The Governor’s Opioid Addiction Crisis Datathon is a groundbreaking competition that will bring together multi-discipline teams comprised of individuals from government, higher education, private industry, and non-profits together to take new and existing datasets and turn it into actionable information that will support the Governor’s goal of using data and analytics to stem the tide of the opioid crisis, reduce addiction harm, and save lives.
Data stewards from health, public safety, and other agencies across federal, state, and local government and health systems will provide existing and new non-sensitive de-identified data to this challenge for the teams to explore and use.
In 2016, Virginia’s Commissioner of Health, Dr. Marissa Levine, declared the opioid and heroin overdose epidemic to be a public health emergency. Overdose deaths have continued to increase year after year; over 1400 Virginians lost their lives to drug overdose in 2016, and 1133 of those were attributable to prescription opioids and heroin. This is a war being fought on two fronts. While prescription opioids are driving the epidemic in Southwest Virginia, heroin/fentanyl pose an additional factor in the rest of the Commonwealth. This indicates that the ways in which we address prevention, treatment, and interdiction must be data-driven and tailored to the needs of individual communities. Collaborative responses and strategies involving both public safety and public health are imperative in ending this crisis in Virginia.
Recent data shows these alarming trends in Virginia:
The teams will be encouraged to develop new insights and applications that will answer burning questions and address challenges that face government and community stakeholders.
Battling the opioid epidemic requires knowledge, research, and resources to ensure change occurs at the required levels and that policy supports those changes. The Federal Government and the Commonwealth of Virginia’s state and local Governments allocate millions of dollars annually to help fund treatment and prevention services for opioid and substance abuse in the Commonwealth. However, how can Virginia effectively measure how resources are allocated? Better yet, how can Virginia use data analytics to develop an approach to address this growing epidemic across the Commonwealth?
Solution: Kearney & Company, P.C.’s (Kearney) solution focuses on the accountability and transparency of the resources (e.g., money, personnel) spent towards treatment and prevention services and how those resources align to risk and need (e.g., opioid related deaths, economic factors).
Impact: Based on the data, there was a potential $22 million in Fiscal Year (FY) 2016 that was available but went unused. The Commonwealth could have redistributed this unused money to locations with the highest risk and greatest need.
Using Virginia Department of Health’s socio-economic factor tracking, state and local financial statements and data, and Federal award data, Kearney developed a tiered approach to align spending data with rates of abuse by drug type (i.e., fentanyl, heroin, and prescription opioids). This risk-based approach identified areas of highest risk as compared to the distribution of funds.
Way Forward: Limited resources must be prioritized and monitored to ensure effectiveness. This requires data-driven policy decisions and transparency over spending. In the next FY, the Federal Government is moving towards establishing a dedicated fund to monitor opioid spending. The discussion for prioritizing resources should be based on a data-driven, risk-based approach that evaluates funding to areas requiring the greatest assistance. Kearney’s solution highlights the need for increased transparency and oversight related to opioid prevention and treatment programs.
In response to the challenge of the Governor’s Datathon to respond to the opioid crisis, the VDH team chose to focus on a message that team members had heard called out during the Population Health Summit earlier in the year; the message being that many resources exist within the community for people impacted by opioids but the challenge is connecting people to the resources. As a result, the solution the team wanted create was a mobile application that would provide three types of information; first, the rate of opioid prescriptions and substance abuse in the user’s locality of residence, second, the users risk for experiencing a fatal overdose based on their residence, sex and age, and third, provide users awareness of and access to resources in the community around them.
The team was able to produce a working version of the application for iOS utilizing Tableau for the visualizations. The team is planning on continuing development of the application, incorporating additional resources and expanding the user risk component to include non-fatal overdose. The team hopes to make a version of the application available by the end of the year.
Our team created an intake process designed to help doctors measure patient risk of opioid addiction. The solution is an early intervention approach. The process takes inputs from a patient health survey and from toxicology reports and stores the information in a centralized database. An algorithm is then applied to the collected data and a risk score is output to the patient’s doctor. This algorithm would also become more accurate as more patient data was obtained, allowing patients and physicians to better manage the over-prescription of drugs that could lead to addiction. If implemented properly and mandated by the appropriate legal framework, this solution has a vast potential to help data sharing, outreach, education and law enforcement to fight the opioid crisis.
