Opioid Mapping Convention

Today I had the chance to attend Jeremiah Lindemann’s Opioid Mapping Convention here at the Broomfield Esri office. Information about this convention had actually reached me this spring, before I had even started my Esri internship, as a forward from a friend who had received this email through a CU Boulder email list. I contacted Jeremiah a few weeks ago to check on my attending this convention during my workday, and he graciously welcomed me!

The convention was packed! Myself and other Esri employees sat to the side of a room filled with a variety of people. We had representation from law enforcement, around five local governments, scientists, health care professionals, and more. I didn’t get a chance to hear personal introductions from everyone, but Jeremiah but together an amazing visual representing the diversity of fields and expertise in the room for the convention using Survey123 data collected from a convention RSVP form.

Jeremiah opened the meeting with a brief introduction to the context of the convention. Jeremiah is a Solutions Engineer for Local Government at Esri. A number of years ago, he started working with models for how mapping can assist in opioid awareness with the goal of mitigating over-prescriptions, overdose, and death. He entered into this area as a passion project, and shared with us his moving personal story of how his brother’s opioid overdose led him to not only grieve, but use this impetus to start using mapping solutions to address this growing area of concern. He shared with us two of the first Esri products he developed in this area to help share resources and awareness with a larger audience. He shared an example from the Northern Kentucky Health Department, which had put together a collection of opioid information resources in an easily digestible and sharable format. Many of the opioid awareness resources were built with Story Maps, which was fun to see real-life examples of Story Maps reaching a large audience. One Story Map Jeremiah had developed early on was a tragic collection of photos and short biographies of individuals who had died from opioid overdoses. This Story Map attempted to destigmatize opioid overdose through putting faces and personal stories to the deaths related to this epidemic. It was hard to read, but also provided a necessary grieving platform for families and friends directly affected by the opioid epidemic.

Opioid mapping with Esri has now expanded beyond these first steps and products to be a full Opioid Mapping Initiative. Jeremiah, as an Esri Solutions Engineer, builds maps and apps that can be used widely for opioid mapping across state and local governments and beyond. Nation-level maps are now being developed to visualize the opioid epidemic, including the National Naloxone Map, which includes prescription drug drop-off locations.

To finish up an introduction to this Opioid Mapping Initiative, we went some options for how to visualize incidents well. Since opioid mapping most often deals with confidential medical and death data, anonymity must be preserved even in mapping. We discussed Bloomington, IL as a counterexample. Bloomington has displayed addresses on a map of what should have been confidential data, and made news for it in a bad way. We then discussed a few examples positively modeling how to show representations of confidential medical data. Tempe maps data through randomizing data location in a block area. This preserves some degree of precision in the data display, while still preserving privacy. Another way people map this data is by generalizing by region or county. However, this technique is really too broad to actually understand the epidemic on a community level and produce effective insights (we discussed this to a greater degree in the presentations from CO county governments who also have different generalization techniques). A great way to visualize data is in a heat map, which blurs data location, but displays as aggregation, which can show hot-spots of activity. Lastly, a new technique pioneered by New Orleans was brought up: the hexbin method. This one seems most exciting to me, and will probably grow in popularity. Displaying data in hexbins can still generalize location while establishing fine-grain data visualization. Hexbin visualizations can also be shown as modified choropleth maps, which can provide density insight similar to a heat map. All of these visualization techniques help with measuring impact at different scales. Coursework in university settings exploring opioid mapping could further improve how to visualize this data, which often has varied privacy standards in place across the country.

After this discussion of some mapping examples, we moved into the presentations that made up the bulk of this Opioid Mapping Convention. Counties from across the CO Front Range joined us and discussed their current work with opioid mapping and response.

Tri-County

Tri-County Health Department opened the presentation session. Adam Anderson gave the primary presentation. He discussed how they attempted to showcase easily reproducible data. They were also exploring how to best show and analyze rate data for the opioid epidemic.

