epidemiology (TIDE) group in the Department of Population Medicine at Harvard Medical School and Harvard Pilgrim Health Care Institute worked with the Massachusetts Department of Public Health to develop the. Electronic medical record Support for Public health (ESP) platform.
ESP enables medical practices and hospitals to automatically track notifiable diseases and chronic conditions using their electronic health record data to support public health reporting, population management, and community needs assessments. It uses algorithms to detect infectious diseases, chronic conditions, and other events of public health interest (e.g. opioid prescribing, influenza vaccination, hepatitis C screening) using vital signs, diagnosis codes, laboratory results, social history, vaccine, and prescription data.
“Rather than relying on diagnosis codes, it uses the totality of information to find the things we want,” Klompas explained. “That can be more sensitive and specific.”
Klompas described how it works: The clinicians interact with the EHR entering information about patients. Every night, ESP does an extract of structured data from the last 24 hours, which is ported to an ESP server that resides behind the organizational firewall. Algorithms identify data of public health interest and can generate case reports to be sent to public health agencies. Chronic disease data can be aggregated as de-identified summaries.
In Massachusetts four organizations — Atrius Health, Cambridge Health Alliance, Planned Parenthood of Massachusetts, and Fenway Health — are now sending automated reports of notifiable diseases via ESP to MAVEN, the Massachusetts Department of Public Health's (MDPH) web-based integrated surveillance and case management system. Longitudinal reporting for chronic HCV, HIV, and TB is currently in development, enabling MDPH to monitor the continuum of care of individuals with these infections. ESP is also being used by MetroHealth in Cleveland and by Tarrant County, Texas.
Massachusetts has also implemented aggregate-level querying and reporting capabilities (MDPHnet) at three clinical partners, Atrius Health, Cambridge Health Alliance, and the Massachusetts League of Community Health Centers. Those three collectively provide care for more than 20 percent of the state’s population.
According to the ESP web site, the MDPHnet is being used for several purposes, including:
• Identifying patients with high dose opioid prescriptions for targeted management;
• Providing data for Community Health Needs Assessments; and
• Evaluating the impact of public health programs to address chronic conditions in targeted communities.
The ESP team has created a web-based data visualization tool called RiskScape for the Massachusetts Department of Public Health using data from MDPHnet partners. Users can explore health conditions of interest by variables such as gender, age, race, city/town, and other co-morbidities.
Klompas did a live demonstration of RiskScape, looking at Type 2 diabetes by zip code, and showing how researchers or public health officials could study demographics and co-morbidities of patients by zip code. Rather than being two years old, the diagnosis information is almost real time. Users can also use it to track trends over time.
Most of the heavy lifting in getting started is the initial installation and building a partnership with public health. Or ESP could be used strictly for internal use for population health management, Klompas said.
The ESP source code is open source and downloadable, he added. The underlying software is popmednet, the same software used in PCORnet, the Patient Centered Outcome Research Network. A health system’s IT department has to set up an extract from the EHR to the ESP system. There is mapping work to be done to identify labs and tests and validate disease detection.
He estimated the initial cost at approximately $75,000, with perhaps $10,000 per year for maintenance.
The main point he stressed is that automated electronic reporting improves the timeliness and completeness of case reports while reducing reporting burden on providers. It also allows public health researchers to get a real-time window into the impact of interventions.