Mountain States Health Alliance operates 14 hospitals and multiple physician clinics in Northeast Tennessee and Southwest Virginia – an area hit especially hard by the opioid addiction crisis. The health system has undertaken a number of initiatives to help reduce the burden of addiction in the Appalachian region, including recently opening a recovery-based addiction treatment center that combines medication, counseling, and social services with the research capabilities of East Tennessee State University to not only help individuals in crisis but also gather meaningful research data about addiction treatment that can help future generations.
The Datathon team from Mountain States created a cloud-based data platform that may compliment treatment-based research by compiling other types of data relevant to the opioid crisis from external, publicly available datasets and organizing the public data in a single repository. The platform features visual analytic dashboards that allow users to drill down into various geographic areas to spot trends and identify correlations that are useful for strategy creation. When fully developed, the resource could be used by city or county officials to create proactive solutions in certain hot spots before a particular crisis escalates.
Since 80% of heroin users first abuse prescription opioids, VDOT’s Datathon team sought to reduce the likelihood of opioid addiction by changing patient and prescriber behavior.
Data on whether patients are holding onto excess opioids was not available, but would be a powerful indicator of over-prescribing and could provide a feedback loop to prescribers that does not currently exist.
Using a combination of real and test data, the team built four apps and a map, and proposed a pilot to improve the ecosystem around prescription opioid return, collect new data on excess opioid levels and provide new behavior-changing insight and transparency to prescribers and the public.
We looked at the data and at scholarly research, and determined that implementing a version of the Icelandic Model in Virginia would help. Iceland had a huge drug problem with Adolescents that grew as those Adolescents grew up into Adulthood. They asked America for help, and through a Structured, Scientific Approach, were able to turn their country around. They have exported this model with success to many places in Europe. It involves facilitating communication between parents, teachers, and government researchers. They found that the more unstructured activities that Adolescents did, the more likely they ended up a Drug User.
The more Structured Activities that Adolescents did, the LESS likely they ended up a drug user. Also, Peer Group attitudes and behaviors were the major predictor. This later was found to also be applicable to adults as well. So a two pronged approach of instituting the Icelandic Model for Adolescents and facilitating Peer Recovery for Adults is our prescription. Setting up Recovery Schools and using Certified Peer Recovery Specialists in Medical Entry Pathways such as Hospital Emergency Rooms. Also, instituting a 5 day prescription for most outpatient pain relief instead of the now normal 30 day supply. And if they ask for refills, involve a Peer Recovery Specialist to talk to them and their family about the dangers and monitor their pain and recovery.
Limited state and local resources in the form of federal funding, personnel, and infrastructure must be managed across multiple counties and townships to manage the opioid crisis. From locating enough treatment facilities to scheduling patrol cars to scan for drug-related crimes, public resources to combat opioid overdoses and fatalities require data-driven methods for targeting policy interventions. At the Governor’s Opioid Addiction Crisis Datathon in September 2017, the Booz Allen team of data scientists developed a prediction model to locate areas of high risk across multiple indicators in the State of Virginia. We believe this approach could be applied for states across the nation, particularly those with high rates of death due to drug overdoses and those with significant increases in death. The Datathon provided a combination of publicly available data and State of Virginia datasets consisting of Census data, socioeconomic data, crime data, treatment center data, funding data, mortality and morbidity data for opioid, prescription drugs (i.e. oxycodone, fentanyl), and heroin cases, where dates started as early as 2010.
All data was cleaned, formatted, and analyzed at the county-level using FIPS codes. The goal of the model was two-fold. First, to predict counties of similar populations and behaviors using an unsupervised machine learning to identify counties that were similar to each other in terms of socioeconomic, demographic, and crime indicators. This was important because neighboring counties like Goochland and Henrico Counties, while sharing a border, do not necessarily share similar behavioral and population characteristics. As a result, counties in close proximity may require different approaches for community messaging, law enforcement, and treatment infrastructure. Next, the team developed a risk model that calculated across fifty data elements such as poverty level, age range, physicians, median income, crime types, and federal and state funding figures, to measure the risk of opioid mortality and morbidity at any of the 103 counties in Virginia. Identifying how counties are similar or dissimilar to each other, and running the risk model to calculate their level of morbidity and mortality severity based on adjusting any number of indicators will allow a policy maker to pin-point where to allocate funding.