A State Financial Incentive Policy to Improve Emergency Department Treatment for Opioid Use Disorder: A Qualitative Study

Financial incentives have not been widely implemented to improve and expand the treatment of patients with opioid use disorder, although these incentives have been shown to be effective (1, 2). Value-based payment models are emerging to support substance use disorder treatment in the form of pay for performance, health homes, and accountable care organizations (2–8). Incentives for clinicians have increased prescribing of medication treatments, such as buprenorphine (9). However, hospitals have not received inducements to provide or expand access to treatment for opioid use disorder.

Hospital emergency departments (EDs) not only care for patients with overdose and other complications from opioid use but also serve as vital touch points to engage patients into longer-term addiction treatment (10–14). After an overdose, patients are at risk for repeat overdose and death (15, 16). Initiation of buprenorphine in the ED improves patients’ health outcomes and retention in treatment (17, 18). Patients also benefit from navigation and counseling during the vulnerable transition following hospital discharge, an approach described as a “warm handoff” (19, 20).

Policy makers have recognized that strengthening the linkage from the ED to treatment is an opportunity to combat the opioid epidemic (14). Patients with opioid use disorder engage in follow-up treatment at low rates after emergency care, including low rates of treatment initiation with opioid medications such as buprenorphine (21–24). Several states have developed guidelines, regulations, and initiatives to facilitate linkage to such treatments (25). For example, Massachusetts mandates that hospitals provide specialized evaluations for patients visiting the ED because of opioid-related illness and arrange transitions to long-term treatment for patients who express interest in such treatment (26). In contrast, Pennsylvania chose to create a financial incentive program for hospitals (27). After an 80% increase in ED encounters for opioid overdose in preceding years, the Department of Human Services (DHS), in collaboration with the Hospital and Healthsystem Association of Pennsylvania (HAP), established the Opioid Hospital Quality Improvement Program (O-HQIP) in 2019 (13, 28). The program seeks to increase the rate of follow-up treatment for Medicaid patients within 7 days of an ED encounter for opioid-related illness.

The O-HQIP comprises two phases (28). The first phase offers a one-time process incentive for hospitals that attest to implementing four distinct treatment pathways. The pathways are initiation of buprenorphine treatment during the ED encounter, warm handoff to outpatient treatment, referral to treatment for pregnant patients, and inpatient initiation of methadone or buprenorphine treatment. The full incentive was paid in 2019 and was contingent on participation in all four pathways, with stepwise payments for partial participation: $25,000 for one pathway per hospital, $62,000 for two, $108,000 for three, and $193,000 for all four, although the final amounts were higher because of outstanding funds allocated for this purpose. No penalties were associated with this program. The second phase of the program, begun in 2020, consists of annual performance incentives as determined through Medicaid claims analysis. All Pennsylvania hospitals are eligible for performance incentives regardless of participation in the first phase.

Our objective was to evaluate implementation among hospitals of this state policy to create financial incentives focused on opioid use disorder treatment. We examined how hospitals made the decision to participate in the first phase of the O-HQIP, with specific attention to barriers and facilitators for implementing treatment pathways. We used qualitative methods to examine operational, cultural, and financial influences on the implementation of this program in order to gain insight into how such incentives may be modified and extended to other settings.

Methods

Study Sample and Participant Recruitment

We conducted semistructured telephone interviews with hospital leaders across Pennsylvania. We obtained a broad representation of hospitals in the state with regard to key characteristics, including location, size, and health system affiliation. We purposively sampled study hospitals from a publicly available, comprehensive list of Pennsylvania hospitals, stratified by participation status (28). We excluded pediatric, specialty, and federal hospitals. We obtained data on hospital characteristics from the Pennsylvania Department of Health. Hospitals varied with respect to their level of participation in the one-time process incentive; hospitals were either full participants (i.e., they had adopted all four treatment pathways), partial participants (adopted between one and three), or nonparticipants (which declined to adopt treatment pathways).