Tri-County Health Department serves over 1 million people, which is huge. This, especially in my Idaho terms, is so massive. I forget what a huge and sprawling urban area the Denver region is. Tri-County, with this large population, has the goal of aggregating and communication a massive amount of data across the region. Before mapping, Tri-County had been communicating data in trend graphs and tables. These were all over the place and hard to make sense of, however, and often required large time expenditure to go through tables for in-depth analysis. As they now transition to mapping, they can display large and complex data sets in a spatial way that is easier to make sense of, visually, and can display data more specifically than trend graphs.

Tri-County brought up an important note about data privacy in Colorado that would come up in each following presentation. The Colorado Department of Public Health doesn’t release private medical data unless more than 2 events occur in the same geographic location. This is a confusing restriction, and one that each county had to work with creating opioid maps for the public. This data visualization restriction can result in data being aggregating into larger regions, which allows for it to be represented, but with a decreased degree of specificity, or modified in ways similar to those discussed above.

Tri-County’s data sources include death certificate data, New America, and State Health Department emergency preparedness data. They noted in their presentation that death data is useful, but is not accessible for the public until ~1.5 years past death, which can make this data less current and applicable. Later presentations also addressed how they are looking to alternate data sources, such as emergency response coding, to gain more current data in place of or in addition to death data.

Tri-County showcased some of their mapping and the insights about the region of their area that had hot-spots of opioid activity. They used a heat map in visualization, and displayed this in action. They have also put together a fantastic web presence that clearly showcases data on the opioid epidemic, and allows for very easy browsing and knowledge transfer around opioid subjects and initiatives. This is one of the first data sites they have put together, and I think it looks great. Their website and most of their current maps can be found on their Opioid Crisis website.

Boulder County

Three individuals from Boulder County joined us next for a discussion of their work in opioid mapping. They showcased a unique collaboration between the county IT department and Public Health. Molly Watson is a GIS Specialist in IT and a biostatiscian. She meets with Public Health officials through a special Opioid Advisory Committee the county has formed, which meets every month, and is also open to the public. They highlighted that there are approximately 80 people in each months meeting!

Boulder County is working with AMR for data acquisition. Boulder as a county has fewer prescription deaths. For more in-depth and accurate data analysis of the opioid epidemic in Boulder County, they are pioneering different data capture from AMR, which will hopefully also incorporate greater data on fentanyl and other drugs.

They are working on a number of projects, including:

• Death rate by city, though this is a proving too large of an aggregation to come up with specific community initiatives to offset opioid overdoses
• Narcan locations: Developing an app to display nearest Narcan location, as well as public transportation options for how to get to them. Through this project, they are also attempting to standardize data of where Narcan/Naloxone is being distributed.
• Drug disposal and needle drop-off locations

Boulder County also brought up some innovative new data management techniques. They noted that Google Assistant could start assisting with manual data maintenance, including data acquisition through calling. They also mentioned public editing and contribution to open data initiatives. Public Health also doesn’t have a dedicated GIS team yet, which would be helpful for more GIS analysis. Additionally, they are working with privacy laws, but are confronted with the issue that privacy laws often cause the sequestration of accurate rural data around opioid use. They are exploring ways to work with privacy laws without sequestering rural data, including the hexbin method.

N. Larimer County Health District

N. Larimer County had a fantastic presentation, showcasing their multi-faceted approaches to projects to understand and mitigate the opioid epidemic. They shared their overarching project umbrella: get standing orders of Naloxone into the hands of those who need it.

I really appreciated how they broke down their overarching project into a series of goals and approaches to best meet this goal. They took an innovative, smart, and unique approach to data acquisition, which used new sources other counties had not mentioned, as well as ground-surveys for a more qualitative analysis approach.

