Conceptualizing and Measuring Abatement from the Opioid Epidemic: A Case Study from Pennsylvania

Abstract

Policy Points.

• In 2021, four major pharmaceutical manufacturers and distributors reached a proposed settlement agreement with 46 state Attorneys General of $26 billion to address their liabilities in fueling the US opioid epidemic. It raises important questions about abatement conceptualization and measurement for allocating settlement funds among substate entities.

• We outline the political economy tensions undergirding the settlement and allocation, introduce an abatement conceptual framework, describe how an abatement formula was developed for Pennsylvania to allocate settlement funds, and summarize considerations for future settlement allocation efforts.

• Documenting the challenges and experiences of this task is essential to inform future efforts.

The impacts and costs of the opioid epidemic in the United States have been well documented in the popular press, in public health circles, and from local, state, and federal policy perspectives. ¹ , ⁶ A common response from states and municipal governments bearing the burden of addressing the opioid epidemic was to file lawsuits as plaintiffs alleging misrepresentations, fraud, and negligence by prescription opioid manufacturers and distributors, retail pharmacies, and individual providers and clinics. Plaintiffs claimed that the harms associated with prescription opioids were downplayed and even intentionally mischaracterized. ⁷ , ⁸ They also argued that these same parties did not meet Drug Enforcement Agency (DEA) requirements to monitor suspicious orders and investigate situations where opioids were diverted from appropriate clinical use to illegal distribution, sale, and use. ⁷ These claims sought remuneration for the cost of past damages, as well as the cost to remedy or abate the lasting impact of the epidemic moving forward. More than 3,000 US cities, counties, and sovereign tribal nations sued a similar set of manufacturers, distributors, providers, and pharmacies, and were subsequently “bound together” as part of Federal Multidistrict Litigation (MDL), through which parties can coordinate pretrial activities including settlements. ⁷ , ⁹

While from a strictly legal perspective, the civil claims brought by states’ Attorneys General (AG) and those brought by counties, cities, and sovereign tribal nations are separate and distinct by different entities, in practice they have become inextricably linked for purposes of moving meaningful settlements forward. Specifically, the industry defendants have offered financial payments and conduct reforms in exchange for the withdrawal of legal claims by almost all states and their political subdivisions. If full participation is not achieved, the total settlement would be reduced accordingly. ⁹ As a result, robust conversations have taken place between state AGs and their public health consultants with the various counties, cities, and other political subdivisions, about how to fairly allocate proposed settlement dollars across all entities. The goal of this allocation was to incentivize as many political subdivisions as possible to participate in the joint settlement. As this paper describes, the conversation between the state level constituents and the substate level entities has centered around the equitable distribution of settlement dollars across political subdivisions to address abatement needs and reduce the effects of the opioid crisis moving forward.

From a political economy perspective, the opioid settlement discussions between the state and substate level municipal entities are haunted by the historical precedent created through the 1998 Master Settlement Agreement (MSA) between the tobacco industry and forty‐six state AGs. The 1998 MSA funding remained under control of state‐level entities. ¹⁰ , ¹¹ Critics of this agreement note that funds from the MSA were not generally made available to local communities burdened with the impacts of tobacco use. ¹² Unlike the 1998 MSA, the opioid settlement proposals offered by industry were conditioned on the substate entities dropping their legal complaints. ¹³ These substate entities made it very clear they were unwilling to agree to the settlement unless they would be assured that a fair and equitable portion of the settlement dollars would flow directly to the substate level, rather than solely to the state, to address local impacts of the opioid epidemic. ⁸

As a result, parties in Pennsylvania agreed to a split of the settlement funds with 70% going to the counties for abatement programs, 15% to the Commonwealth for statewide abatement programs, and 15% to litigating subdivisions. ¹⁴ In addition, industry and plaintiffs have specified in the proposed settlement that monies must be spent to address the opioid crisis through a number of identified abatement activities, notably focused on future abatement. ¹⁵ Not only does this language prevent settlement funds from being used for non‐opioid crisis remediation purposes as was the case in the MSA, ¹⁵ but this also prohibits the use of the funds to compensate for past damages and instead requires funds be used to abate future problems.

