Tackling the Opioid Epidemic: Research and Policy Perspectives

The U.S. faces a growing opioid epidemic that has impacted nearly every family. Over the last few decades, drug overdose deaths from synthetic opioids, including fentanyl, have surged in the U.S. The opioid crisis is multifaceted, requiring collaborations among communities, policymakers, and public health professionals across sectors to fully understand the complexities of addiction, and apply our combined knowledge and expertise to identify and develop effective mitigation strategies. Statisticians play a critical role in providing data-driven insights into this problem and in evaluating potential policy options to curb opioid use. To learn more about the current situation and challenges, and to gain policy perspectives on the issue, we are happy to interview Dr. Beth Ann Griffin, Senior Statistician at the RAND Corporation and Co-Director of the RAND/USC Opioid Policy Tools and Information Center (OPTIC).

Robert: Could you start off by telling us a little bit about yourself and the RAND Corporation? What are some of RAND’s goals/visions?

Beth Ann: Yes, I am delighted to speak with HPSS today, and happy to introduce myself and RAND! First, the RAND Corporation is an impactful research organization that develops solutions to public policy challenges with the goal of helping to make communities across the world safer, healthier, and more prosperous. Notably, RAND has about 1,200 researchers working across a vast number of different topics and substantive areas including public health, public policy, education, criminal justice, the environment, and national security. It is hard to believe that I have worked at RAND for the past 18 years. I love that I have great colleagues and that we are taking on some of the most significant problems that impact the world. At RAND, I also have the privilege of being part of an amazing community (Statistics Group) that has been pivotal to my growth, success, and well-being. My statistical research involves the development and dissemination of methods for estimating causal effects using observational data in various applications including addiction, mental health, rare diseases, military health, and education. Of note, I also currently co-direct OPTIC, whose goal is to foster innovative research, tools, and methods for tackling the opioid epidemic.

Robert: Does RAND offer any opportunities for students or recent graduates who are interested in the organization?

Beth Ann: Yes! We have wonderful opportunities for students and recent graduates who are interested. For current graduate students, we offer a Summer Associate Program that runs for 12 weeks and introduces graduate students to policy, projects, and working at RAND. Typically, a Summer Associate will work on one or two projects during the summer to see how RAND really works. I have worked with several Summer Associates over the past 18 years and find the experience to be rewarding for both the student and the project team. It often results in a peer-reviewed publication or co-authorship on a RAND report chapter. Many Summer Associates end up applying for permanent research positions at RAND. We are also actively hiring recent and rising graduates, with or without a postdoc. As I noted earlier, one of the best parts of my career at RAND has been the Statistics Group. Our group has a commitment to mentoring, and we work hard to grow our new hires and ensure they find meaningful projects and collaborators; we also consult with them on any statistical needs for their projects. Our group is a great place to be for a recent graduate given the priority we place on both mentorship and community.

Robert: What is a typical workday like for you at RAND, and what do you enjoy most about the organization and your job?

Beth Ann: My typical workday can be divided into two general formats. I will have a couple days a week where I keep my calendar free so I can focus on my research. These days will include running analyses for a project, running simulations for a recently developed method, or writing for publications/proposals. For example, this past month, I have been busy with producing a publication targeting the Annals of Applied Statistics with my OPTIC team that uses Bayesian debiased autoregressive models to examine the impact of four policies on opioid-related outcomes. I am very excited about this paper! My other type of day tends to be “collaboration days,” which include a full day of meetings with my project collaborators or mentees. With my project calls, we use the time to catch up on project needs, challenges, and potential solutions. With my mentees, we use the time to catch up on RAND work, work/life balance, power calculations for proposals, and any other topic that might be of interest to them.

Robert: Before diving a bit deeper into your opioid-related work, could you briefly talk about how you got involved in this area as a statistician?

Beth Ann: Great question! I am relatively new to state policy evaluation methods and applications. The first 10 years of my career at RAND focused on developing and disseminating best statistical practices for addiction researchers. I served as a primary investigator (PI) on four R01 grants funded by the National Institute of Drug Abuse (NIDA), each of which combined novel methodological research with the estimation of causal effects for substance use treatment research for adolescents in observational settings. As part of these efforts, we developed over 20 software tools and tutorials related to the use of the Toolkit for Weighting and Analysis of Nonequivalent Groups (TWANG) package, which uses a nonparametric machine learning algorithm to estimate propensity score weights. In 2016, I found myself in a lull in funding and reached out to a close collaborator at RAND to inquire if he needed a statistician for his work. That email request led to me joining his research collaborative called the RAND Gun Policy in America Initiative, whose goal is to provide rigorous information around what research can tell us about the effects of gun laws. In that work, we developed and executed a critically needed set of simulation studies to identify optimal methods for evaluating the impact of gun policies and subsequently used the results of the simulation studies to design a series of high-profile studies about different classes of gun control policies and their potential impacts on outcomes. This project led me to start working with my OPTIC co-directors (Bradley Stein and Rosalie Pacula at USC) on efforts to secure funding for OPTIC, where we proposed to expand and reproduce the simulation studies within the opioid context as part of a national center. As you can see, methods from one policy area can often lend themselves well to other emerging topic areas.

