MITRE Corporation to showcase SMART Scatter analytics at HIMSS19
There are lots of ways to analyze data for better quality outcomes and improved pop health – and there are also many types of data that can be analyzed.
At HIMSS19 next week, Kevin Gormley, principal data scientist at The MITRE Corporation, along with his colleague Dawn Heisey-Grove, principal epidemiologist at MITRE, will show how tapping into multiple sources of data from specific communities can help detect at-risk patient populations and target needed interventions.
By probing risk factor estimates and heat maps of higher-risk geographic locations, providers and public health officials can detect certain health patterns that might have been previously overlooked. But putting that data to work requires that certain hurdles be overcome – especially with regard to access and aggregation of vastly different and often very siloed data types: state or county information, census block group, postal code, etc.
A new open source tool aims to help address some of the biggest challenges for communities and health organization hoping to leverage multiple data sources at once. Called Simulated Multivariate Adaptive Regression Technique, or SMART Scatter, the technology allows easier access to local administrative data for the development of pop health models.
In Orlando, Gormley and Heisey-Grove will discuss how various different datasets can inform public health policy, surveillance and intervention; explain how partnerships between local government and the the private sector can boost the success of those efforts, and show how heat maps can be leveraged to improved population health health management.
In advance of HIMSS19, Gormley answered a few questions about SMART Scatter tool for Healthcare IT News.
Q. What is SMART Scatter and how does it work?
A. SMART Scatter integrates data from different sources stored at different geographic levels (e.g., county, ZIP code, census block group, and household). Before SMART Scatter, analyses could only be conducted with data at the same geographic level. SMART Scatter increases the amount of data that can be used in a single analysis thus enabling estimates that help public health decision makers better evaluate the implications of certain policies. Our tool captures valuable information from each data source (e.g., correlations between variables, distributions of variables such as income) and uses all the available data sources for each household (its respective census block group, ZIP code, county, etc.).
Q. What are some of the ways it's helping improve public health, surveillance and intervention? What challenges does it solve?
A. Our community partner confirmed that the heat map of domestic violence hot spots fit their public health experience. Knowing that the model is accurate, we are shifting this year to building a public health policy simulation infrastructure that evaluates public health intervention policies applied to community or communities. Based on their experience and expertise, public health staff could use the model to simulate the effects of interventions they predict will be most impactful within targeted portions of a community (e.g., school districts). Our tool will estimate impacts for different policy interventions, enabling public health staff to confirm their strategies and consider refinements.
Q. Are some other specific use cases that could be promising besides the ones that will discussed at HIMSS19?
A. Our current model is based on incidents of domestic violence, but other conditions could be simulated similarly if their risks can be modeled at the geographic level. For example, residents close to wooded areas may be at greater risk of tick-borne illnesses, people living near sources of air pollution (factories, highways, etc.) may be at higher risk of asthma, or those who live in "food deserts" may be more susceptible to obesity.
Q. What are the specific geographic data that are most useful here?
A. What is most useful is combining multiple local data sources that a county or community already have access to with Census data, which are publicly available.
Q. Beyond public health and safety, do you feel that this is a tool that has great promise for larger population health management challenges, esp. as health systems grapple with value-based reimbursement?
A. SMART Scatter has the potential to estimate the social determinants of health and other risk factors for communities. As healthcare organizations consider how to control costs while improving care, there are many aspects related to social determinants of health and access to care that could be modeled with SMART Scatter. Access to transportation, for example, is an ongoing challenge for patients travelling to appointments. In the future, perhaps SMART Scatter could be coupled with patient rosters to estimate how public transportation could be modified or supplemented to improve access to care, or to better understand community needs around transportation. Another example is among clinicians who are treating patients with obesity; they may wish to better understand the community assets around the patient’s home or place of employment, and SMART Scatter, with its ability to leverage data from multiple sources, could be a tool to facilitate that view.
Q. Any other thoughts for those interested in learning more?
A. SMART Scatter is representative of the type of innovation that can be realized from research collaborations – in this case between county public health staff, university researchers (Virginia Tech), and MITRE. We are careful to always recognize these indicators do not suggest that a single individual is at greater risk. Rather, these SMART Scatter estimates show risk potentials within small populations, possibly within a few blocks or a ZIP code.
Kevin Gormley and Dawn Heisey-Grove's presentation, "Leveraging Local Data for Public Health with “SMART Scatter" is scheduled for Wednesday, February 13, from 11:30 a.m. to 12:30 p.m. in room W206A.