They are pioneering data acquisition from police calls. This was where they started some analysis. They are able to get closer to real-time overdose data not through death certificate data, which can be delayed (as mentioned above), but through gathering information from police calls where “overdose” is in the report. They also use a hospital discharge dataset, as well as CDPHE Vital Statistics. Using this overdose data, they initially found that overdoses were clustered around downtown Fort Collins. However, they were incredibly admirable statisticians, and did not stop the analysis here. They recognized that overdoses may not be directly tied to community location, and may occur with greater incident rate because of the density of activity by a diverse group in the downtown area. When looked at in isolation, this overdose visualization may even suggest that overdoses are related to the homeless population, because the downtown region also had the greatest homeless populations. However, through concurrent data acquisition, they added unprecedented depth to this initial analysis, and provided visualizations showing this was a housed issue facing the country on a broad scale.

This team implemented an incredible approach combining these data sources with on-the-ground survey data. They partnered with UNC School of Public Health to perform a walking audit of downtown businesses, following this discovery of high overdose rates in the downtown area, to identify target areas for Naloxone distribution. However, this audit actually revealed that most downtown businesses were not familiar with Naloxone or the opioid epidemic in general. They took this data, and shifted their priority from distribution to education, in an attempt to first start with Naloxone and opioid education to help downtown awareness of these issues and proper response prior to drug distribution. This was an intimate and actionable look at opioid awareness, and one that they are now acting on.

Another data acquisition technique they implemented that added a wealth of insight to their project was a hand collected Community Health Survey of Larimer County. This was distributed to randomized addresses, and was to be filled out by one member of the household. This survey allowed them to look at data on a census level and fold in this local survey data. This was an incredibly smart and insightful approach, and provided tremendously great insight. This combination of survey and census data allowed they to identify the geographic distribution of opioid use. They masked data with kernel density to preserve privacy, but this still allowed for neighborhood-level visualization (note that this still may not address rural communities in the county). This data aggregation and visualization showed true areas of need for opioid education and response. In distinction from the initial overdose visualization, these data sources showed that the opioid epidemic was a housed, neighbor-hood level, community issue. The survey realized the need for more opioid education. Using this information as well as the insight of the downtown audit, the county is now going to hit Overdose Awareness Day hard in an attempt to greatly increase public education around opioids.

I so respect this multi-faceted, flexible, and informed approach that based off of a guiding question, but relied on multiple data acquisition techniques and combinations to pinpoint first steps to increase opioid awareness and response on a community and county level. Thank you, thank you, for capping off the county presentations with this in-depth showcase of a comparative approach reaching many insights about this epidemic.

Questions and comments

Following these county presentations, a police chief in attendance noted that a large dataset missing from these county analyses was a mental health dataset. This is an area yet to be explored in combination to these previous approaches, and is incredibly important to add in.

We also had a short presentation from a representative from the state level. He was a whip-smart presenter that went through so many facets of the opioid epidemic in a flash. Importantly, he complimented these county initiatives, and noted that the opioid epidemic will not be fought as effectively on a state level; this issue requires community-level analysis, education, response, and funds, and the state is just too distant and large of a scale to have real, actionable impact on this multi-faceted issue facing communities. Public awareness happens locally, and he encouraged more county level work, followed by local coalitions that could feature more sharing out of best practices similar to what we were doing today at Esri. He thinks the state could help with “delivering collaboration” scaffolding and supporting these exchanges for the county. The state can also help with data acquisition and transparency of methodology in mapping this epidemic.

Wrap-up

The most important note about data acquisition and display that came up over and over again was how to deal with restrictive privacy laws to still show fine-grain data. As death data has many privacy laws around in, the new frontier in opioid mapping, especially for the public, may be to turn to new, innovative, and reliable data sources. A very promising area for this could be EMS coding of specific, query-able terms that can provide closer to real-time insight on opioid and other drug use.

Thank you

Thank you again to Jeremiah Lindemann for putting on this convention, to all of the county representatives who presented, and to the diverse and engaged audience! Thanks for letting me attend.