Robert: Just to give people a little bit of background on the issue, could you briefly talk about the current state of the opioid epidemic and some of the major issues at hand where statisticians can offer their skills?

Beth Ann: The opioid epidemic is a public health crisis that began in the 1990s. Between 1999 and 2019, nearly 500,000 people died from an overdose involving opioids. The impact can be felt far and wide, with around 40% of adults in the U.S. personally knowing someone who died from an opioid overdose. We are currently in what is called the “fourth wave” of the epidemic with fentanyl and stimulants driving the latest increase in fatal overdoses, likely reflecting the spread of fentanyl and greater drug misuse associated with the increased stress, social isolation, and job loss stemming from the COVID-19 pandemic. It has caused extensive harm and devastation in the U.S. In response, states continue to implement an array of policies meant to combat this poly-substance crisis, producing a policy landscape that is complex and dynamic. This where statisticians can offer their skills! Unfortunately, obtaining credible estimates of the causal impacts of opioid-related policies is challenging for a variety of reasons. First, the sample sizes are highly constrained in state policy evaluation studies – with at most 50 states in your sample – resulting in a large amount of uncertainty around any estimate. Second, policies do not happen in a vacuum and disentangling the effects of specific policies from other policies is crucial in this setting so policymakers can understand what is and is not working. Finally, the likely effects of many state policies will be small, but even small effects here are important because they will translate into saving thousands of lives. With so many complexities, this is where statisticians are needed to help provide data-driven insights. Statisticians can work in multidisciplinary teams that include healthcare providers, policymakers, law enforcement, and community organizations. Their expertise in data handling and analysis is crucial for the integrated approach needed to tackle this complicated issue.

Robert: Could you describe some of the work you have done, or are currently doing, related to opioid policy and how you apply statistics in this space?

Beth Ann: We spent the first 5 years of OPTIC testing and refining simulation methodologies to assess the relative performance of different statistical methods and providing critically needed information to the field on best practices for estimating the overall average effects of opioid policies. We published papers describing key findings from our simulation studies related to three different cases: no confounding, cooccurring policies, and confounding. We believe findings from our simulations can help researchers identify which models provide the most accurate estimates of state-level policy effects in the presence of concurrent policy implementation and confounding between enacting and non-enacting states. We also developed an R library called ‘optic’ that can be used to run all of our simulations on any repeated measures data. This valuable resource enables researchers to routinely perform similar simulations when beginning to work with a new outcome series. In general, I love creating R packages for any new method as well as ensuring those packages have user-friendly tutorials to go along with them. I believe such efforts as statisticians are critical to ensure reproducibility and a broad uptake of new methods, which is an important part of our role as statisticians in this world. We have also written several publications that provide guidance for researchers on the proposed potential solutions for common challenges in opioid and other state policy evaluations, and developed new methods to address several of these challenges. I lead this third project together with fabulous collaborators from inside and outside of RAND, including Liz Stuart from Hopkins who is a member of HPSS (shout out to Liz!). It was a busy 5 years and we just recently secured funding for an additional 5-year period. The next five years will be focused on understanding how to best estimate policy effect heterogeneity. Currently, statistical methods for policy evaluation generally estimate overall average effects and do not consider effect heterogeneity. Methodological work is critically needed to identify optimal approaches to identify and characterize effect variation both within and across states.

Robert: What have been the broader implications of your work in this area? Have you engaged with any policymakers and if so, how did you successfully convey your statistical findings to a non-statistical audience?