Interview participants were selected among hospital and health system leaders who had direct involvement in the decision whether to participate in the program as well as in its implementation. These stakeholders varied among institutions and included chief medical and operating officers, directors of behavioral health, addiction medicine specialists, ED chairs, and physician leaders. Study hospitals nominated potential interview participants, who we then approved for an interview after reviewing their role at the hospital; in two instances, we requested an alternative participant who had greater knowledge and awareness of the O-HQIP. Some interviews included multiple participants, although primary respondents were specified in advance. Initial recruitment inquiries were conducted by e-mail with assistance from HAP. Telephone interviews were conducted from April to June 2019, following the participation deadline for the process incentive (March 2019). A single study author (A.S.K.) conducted all interviews, whose mean length was 48 (range 29–74) minutes.

Interview Procedures

We developed the interview guide by using the Consolidated Framework for Implementation Research (CFIR) (29, 30). The CFIR model draws from multiple evidence-based theories to offer a unified and practical framework for organizing various influences on implementation (i.e., constructs) (29). The primary goal for the interviews was to understand the willingness of organizations to engage in the O-HQIP financial incentive program, with a focus on barriers and facilitators to implementing opioid treatment pathways.

We designed open-ended interview questions to allow for inductive examination of the decisions, experiences, and processes in hospitals that were considering O-HQIP participation. Questions were framed according to the CFIR constructs identified as most relevant to study objectives, including implementation climate, culture, and readiness for implementation (inner setting); design quality and packaging (intervention characteristics); planning, engaging, and executing (process); and external policies and incentives (outer setting) (30). We also included questions to deductively examine specific a priori hypotheses for potential influences on O-HQIP participation, after a literature review and a quantitative analysis of hospital participation in the program (27). We revised interview questions iteratively after piloting the guide within the University of Pennsylvania Health System and after each of the first three study interviews.

We assessed saturation after each set of three interviews through review of the transcript and preliminary notes. Saturation was defined as informational redundancy—or the point at which new data were redundant with previously collected data (31, 32). We determined that saturation was achieved after 15 interviews; five additional interviews were completed to ensure broad representation of hospitals in the sample. Data saturation was assessed across all hospitals, regardless of participation status.

Analysis

The interviews were recorded, transcribed verbatim, lightly edited for clarity, and entered into NVivo, a qualitative analysis software package. We created a preliminary codebook, with individual codes representing specific CFIR constructs that were included in the interview guide (30). Using a qualitative content analysis approach, two authors trained in qualitative techniques (S.F.L. and J.D.) used NVivo to systematically code the interviews (33–35). The first five transcripts were independently double-coded, with >95% agreement on each transcript. We discussed any discrepancies and refined the codebook through an iterative process that added emergent codes, standardized code definitions, or removed unnecessary codes. After the remaining interviews were coded, interview responses with attributed codes were extracted and analyzed in discussion by the study team. We then assembled key patterns and themes through a consensus process. Analysis was pooled for hospitals with varying participation status, but we also examined trends in themes according to participation status in an exploratory analysis. The institutional review board at the University of Pennsylvania approved this study. While conducting this study, the study team followed the Consolidated Criteria for Reporting Qualitative Research (36).

Results

Characteristics of Participating Hospitals

We conducted 20 interviews across various hospitals and health systems in Pennsylvania; the hospital characteristics are presented in Table 1. Study hospitals varied with respect to key characteristics, including location, size, profit status, and health system affiliation. Seven hospitals were full O-HQIP participants, seven were partial participants, and six did not participate. The roles of specific interview participants are shown in Table 2. Interviewees held different positions in their hospitals, but all had a direct role in adoption and implementation of the opioid treatment pathways.

Key Themes

Table 3 summarizes six key themes that emerged from the interviews and presents representative quotations. These themes were mapped to specific CFIR constructs that were used as the framework for the interview guide and data analysis. (A table in an online supplement to this article shows how key themes tended to differ by hospital participation status.)

Resources and Community Partnerships

Implementation of the O-HQIP pathways required resources—specifically, trained staff to counsel patients and coordinate follow-up care. The presence of in-person recovery specialists or case managers, although not specifically mandated by the program, was thought to be effective in ensuring successful care transitions. As the interviewee from hospital 5, a partial O-HQIP participant, noted, “Having [certified recovery specialists] sitting in the ED, and in other cases on call, available to just step in with those cases, is why folks think positively about the warm-handoff program, because there is someone trained in how to deal with them.”

