Researchers Uncover Potential Opioid Abuse Hot Spots with Network Scan Statistics - Methodology May Ultimately Enable Early Interventions in High-Risk Areas
According to the Centers for Disease Control and Prevention (CDC), more than 63,600 people died from opioid overdoses in the United States in 2016, and more than 70,000 people died from the same cause in 2017. The opioid epidemic in the United States is showing no signs of abating, but researchers from the Biocomplexity Institute have identified potential geographic opioid misuse and abuse hotspots in Virginia, West Virginia, and North Carolina using network scan statistics – a methodology that may ultimately be used to save lives through targeted interventions in high-risk areas.
The research study, "Detection of Spatiotemporal Prescription Opioid Hot Spots With Network Scan Statistics: Multistate Analysis," was led by Biocomplexity Institute Professors Achla Marathe and Anil Vullikanti, who work in the Institute’s Network Systems Science and Advanced Computing division. Their research was published in the June 17, 2019, issue of JMIR Health and Surveillance. The researchers examined spatiotemporal hot spots of opioid users and prescription claims through a combination of county-level demographics data from the United States Census and Medicare claims data, and applied network scan statistics to detect significant anomalous spatial clusters of users and opioid prescription claims.
"The application of this methodology and the fusion of data sets is fairly novel, and can be used for surveillance of new and emerging hot spots," Vullikanti said. "Beyond the identification of spatiotemporal clusters that are likely to encounter prescription opioid misuse and overdose, our research also provides characteristics of the counties and providers’ specialties that are at higher odds of being anomalous in terms of the number of opioid beneficiaries and prescriptions."
Among other findings, the research identified some common patterns across all three states:
- The counties in top clusters have similar demographic characteristics such as:
- Racial features: Predominantly white, except in North Carolina
- Income and employment levels: Lower than statewide levels
- Number of housing units: Significantly lower than statewide levels
- Healthcare coverage: Generally, more people have public insurance, including Medicare and Medicaid; fewer have private insurance; and, fewer people have employee paid plans.
- Healthcare providers with an unusually high number of opioid beneficiaries and claims include specialties such as physician’s assistant and nurse practitioner.
- The same counties appeared in the most anomalous clusters in 2014 and 2015, but they were far fewer than in 2013.
"Considering the severity of the opioid epidemic in the United States, we plan to work with both the UVA School of Medicine and the Virginia Department of Health to assess the practical applicability of our research and unique methodology," Marathe said. "We’re hopeful the data could be applied to deliver targeted early interventions in the regions that are anomalous or likely to become anomalous over time."
While the Biocomplexity Institute team focused this study on Virginia, West Virginia, and North Carolina, the methodology is generic and can be applied in other contexts for anomaly detection. Its application to spatial epidemiology can help detect unimmunized or under-immunized clusters of individuals in an otherwise well-vaccinated population. Other possible application areas include detection of potential hot spots for suicide, teenage pregnancies, and criminal activity.
The full research study is available for download here.