After a six‐month period of discussions and negotiations, Pennsylvania met the terms of the proposed settlement and came to an agreement on how to allocate the $1.07 billion in settlement funds to be distributed to Pennsylvania over the next 18 years, including how those funds would be allocated to state and substate entities. ¹⁶ In the following sections, we use our vantage point as the independent analysts contracted as a technical assistance team by the Pennsylvania AG's office to assist claimants in Pennsylvania with research and analysis to inform their ultimate decision of whether to participate in the proposed settlement with the four industry defendants to describe:

1. A conceptual framework for relative abatement —or the prevention and remediation of impacts from the opioid epidemic—given that settlement funds were conditioned to be used for abatement purposes only;

2. How relative abatement needs were characterized and measured in a formula developed to determine each political subdivision's share of the 70% of overall settlement funds allocated to substate entities within Pennsylvania; and

3. Important considerations for stakeholders engaging in similar allocation efforts in the future.

An Abatement Conceptual Framework

Abatement in the legal context refers to ending or lessening a public nuisance, ¹⁷ , ¹⁸ in this case the impacts of opioid use and opioid use disorders. In the public health context, the concept of abatement of a particular condition (in this case, opioid use disorder) arises from a set of interventions known to end or lessen its impact in the future. ¹⁹ An abatement paradigm requires different methods than a damages paradigm. In a damages paradigm, the emphasis is on summing the cumulative societal costs of a problem from the start of its impacts to the present, and then identifying remuneration options to address those prior costs. However, in an abatement paradigm, the aim is identifying measures that capture current and future impacts of a problem, to distribute a finite sum of funding to meet future needs. The measures in damages and abatement paradigms may overlap, but their goals are distinct.

Therefore, we approached the task of conceptualizing and assessing relative future abatement needs by focusing on a framework that included prevention and remediation. Prevention refers to reducing future prescription opioid misuse (referred to as any use of prescription opioids outside their intended clinical use), illicit opioid use, and opioid use disorder (OUD). Remediation refers to the provision of societal services and infrastructure to address immediate treatment and intervention needs of constituents impacted by the opioid epidemic. This conceptualization treats a community as a patient that requires ways to stop the condition from (re)occurring (prevention) as well as immediate triage to address the existing condition and a pathway to healing (remediation). We then sought to identify relevant measures to create a formula to approximate the remediation and prevention needs in the future, which can also be viewed as a relative severity score of the opioid crisis, for each county. This formula determines the share of the settlement to be allocated for each county within Pennsylvania.

Characterizing and Measuring Abatement Needs

Choice of Measures

Estimating future abatement requires observing prior and existing patterns. Simple extrapolation of historical trends from past observations may not accurately reflect the future trajectory of the ongoing opioid epidemic considering its complex dynamics and interactions with many external factors. There are system dynamic models for projecting the epidemic under various intervention policies, but the detailed modeling results tailored to the counties in Pennsylvania are not directly available. ²⁰ , ²¹ , ²² The impact of the COVID‐19 pandemic further adds unpredictability and uncertainty of the future trends in substance use patterns. ²³ Additionally, measures vary substantially in their responsiveness of capturing the changes in substance use patterns. Some measures reflect the current changes in substance use outcomes such as overdose deaths, whereas others may have a long lead time before showing the effects, such as measures in prevention or prescription regulations. We therefore chose to focus on measures within a recent timeframe that would provide the best understanding of the potential future impacts of the opioid epidemic.

To explore appropriate measures for approximating remediation and prevention, we selected data using three criteria: availability, comprehensiveness, and efficacy. Owing to the limited time allotted to developing a formula that would need to be discussed and approved by diverse stakeholders for the settlement, we required publicly available data for easy access and transparency. We did not consider proprietary data (e.g., commercial health and Medicaid claims databases) because of the prolonged approval process of acquiring data and its non‐transparency for reproducing results. Since we developed the formulas to inform the allocation across subdivisions, we chose county‐level measures that were available for all counties or where missing county‐level data was minimal and could be reasonably imputed in order to ensure no substantial biases were introduced to rural counties because of missing data or incomplete records based on voluntary reporting in practice. Finally, we sought data that was as efficacious as possible to ensure that any data used would be accurate in what it was intended to measure and valid in capturing impacts.