However, most hospitals—specifically, smaller or independent hospitals with lower volumes of patients with opioid use disorder—were unable to justify investing in these resources internally. Some hospitals noted resources as a barrier to participation, despite the incentive payments: “We would have to add additional ED staff to manage a small population. We already staff that department pretty lean. We really don’t have the resources. I know that there is funding, but I don’t know if that is worth being spent on a small segment of our population” (hospital 10, nonparticipant).

For several hospitals, the strategy for overcoming resource limitations was through partnerships with external community organizations. These relationships were cited as necessary for hospitals to participate in the O-HQIP. Two key characteristics described these community partners: geographic proximity and proactive engagement with patients. For example, effective partners could provide in-person consultation to patients. Community partners could guarantee destinations for patients so that they could receive care after hospital discharge, reassuring providers who were interested in prescribing buprenorphine but who lacked experience with this treatment. “Our relationship with a behavioral health organization that is literally steps away from our ED is really key to our success. We like to work in partnership with them because they are really the experts” (hospital 20, partial participant).

In many cases, community partners were those that had been designated as single-county authorities or centers of excellence by the state of Pennsylvania; these organizations receive separate funding to support opioid treatment and assume local responsibility for guiding patients to care. Although some hospitals described these external services as robust, the quality and alignment of services varied. Hospitals and health systems that provided internal addiction services often pursued external grant funding, which created concerns about sustainability.

Organizational Priorities and Funding

All hospital representatives expressed interest in developing initiatives to improve treatment for opioid use disorder, but some had to balance participation in this program with competing priorities, including other opioid-related initiatives, such as improving prescribing behavior. “I think the goal of the program is great. But jumping to, ‘We’re going to focus on getting people from the hospital into therapy’—I didn’t think we were ready to start. This was putting the cart before the horse . . . we had to think about prescribing [first]” (hospital 4, nonparticipant).

For many hospitals, the O-HQIP focused organizational attention on this specific issue. Many leaders interpreted the policy as an endorsement from the state that this problem was urgent and should be prioritized. In many cases, the existence of the policy allowed internal champions to proceed with existing plans that had yet to be implemented. “I thought [linking patients to treatment] was important to be doing anyway. But because of the lack of resources, it was hard to get momentum to really push forward. Not that people were resistant. It’s just that people have jobs. It’s hard to fit it in to everything else you’re required to do. . . . Our chief quality officer was on board very quickly, and then the financial benefits brought along the chief financial officers” (hospital 11, full participant).

In most cases, the financial incentive provided additional momentum for efforts to improve treatment access for opioid use disorder. However, pledged incentives were not incorporated into hospital budgets. More than the specific amount of funding, the fact that the program was tied to any funding encouraged hospital participation. “It was something that we were going to do anyway in terms of starting the program. But I felt like any time there’s a financial incentive . . . the administration would be more likely to support our projects, so I thought that would be helpful. It was financial reinforcement of the original motivation for what we wanted to do” (hospital 15, partial participant).

Buprenorphine Practice Changes

One treatment pathway focuses on treatment initiation with buprenorphine by an ED provider. All partially participating hospitals chose not to implement this specific pathway. Although there was interest in introducing buprenorphine to the ED setting, several participants noted operational and cultural barriers, including physician training. Buprenorphine prescribing requires special training and approval of a specific waiver from the Drug Enforcement Agency (DEA), known as the DATA 2000 or X waiver. As a physician leader noted, “There’s not an ED that doesn’t want to have [crisis recovery specialists], and yet when you go to them and say, ‘We can train your doctors to prescribe buprenorphine, it will be free—just bring them in,’ nobody shows up. . . . We need more education, and yet you provide the education, and nobody [takes advantage of it]” (hospital 5, partial participant).

Participants noted perceived concerns that buprenorphine prescribing would increase the use of the ED as a primary site for addiction treatment. As one department chair said, “I took the lead, started prescribing [buprenorphine], and these people didn’t come back the next day or the day after that. I was able to show my [staff] this, and now maybe half of my [providers] are at least comfortable giving [buprenorphine] in the ED. I’m the only one who is X-waivered, though” (hospital 3, full participant).

Hospital leaders noted that stigma remained a barrier to changing practice among the clinical staff, including ED physicians and nurses. A director of behavioral health said, “The emergency physicians are not starting [buprenorphine]. They’re totally philosophically against that, actually. . . . They will use [benzodiazepines] for these patients, but they will not start [buprenorphine]” (hospital 8, nonparticipant).

