With a 200% increase in the rate of opioid-involved deaths since 2000,1 many US states and the federal government have declared states of public health emergency.2 Government responses to the epidemic include focused research on and treatment of opioid dependence3 and system-level interventions and policies to reduce overdose.4 As part of this work, various local and state health departments are developing opioid data dashboards containing visualizations, descriptive information, and downloadable data or reports.
Opioid data dashboards can potentially improve our understanding of the opioid epidemic, facilitate community planning, promote evidence-based decision making, and support monitoring and evaluation. Yet, will these data dashboards meet these goals? Past experience from working with local and state health departments has shown that government resources are primarily devoted to collecting and using data for analysis, rather than for communication or orienting communities toward action. As health agencies develop opioid data dashboards, we offer 10 recommendations for improving data-driven communication about the opioid epidemic. These are based on our experience studying and promoting open data platforms, working on data dashboard platforms, and consulting with government agencies and others in their data orbit (eg, foundations, nonprofits, university research centers) about data communication strategies.
Lesson 1: Use Available Data, Rather Than Waiting for That Ideal Data Set
Perfect data are elusive. Mortality data limitations include incomplete and inconsistently reported toxicology information across jurisdictions1 and biased reporting from elected coroners about controversial deaths such as suicides.5 Other data challenges relevant to tracking opioid trends include lag times, heightened awareness of overdose leading to increased reporting, uneven uses of electronic versus paper-based ambulance data, missing data on naloxone administrations reported by first responders,6 known differences in metrics derived from administrative versus clinical data,7 and disjointed trends in opioid-related inpatient stays after the implementation of the updated International Classification of Diseases manual.8 These limiting factors could persuade agencies that their data are insufficient for communication purposes. Yet, following Crosby, Stills, and Nash9 (“If you can't be with the one you love, honey/Love the one you're with”), if you can't be with the data you love, love the data you're with. New York publishes quarterly county reports of opioid-related deaths, emergency department visits, hospitalizations, treatment admissions, and naloxone administrations based on data through the preceding quarter.6 To reinforce the provisional nature of these data, reports contain detailed explanations of how counts evolve as data are received by the health department and cautions on interpretations.
Lesson 2: When Data Do Not Exist, Turn to Others Facing a Similar Challenge
A local health department might simply lack data on opioid overdose outcomes that can help make a persuasive case for action. This may make it difficult to change the narrative of the local opioid problem from an issue of individual responsibility requiring law enforcement solutions to a broader public health issue that also encompasses prevention, treatment, and harm reduction. In these instances, it can be helpful to find other local communities with similar profiles that have seen success with public health-oriented solutions. Their outcomes can be shared to support a new approach even when local data are not available.
Lesson 3: Create Data Stories, Not Just Data Visualizations
The data visualization field has evolved from creating visualizations to telling stories. Since the publication of Edward Tufte's classic The Visual Display of Quantitative Information,10 diverse disciplines have evaluated effective graph design and synthesized recommendations on communicating information visually.11,12 Yet, similar to how stories are persuasive in advertising,13 there is a role for scientific storytelling.14 The California Health Care Foundation funded LiveStories, a data storytelling start-up, to work with roughly 10 county-level opioid coalitions to create dashboards. The dashboard templates15 provided local-level data and encouraged local coalitions to add personal stories to supplement the data with memorable information about affected individuals.
Lesson 4: Appeal to Both the Heart and the Head
Data represent facts, and researchers are trained to search for truth and present findings neutrally. Yet, persuading communities to take action may also require tugging at the heart strings, which the Sonoma County Department of Health Services sought to do in its farmworker health survey.16 The report incorporated poignant photographs alongside objective data and charts to make a persuasive case. If a health agency is unable to convey emotion due to political constraints, community partners can take that role. In these partnerships, health departments can provide the data and community organizations working directly with local populations can supply the stories.
