Increasingly, the public health sector is tapping into the same big data and analytics frameworks used by commercial and clinical organizations and applying them to address a multitude of community health challenges. A recent All In project showcase webinar shared two examples of collaborations led by researchers and public health departments that are at the forefront of the movement to leverage big data to drive community health improvement.
Allegheny County Health Department and Vanderbilt University Department of Health Policy are each actively working to leverage multiple big data streams and apply advanced analytics to improve health—specifically, to address cardiovascular disease and infant mortality, respectively. Both communities learned valuable lessons in the process and shared advice on finding the right balance between using complex algorithms and more modest methods for identifying at-risk individuals and prioritizing interventions. Below are key takeaways from their presentations.
1. Keep it simple
Dr. Melinda J. Beeuwkes Buntin of Vanderbilt University’s Department of Health worked with partners to develop a data-sharing network including various sectors to enable real-time identification of at-risk pregnant women and referral to appropriate interventions to reduce infant mortality.
The project updated the Welcome Baby algorithm, which was already an innovative and sophisticated model for understanding the predictors for infant mortality. Although altering the model did improve their prediction of infant death, Dr. Buntin noted that even without adding complex modeling techniques, the original vital record data was still very powerful for identifying at-risk families. She added:
“No community should be deterred or concerned because they don’t have these tools or techniques at their fingertips.”
2. Be flexible and open to changing course
Dr. Karen Hacker, director of the Allegheny County Health Department, helped bring multiple sectors together to form a connected data warehouse that combines data and exports it to a modeling platform, FRED, to develop a geographically accurate model of the complex distribution of cardiovascular disease risk factors in the county.
When the project began, very little was published in the literature about how social determinants of health effected risk for cardiovascular disease. The project team hoped that the data would illuminate a specific social determinant or intervention that would lead to a decrease in cardiovascular disease, but Hacker noted:
“It began to appear that no one single social determinant was going to have the kind of impact that all of them together would have. They are all so connected to each other and so correlated with one another that we realized we needed to look at addressing a multiplicity of social determinants to see the needle move.”
Although the results weren’t as expected, the health department plans to continue to use the data warehouse to examine variety of other issues moving forward, from asthma to opioid overdoses. They are continuing to work with partners to sustain the rich data they have aggregated and refine the synthetic model.
3. Make sure you have the capacity to act on the results
Dr. Buntin explained that even if the algorithm adds greater precision for identifying at-risk families, the ability to act on that information is constrained by the capacity of the program to deliver interventions of the correct intensity. For example, The Welcome Baby algorithm divides births into risk categories (high, medium, and low), with higher-risk groups receiving a more intensive intervention (eg., a home visit versus a phone call). If the model were to change the cutoff so that more families were considered high-risk, it would be critical to ensure the resources are in place to deal with the added cost and staff time.
4. Leverage the data to attract new investments
Having a sophisticated analytical model that targets the highest risk cases can help demonstrate that resources are well-directed and make the case for maintaining investment in a project. Dr. Hacker explained that the social determinants of health data collected in Allegheny County has helped encourage managed care organizations and other partners to shift their investments into the high-need communities illuminated by the data. Dr. Buntin also stated that the results from the infant mortality analyses have been helpful in raising public and private funds to support their ongoing work to develop targeted interventions for pregnant women and new moms.
Learn more
To learn more about projects that share data across sectors to improve health, sign up for the All In newsletter.