Coordinated Planning and Champions

Hospital leaders described an extensive process of planning treatment pathways when considering whether to participate in the program. Planning required involvement of multiple stakeholders, including representatives from the ED, behavioral health, addiction medicine, pharmacy, obstetrics, hospital medicine, nursing managers, and hospital administration, including financial officers. Health systems with multiple hospitals also engaged in planning across different facilities. In most cases, hospitals attempted to balance consistent implementation across the health system with the particular circumstances of individual hospitals and local communities. Many health systems decided to collectively participate at a certain level but allow for local adaptation. “Our goal at the beginning was that all hospitals would be able to attest to all four [pathways]. Also, our approach was to go to each hospital ED and hospital leadership to identify at least one or more champions from their facility [who is] able to participate, so we could discuss implementation, barriers, share best practices, what worked in one place, and why that would or would not work in other places” (hospital 11, full participant).

As part of planning, identification and engagement of internal champions were essential to adopting and launching the pathways. These champions often had preexisting interest and expertise. As a chief medical officer stated, “You need staff, and we did—because we had a nurse champion that took this [program] and ran with it. I could see other places that don’t have that champion could have some struggles” (hospital 20, partial participant).

Program Design and Clarity

Nonparticipating hospitals cited difficulties in understanding program requirements as a barrier to adopting the O-HQIP pathways. Even partially and fully participating hospitals highlighted challenges to understanding program requirements when they were initially announced. Although most questions were resolved through communication with DHS, when HAP acted as key mediator, many hospitals reported that the time between the program announcement and the participation deadline was not sufficient to prepare for pathway implementation. Some hospitals appreciated that the open-ended program design allowed for adaptation to local circumstances, but some also stated that the lack of details or model pathways generated uncertainty. “I think it was open ended purposefully, probably to allow some innovation and local implementation. But there were just some details that weren’t entirely clear to us” (hospital 5, partial participant).

Technology and Data Collection

Nearly all hospital leaders underscored the need to employ technology to implement O-HQIP pathways and to collect data on their effectiveness. Some hospitals used more advanced approaches to implement pathways through the electronic health record (EHR) and to generate reports on patient outcomes. But most hospitals reported difficulty in collecting accurate data on whether patients could actually obtain follow-up treatment. The main reason for this difficulty was that patients obtained follow-up care at community sites external to the health system, without interoperability between the sites and the system. Some hospitals attempted to manually collect external data, as described by one director of toxicology: “When you refer somebody [to a site] outside the system, there’s no great way of knowing if that individual engaged. It becomes an active process. In my county, I’m working to make phone calls—and educating my providers on how to do [the referral], including social workers” (hospital 11, full participant).

Many hospital leaders expressed the desire to receive outcomes data from the state, noting that more frequent or even real-time feedback would be useful in the implementation and sustainability of treatment pathways. “I’m kind of in the dark as to whether we’re meeting our goals or not. It would be nice if we could get some type of a quarterly feedback or report of, ‘You have these many patients,’ and ‘These many patients got their treatment’” (hospital 14, partial participant).

Discussion

Financial incentives and other value-based approaches have the potential to drive swift system and practice changes needed to intervene in the opioid epidemic. Our study examined hospital participation in a voluntary financial incentive program in Pennsylvania, O-HQIP, to understand why hospitals accepted or declined participation in this program. The themes we identified reflect the willingness and ability of hospitals to enact evidence-based practices, as well as the many internal and external influences on implementing treatment pathways.

Other states have elected to follow other policy approaches to improving follow-up opioid treatment (25). Massachusetts mandates that patients receive a specialized substance use evaluation within 24 hours of presenting for emergency care (26). Rhode Island imposes stricter rules, requiring this evaluation before ED discharge, in addition to the use of evidence-based protocols for discharge planning and coordinating with outside providers (37). Pennsylvania is unique in establishing a voluntary incentive program that specifies goals for hospitals but, by design, allows hospitals to adapt to local practice, resources, and environment (28). The effectiveness of all of these strategies for improving patient outcomes has yet to be evaluated, which will require quantitative analysis of patients’ health outcomes in relation to pathway implementation.