The political economy of decision‐making is a key factor that is often overlooked in developing a measurement scale with diverse stakeholders. While we were assessing measures and data based on our criteria of availability, comprehensiveness, and efficacy, the stakeholders who would eventually decide whether to buy in had their individual motivations to include measures that better represented the needs of their constituents. Therefore, we were required to reframe our consideration of what the “best” formula would be, recognizing it may not necessarily include the measures with the highest quality by our criteria. Rather, the “best” formula would be the most practical set of measures that would assess relative abatement needs and that stakeholders believed would do the same. This required several discussions, reconsiderations of data and their sources, and compromise, which should not be overlooked in the decision‐making timeline. More details about all measures discussed in this paper are found in Online Appendix A.

Modeling Relative Abatement Needs

Since the total amount of settlement money is fixed, it is sufficient to generate the shares (or percentages) of the final allocation for each county from the final formula, which can also be viewed as a relative severity score for each county. To do this, for each selected measure, we calculated the share in each county relative to the total amount in the Commonwealth. We then averaged relative shares of each measure to create a total allocation share for each county. Measures were not always weighted equally. Unequal weights were determined based on data efficacy and relative importance through deliberation with stakeholders. Based on discussions with stakeholders, we developed multiple models for consideration within Pennsylvania. Three of these models are described below and summarized in Table 1: 1) a replication of the federal interstate allocation model, 2) a formula initially proposed by us, the technical assistance team, and 3) the final formula adopted that reflects the political economy considerations described above.

• Replicating the Interstate Allocation Model. The original interstate allocation (i.e., percentage to be allocated to each state) used in the MDL was based on a formula of four measures (calculated as the percentage representing the share of each state): amount of opioids dispensed (measured in morphine milligram equivalents or MMEs; 2006‐2014), ²⁴ prevalence of pain reliever use disorder (2015‐2017), ²⁵ overdose deaths (defined as the average share of opioid‐related overdose deaths and all drug‐related overdose deaths; 2007‐2017), ²⁶ and population of each state (2018). ²⁷ To replicate this at the substate level, we made several adjustments due to data availability. In particular, prevalence estimates of pain reliever use disorder are readily available from the public report of the National Survey of Drug Use and Health at the state level and substate regional level, but not at the county level. Therefore, we applied the same prevalence rate (%) of the substate region to all counties within this region, multiplied it by the county population to estimate the counts, and used the counts to determine the relative share of the number of individuals with pain reliever use disorder for each county (i.e., county counts divided by total counts). In addition, the years included for each measure differed slightly from the interstate calculation due to data availability.

• Proposed Formula . We proposed a two‐measure formula to stakeholders that included each county's percentage share of all drug‐related overdose deaths (2010‐2019), ²⁸ and total prescription opioids dispensed in MMEs (2006‐2014), ²⁴ weighted equally. Our team assessed these measures to be the most comprehensive, available, and efficacious for each county in Pennsylvania. MMEs are the most accurate county‐level measure of supply and are considered public information. All drug‐related overdose deaths capture the evolving nature of the opioid epidemic, which has transformed through prescription opioids, heroin, illicitly manufactured synthetic opioids (e.g., fentanyl), and an emerging polysubstance (involving stimulants) epidemics, and thus represent more inclusive and comprehensive measures of the impact of the ongoing epidemic. Using all drug‐related overdose deaths also side steps issues associated with unequal reporting of drug types in overdose death cases across counties. Conceptually, these measures provided the best understanding of the existing need for remediation, while also providing a future understanding of prevention potential.