Lesson 5: Enlist Writers, Designers, and Community Members When Crafting a Story
Most health data experts are not data communicators. In the Sonoma County farmworker survey, epidemiologists collaborated with the health department's communications staff. In a different example, the digital agency Velir and the communication firm Purpose collaborated on a report about the number and demographic characteristics of shift workers and their financial, health, and other daily challenges. Although rooted in data, the information was presented in a story that incorporated numerous quotes and videos to help readers understand the experience of irregular and unpredictable work schedules.17
Lesson 6: Don't Start With the Full Story
Health researchers are trained to share extensive information on the data and its analysis for transparency and replicability. However, this approach is a quick death for communicating stories to broader audiences. A more productive strategy is starting with a catchy visualization to draw in readers, who can subsequently read a short summary and then a full report if interested. The dashboard templates created for the California opioid coalitions provided an option of a teaser to briefly set the stage and pique a reader's interest. In 1 example, the teaser was a statement in large font that “prescription opioids now kill more people than firearm homicides” and accompanying line graph. This was followed by a question, “How many people died from prescription opioid overdoses in your county?” The page subsequently scrolls to a map of California counties and hyperlinks to counties' opioid report cards.15
Lesson 7: Make Large Numbers Relatable
Large numbers are hard to grasp. The Centers for Disease Control and Prevention's opioids landing page starts with a striking statement: “Opioids (including prescription opioids, heroin, and fentanyl) killed more than 42,000 people in 2016, more than any year on record.”18 But are 42 000 annual deaths large or small? The field of gun control faces a similar challenge when portraying the number of people killed by gun violence. The Washington Post showed visually the 1102 individuals killed by mass shootings through June 29, 2018.19 The focal point is a visualization containing 1102 silhouettes representing each victim. Scrolling a cursor across the silhouettes triggers pop-up boxes describing details of each victim and highlights the silhouettes of other victims from that shooting. Users can click silhouettes for further information about each shooting. Similar visualizations provide information and context for the 298 guns, 158 shooters, and 154 shootings. In making large numbers relatable, consider cognitive psychologist Paul Slovic's reflections on psychic numbing to mass genocide:
Our cognitive and perceptual systems seem to be designed to sensitize us to small changes in our environment, possibly at the expense of making us less able to detect and respond to large changes .... Numerical representations of human lives do not necessarily convey the importance of those lives. All too often, the numbers represent dry statistics, “human beings with the tears dried off.”20
Lesson 8: Always Talk to End Users
A major challenge for health department staff preparing data products such as open data platforms is uncertainty about the end users and their desired usage of data.21 Will they be using the data in presentations to the local business community or in one-on-one meetings with policy makers? What is their data expertise and content knowledge? Understanding data consumers can help health agencies more successfully “package” the material. When developing its Prevention Agenda Dashboard,22 the New York State Department of Health solicited extensive input from its intended audience of local planning groups to ensure that the site's content and features would match their data needs and technical skills. While researchers typically want raw data files, some clients prefer data products in the form of 2-page printable handouts or PowerPoint presentations. End users should be consulted throughout the process and not just for beta testing.
Lesson 9: Involve Local Data Ambassadors in Dissemination
We are all part of a data ecosystem comprising data producers (eg, field staff who collect primary data; analytic staff responsible for assembling, cleansing, and de-identifying the data; and information technology staff who release data electronically), data consumers (eg, government personnel, researchers, journalists, entrepreneurs, and general citizens), and facilitators (eg, third-party consultants and vendors who support government agencies' efforts to make data available to the public).23 This larger network can be leveraged to facilitate dissemination. For example, Health Data Matters, a Case Western Reserve Medical School initiative to improve community health in Cleveland and Cuyahoga County, actively involves community partners from more than a dozen local organizations to help formulate messaging, provide input on data, and encourage use of published information.24
Lesson 10: Don't Forget to Make Data Fun
Data production and analysis require sophisticated skills and the opioid epidemic's toll is sobering. But data communication does not need to be tedious. The shift workers' Web visualization described previously also includes a link to a “Schedule Invaders” interactive game to further engage readers while educating them on the challenges faced by shift workers. The New York Times developed an interactive “You Draw It” feature for readers to guess the trajectory of opioid overdose deaths.25 Readers first see a plot of car accident fatalities from the 1970s and the 1980s and are invited to draw a line representing their hypothesis about how the trend changed in the subsequent decades. After revealing the answer, subsequent narrative describes the potential role of secular changes such as car and road safety, recessions, and distracted driving. Readers do similar exercises for deaths attributable to guns and human immunodeficiency virus. The final vignette is guessing the trend for opioid overdoses. This analysis is not sophisticated, but similar to classroom games, allowing readers to manipulate the graphs and check their answers makes the content on opioid overdose trends more likely to stick and fosters curiosity about the causal explanations.