The O-HQIP program exclusively awards funds to hospitals. However, many hospitals required that external community partners, which were not eligible for funding through this program, accept receipt of patients for ongoing treatment. We found that implementation of pathways required robust outpatient services, which in Pennsylvania had been organized and supported through separate initiatives (38, 39). Larger hospitals and health systems maintained internal resources unavailable to small or independent facilities. In the absence of internal or external resources, some hospitals could not participate in the program. Future programs may consider gain-sharing or dual support for referring and receiving facilities. In addition, different approaches may be needed for rural communities, including support for substance use treatment in primary care or telemedicine (40, 41).

We found that few hospitals anticipated incentive payments in their operating budgets. The specific payment amount may not have been as important as the prioritization of treatment access—among competing priorities—tied to funding through O-HQIP. Hospitals were often on the cusp of change and responded to this nudge to prioritize opioid treatment access. Some hospital leaders expressed uncertainty about whether they could successfully implement O-HQIP pathways but chose to participate nonetheless, given the absence of penalties for not meeting follow-up metrics. Yet the lack of risk was not enough to motivate all hospitals to participate. Although some hospitals misunderstood program requirements and therefore declined to participate, others expressed a lack of readiness for change. Alternative or delayed inducements might be appropriate for hospitals in a different phase of readiness. Future studies may assess alternative approaches that were not explicitly addressed in this study, such as different incentive amounts and strategies that incorporate behavioral economic interventions (including loss aversion, relative social ranking, and goal gradients) (42).

A common obstacle to full O-HQIP participation was the requirement to initiate buprenorphine in the ED. Barriers to buprenorphine administration and prescription have been well described (43–45). Buprenorphine prescription requires a DEA X waiver, and stigma against medication treatment likely influences providers’ willingness to offer it to patients. However, buprenorphine (as well as methadone maintenance therapy) has been consistently shown to be the most effective treatment modality for opioid use disorder (46–49). This study’s results support previous findings by Hawk et al. (45) that implementing buprenorphine requires time-intensive training for physicians, in addition to concerns regarding nursing education, pharmacy availability, establishment of clear plans for ongoing care, and effects on patient throughput. Fully participating hospitals that implemented buprenorphine could prioritize this approach over competing needs, develop efficient pathways, arrange provider training, and ensure linkage to care. Partial participants, however, were only willing to connect patients with external buprenorphine providers but were not yet willing to change the practices of their own providers. Given that evidence supports improved treatment retention after ED initiation of buprenorphine rather than referral for initiation, incentives are needed that overcome the multifaceted barriers for this key treatment modality (14).

Finally, challenges related to data collection affected both willingness to participate and implementation of treatment pathways. For other pay-for-performance initiatives, many hospitals monitor performance internally, using EHR-derived reports. However, linkage of patients to treatment often required feedback from community partners without data-sharing infrastructure. Sharing of data for substance use treatment is also restricted under regulations for protected health information. Without active feedback, hospitals could not determine whether O-HQIP pathways were successful, make iterative changes, or use behavioral techniques, such as report cards, to further motivate providers.

This study had several limitations. First, it was qualitative, using semistructured questions that were subject to response bias because interview participants were leaders of their respective organizations. Second, the study had the potential for selection bias among the individuals and organizations that elected to be interviewed, although we attempted to obtain a broad range of perspectives from different hospital types. Third, a related limitation was that because of the small study sample size, we were unable to fully stratify the analysis by the full range of hospital characteristics, although an exploratory analysis considered key themes related to hospital participation status. Fourth, we interviewed stakeholders with varied roles within their organizations, generating diverse perspectives but potentially creating inconsistency. Finally, interviews were conducted in the 3 months immediately after the deadline for hospitals to participate in the process incentive. This timing was purposeful in that all hospitals had declared their participation status but created the potential for recall bias.

Conclusions

The challenge of solving the opioid epidemic requires bold policy making. Value-based payment models, including financial incentives, will become essential to overcoming knowledge-to-implementation gaps in best practices. Pennsylvania created the first financial incentive program to encourage hospitals to facilitate linkage from hospital EDs to sustained outpatient care. Participation in this program and implementation of treatment pathways offer key lessons for future iterations of opioid-centered incentive programs. Future work is needed to determine the effectiveness in improving patient outcomes not only for the O-HQIP initiative but also for emerging delivery models, regulatory changes, and payment innovations in the care of patients with opioid use disorder.