Beth Ann: One of OPTIC’s core aims is the dissemination of our tools, methods, and findings to communities of interest to ensure uptake of more robust methods and more accurate findings around the effectiveness of different policies. As part of this effort, OPTIC team members have formally spoken with multiple individuals and decision-maker groups from government agencies or elected officials and their staff. Additionally, OPTIC team members meet regularly with Advisory Board Members from the National Governors’ Association and the National Conference of State Legislatures, and RAND’s legislative analyst and congressional liaison for ongoing discussions of top legislative priorities. This type of engagement is critical to ensuring that our efforts align with the interests and needs of policymakers. We also recently published a pragmatic guide to help decision-makers determine how much they can trust the results of a health policy evaluation. The guide, lead by Megan Schuler at RAND, offers a few rules of thumb to identify policy evaluations that provide strong evidence about whether there is a cause-and-effect relationship between a policy and an outcome. The guide also highlights some red flags indicating that a study’s findings should be viewed with caution. The exercise of being involved in this effort taught me a lot about the challenges behind ensuring that we communicate to non-statisticians in non-technical language, and the positive reception by policymakers and their staff has highlighted the deep need for such efforts as statisticians. The other product I feel is useful for any work we do under OPTIC is that we develop Key Takeaways which are OPTIC-created, one-page summaries of published opioid research findings tailored for policymakers. We supply these to our legislative partners for distribution, use relevant ones in our conversations with policymakers, and also post them to our website to increase their accessibility. Of note, OPTIC has been vital to our successful dissemination efforts and research from it was recently highlighted by the White House as well as John Oliver on his May 12 show devoted to Opioid Settlements, which was all very exciting!

Robert: How do you think statisticians can be the most useful as we continue to confront the opioid crisis?

Beth Ann: We as statisticians can be incredibly valuable in addressing the opioid crisis in several ways. First, we can analyze existing surveillance data to identify state and nation-wide patterns and trends in opioid use and overdose rates. This type of work helps us to understand the scope and scale of the crisis. We can also develop models to predict future trends and identify areas at highest risk. Second, we can work to help the field by both designing and analyzing the effectiveness of interventions aimed at reducing opioid misuse, including the evaluation of state and local policies. This work requires strong collaboration between communities, policymakers, and interdisciplinary public health professionals across all sectors to better understand the pool of evidence. Third, we can play a vital role in assisting state and local governments on how to best distribute opioid settlement funds by developing and utilizing methods that can identify the most impacted areas and suggesting where intervention efforts should be concentrated. Opioid settlements with pharmaceutical companies have already occurred, and there are more to come. Settlement funds could save lives and mitigate lifelong harms from opioid misuse if they are allocated appropriately to the most effective interventions. Through these roles, we can provide crucial insights and tools that help in understanding and combating the opioid crisis effectively.

References:

1. Griffin, B. A., Ridgeway, G., Morral, A. R., Burgette, L. F., Martin, C., Almirall, D., Ramchand, R., Jaycox, L. H. & McCaffrey, D. F. Toolkit for Weighting and Analysis of Nonequivalent Groups (TWANG) [Online]. Santa Monica, CA: RAND Corporation, 2014. [https://www.rand.org/statistics/twang.html]

2. Schell, T.L., Griffin, B.A., and Morral, A.R. Evaluating Methods to Estimate the Effect of State Laws on Firearm Deaths: A Simulation Study, Santa Monica, Calif.: RAND Corporation, RR-2685-RC, 2018. [https://www.rand.org/pubs/research_reports/RR2685.html]

3. Schell, T.L., Cefalu, M., Griffin, B. A., Smart, R., & Morral, A.R. (2020). Changes in firearm mortality following the implementation of state laws regulating firearm access and use. Proceedings of the National Academy of Sciences of the United States of America, 117(26), 14906-14910. [https://doi.org/10.1073/pnas.1921965117]

4. Griffin, B.A., Schuler, M, Stone, E.M., Patrick, S.W., Stein, BD., Nascimento de Lima, P, Griswold, M, Scherling, A., & Stuart, E. (2023). Identifying optimal methods for addressing confounding bias when estimating the effects of state-level policies. Epidemiology. 34(6):856-864. [DOI: 10.1097/EDE.0000000000001659]

5. Griffin, B.A., Schuler, M.S., Pane, J., Patrick, S. W., Smart, R., Bradley, D. S., Grimm, G., & Stuart, E. (2023). Methodological considerations for estimating policy effects in the context of co-occurring policies. Health Serv Outcomes Res Method. 23(2):149-165. [DOI: 10.1007/s10742-022-00284-w]

6. Griffin BA, Schuler MS, Stuart EA, Patrick S, McNeer E, Smart R, Powell D, Stein BD, Schell TL, Pacula RL. (2021) Moving beyond the classic difference-in-differences model: a simulation study comparing statistical methods for estimating effectiveness of state-level policies. BMC Med Res Methodol. 13; 21(1):279. [DOI: 10.1186/s12874-021-01471-y]

7. Schuler M. S., Griffin B. A., McGinty EE, Cerdá M, & Stuart EA. (2020). Methodological Challenges and Considerations in Opioid Policy Evaluation. Health Serv Outcomes Res Methodology. [DOI: 10.1007/s10742020-00228-2]