• Final Formula. Stakeholder feedback on the proposed formula led to an expanded formula that included measures of naloxone administered using emergency medical services data, ²⁸ and OUD‐related hospitalizations using data from the Pennsylvania Health Care Cost Containment Council ²⁹ to account for opioid misuse that did not necessarily result in death. In addition, stakeholders advocated for more recent years of mortality data and adjustments to the MME measure (See Appendix A). ²⁴ We incorporated this feedback in the development of the final model, which included each county's share (%) of overdose deaths from all drugs (2015‐2019), adjusted MME of prescription opioids dispensed (2006‐2014), OUD‐related hospitalizations (2016‐2019), and EMS‐administered naloxone (2018‐2020). With its higher relative efficacy and completeness compared to the other three measures, we weighted all overdose deaths more heavily (40%) and the remaining measures equally at 20%.

Table 1.

Measures, Weights, and Data Sources used in the Different Allocations

Measure a : Source	Interstate Replication (Years)	Proposed Model (Years)	Final Model b (Years)
Overdose Deaths: Centers for Disease Control and Prevention (CDC) WONDER Multiple Cause of Death	22%c (2007‐2017)	50% (2010‐2019)	40% (2015‐2019)
Amount of Opioids Dispensed: The Automation of Reports and Consolidated Orders System (ARCOS)	34% (2006‐2014)	50% (2006‐2014)	20% (Adjusted) (2006‐2014)
Prevalence of Pain Reliever Use Disorder: The National Survey on Drug Use and Health (NSDUH) administered by the Substance Abuse and Mental Health Services Administration (SAMHSA)	22% (2015‐2017)
OUD‐Related Hospitalizations: Pennsylvania Health Care Cost Containment Council (PHC4) number of unique individuals hospitalized for any OUD‐related diseases			20% (2016‐2019)
EMS‐Administered Naloxone: Pennsylvania Department of Health number of naloxone (Narcan) doses administered by Emergency Medical Services (EMS)			20% (2018‐2020)
Population: US Census American Community Survey (ACS) 5‐Year Estimates	22% (2018)

^a The formulas use county share (%) for each measure

^b The top‐up approach described below was used in the final formula

^c Whereas the final and proposed models used overdose deaths from all drugs, the interstate replication used the average of all overdose deaths and opioid‐specific overdose deaths.

Introducing a Minimum Allocation to Promote Equity

As we reviewed relative severity scores from various models, some rural counties with very small populations had very low scores (< 0.15%). This would have resulted in a very small share of the total settlement allocation (e.g., far less than $1 million dollars over 18 years). We hypothesized that this may not represent a meaningful amount of resources required to address abatement needs in those counties. To protect very rural counties like these, we proposed that all counties would receive a certain minimum amount, regardless of the final allocation formula used. This could be done in two different ways, namely, a base minimum allocation and a top‐up allocation. In the base‐minimum allocation, each county is given a set amount (e.g., $1 million) before allocating the remaining funds using the proposed formula. Alternatively, a top‐up allocation provides any county that did not reach $1 million through the allocation formula with a “topping‐up” to a $1 million share and adjusts the allocation to the other counties accordingly. Both approaches slightly reduced other counties’ shares to ensure all counties received a $1 million minimum. The stakeholders chose the top‐up approach, which impacted 11 counties in total, all of which were rural.

Considerations and Recommendations

The process of conceptualizing and measuring abatement for purposes of allocating opioid epidemic settlement dollars in Pennsylvania yielded several important considerations, which may provide some useful and practical insights into similar future endeavors for other states or entities. We divide these into considerations and recommendations from data analytics, political economy, and evaluation perspectives, respectively.