Implementing These Ten Lessons
Applying these lessons to opioid overdose data is not easy. Data storytelling and Web programming are not core professional competencies in master of public health and doctoral programs, and it may be cost-prohibitive for many health agencies to hire external consultants. The fast pace of the opioid epidemic and the related public health programming make it difficult to redirect staff time to data communication. Data communication also requires a fundamental culture shift, including breaking down data silos, addressing data owners' concerns about end users not understanding the data, and bridging the divide from sophisticated analyses for research and surveillance to data journalism.
But most of these lessons are entirely doable, simply taking practice and a realization of different ways of approaching how best to summarize information. Moreover, there are now multiple data visualization tools such as Datawrapper, Infogram, and LiveStories that are easy to learn and not cost-prohibitive. These tools make it easier to tell compelling stories about how communities are affected by the opioid epidemic and the effectiveness of current overdose prevention programs. While it may be infeasible to apply all lessons simultaneously, with practice and leveraging a data storytelling tool, public health practitioners can communicate about the opioid overdose epidemic much more effectively.
1. Rudd RA, Seth P, David F, Scholl L. Increases in drug and opioid-involved overdose deaths—United States, 2010-2015. MMWR Morb Mortal Wkly Rep. 2016;64(5051):1445–1452.
4. Haegerich TM, Paulozzi LJ, Manns BJ, Jones CM. What we know, and don't know, about the impact of state policy and systems-level interventions on prescription drug overdose. Drug Alcohol Depend. 2014;145:34–47.
5. Klugman J, Condran G, Wray M. The role of medicolegal systems in producing geographic variation in suicide rate. Soc Sci Q. 2013;94(2):462–489.
7. Tang PC, Ralston M, Arrigotti MF, Quereshi L, Graham J. Comparison of methodologies for calculating quality measures based on administrative data versus clinical data from an electronic health record system: implications for performance measures. J Am Med Inform Assoc. 2007;14(1):10–15.
8. Heslin KC, Owens PL, Karaca Z, Barrett ML, Moore BJ, Elixhauser A. Trends in opioid-related inpatient stays shifted after the US transitioned to ICD-10-CM diagnosis coding in 2015. Med Care. 2017;55(11):918–923.
9. Stills S. Love the One You're With. Los Angeles, CA: Atlantic Records; 1970.
10. Tufte ER. The Visual Display of Quantitative Information. Cheshire, CT: Graphics Press; 1984.
11. Asada Y, Abel H, Skedgel C, Warner G. On effective graphic communication of health inequality: considerations for health policy researchers. Milbank Q. 95(4):801–835.
12. Duke SP, Bancken F, Crowe B, Soukup M, Botsis T, Forshee R. Seeing is believing: good graphic design principles for medical research. Stat Med. 2015;34(22):3040–3059.
13. Quesenberry KA, Coolsen MK. What makes a Super Bowl ad super? Five-act dramatic form affects consumer Super Bowl advertising ratings. J Mark Theory Pract. 2014;22(4):437–454.
14. Krzywinski M, Cairo A. Points of view: storytelling. Nat Methods. 2013;10(8):687.
20. Slovic P. “If I look at the mass I will never act”: psychic numbing and genocide. Judgement Decis Mak. 2007;2(2):79–95.
21. Martin EG, Begany GM. Opening government health data to the public: benefits, challenges, and lessons learned from early innovators. J Am Med Inf Assoc. 2017;24(2):345–351.
23. Begany G, Martin E. An open health data engagement ecosystem model: are facilitators the key to open data success? Paper presented at: Proceedings of the Association for Information Science and Technology; October 27, 2017; Washington, DC.