From a data analytics perspective, data availability is a linchpin in the task of developing an allocation formula. Given that many measures that capture the impacts of the opioid epidemic are not systematically collected and publicly reported across all states, ³⁰ , ³¹ , ³² those engaged in this task must work swiftly to identify available data and seek access to restricted data that may be necessary. Priorities should be given to measures based on census data that are designed to be systematically collected state‐wide for maximum possible fair representation of all counties. Caution is needed when using voluntarily reported datasets, as they could potentially introduce unintended biases toward underrepresentation of rural and small counties and thus make these counties underresourced. In addition, it is essential to understand the past and current context of this epidemic when choosing proper outcome measures. For example, we argued that while the settlement was with opioid manufacturers and distributors, the evolving nature of this phenomenon has resulted in spillover into other drugs. Measuring only opioid‐related overdose deaths may not sufficiently reflect the true impacts of the epidemic; instead, a more inclusive overdose death measure including other drugs may be more comprehensive and appropriate. In addition, data analysts should consider keeping models simple and be aware of potential multicollinearity in their data. Measures related to the impacts of opioids can be highly correlated; in particular, we found county level population size to be highly correlated with most measures. Therefore, using a complex model of many measures may not provide substantial added value than a more parsimonious model, the latter of which is more digestible for a lay audience. Analysts should be clear with stakeholders about concerns with measurement error and offer recommendations of measure weighting based on the quality of the data sources being used. Analysts should also be prepared to offer multiple models—speaking to the strengths and limitations of each, knowing that individual interests may, in part, shape stakeholder choices. Finally, we recommend that model results be presented as shares (i.e., percentages of the total amount to be allocated) rather than dollar amounts in order to help facilitate more objective discussions with stakeholders.

From a political economy perspective, those engaged in discussions around measuring relative abatement in future settlements should acknowledge the tensions between developing a model with the most robust measures and developing a model that will yield the most buy‐in from stakeholders. Additionally, using base minimum and top‐up approaches may be one avenue for bringing smaller jurisdictions into the process with relatively small impact for larger jurisdictions. Finally, the historical context (e.g., the 1998 MSA) of a settlement discussion should not be overlooked, as it has potential influence on structures for administering settlement funding. Control and oversight of settlement funding will likely need to be discussed separately from allocation amount decisions, so that a settlement allocation can move forward rather than get mired in the confluence of these issues.

Relatedly, from an evaluation perspective, the shift from the 1998 MSA to the current opioid settlement in allowing local jurisdictions to decide how to allocate funds toward abatement raises a number of questions and opportunities. While the current opioid settlement provides provision for the use of funding to specifically address abatement activities, with a list of approved types of activities, there is no requirement to measure the impact of such activities on abatement of the opioid epidemic. Further, because of the decentralization of decisions around use of the settlement funds, there is no comprehensive strategy outlined for implementing evidence‐based abatement activities. Current federal and nongovernmental resources on evidence‐based activities (e.g., John Hopkins Principles for the Use of Funds from the Opioid Litigation, ³³ the Evidence‐Based Practices Resource Center through the Substance Abuse and Mental Health Services Administration, ³⁴ the Brandeis Opioid Resource Connector, ³⁵ the Community Opioid Resources Engine for North Carolina ³⁶ ) should be promoted more and expanded. Therefore, while local jurisdictions have control over the abatement activities funded through the use of the opioid settlement funds, it remains unclear if the various activities will be efficacious in abatement of the opioid epidemic. Failing to rely on evidence‐based approaches could result in less efficacious—if not countereffective—approaches, resulting in minimal abatement results for communities in need of prevention and treatment. There would be value in the federal government investing in research and evaluation on the effectiveness of the various abatement activities that are used in the coming years. How states require jurisdictions to report on their use of funds will be critical to monitoring effectiveness of various approaches to addressing abatement needs in the future. To assuage this concern in future settlements, approved types of activities should have an additional designation if they have a proven evidence base.

Conclusion

As the opioid epidemic continues in the United States, there will likely be additional settlements with industry to address the impacts of this crisis on the abatement needs of local communities. Important questions related to the conceptualization and measurement of abatement will face stakeholders who must determine how to allocate such settlement funds. In this manuscript, we have presented a case study that documents the settlement allocation work in Pennsylvania. This case study provides a foundation for future discussions on how to navigate similar efforts in a way that prioritizes both an accurate and fair allocation with inclusion of all subdivisions.

Funding/Support: The authors had a technical assistance contract with the Pennsylvania Office of the Attorney General.