Was there something in the weekly content that resonated with you or changed your perspective? If not, do you have thoughts about why not? Is there any part of the unit or previous

Original Investigation | Substance Use and Addiction

Effect of Exposure to Visual Campaigns and Narrative Vignettes on Addiction Stigma Among Health Care Professionals A Randomized Clinical Trial Alene Kennedy-Hendricks, PhD; Emma E. McGinty, PhD, MS; Amber Summers, PhD; Susan Krenn, BA; Michael I. Fingerhood, MD; Colleen L. Barry, PhD, MPP

Abstract

IMPORTANCE Stigma toward people with opioid use disorder (OUD) is pervasive in clinical settings, impeding delivery of high-quality care. To date, no study has evaluated the effect of different stigma-reduction messages or messengers among health care professionals.

OBJECTIVE To evaluate the effect of OUD-related messages delivered by different messengers on stigma and attitudes toward people with OUD among health care professionals.

DESIGN, SETTING, AND PARTICIPANTS This randomized clinical trial examined the effects of OUD-related messages delivered by a visual campaign alone or in combination with a written narrative vignette from the perspective of 1 of 3 messengers. Health care professionals in the US were recruited from 2 national online survey panels (Ipsos KnowledgePanel and SurveyHealthcareGlobus). A total of 1842 participants completed a web-based survey measuring stigma toward people with OUD from November 13 to 30, 2020.

INTERVENTIONS Eight groups were exposed to 1 of 2 message frames. One frame (Words Matter) emphasized the harm of stigmatizing language, and the other (Medication Treatment Works) focused on the effectiveness of medications approved by the US Food and Drug Administration for the treatment of OUD. Message frames were communicated through either a visual campaign alone or a visual campaign in combination with a written narrative vignette from the perspective of a simulated patient with OUD, a clinician, or a health care system administrator.

MAIN OUTCOMES AND MEASURES Dimensions of stigma toward people with OUD were measured on 5-point Likert scales that included items about desire for social distance from people with OUD, perception of individual blame for OUD, perspective of OUD as a medical condition, and support for increased governmental spending on OUD treatment. The level of warmth felt toward people with OUD was measured by a feeling thermometer (range, 0-100 points).

RESULTS Among 1842 participants, the mean (SD) age was 47 (13) years; 1324 participants (71.9%) were female, 145 (7.9%) were Hispanic, 140 (7.6%) were non-Hispanic Black, 1344 (73.0%) were non-Hispanic White, and 213 (11.6%) were of other non-Hispanic race (ie, individuals who did not self- report race as Black or White and did not self-report ethnicity as Hispanic). Compared with nonexposure, exposure to the combination of visual campaign and narrative vignette communicating the importance of nonstigmatizing language from the perspective of a patient with OUD was associated with a lower probability (difference, −16.8 percentage points, 95% CI, −26.1 to −7.4; P < .001) of unwillingness to have a person with OUD marry into the family (a measure of social distance preference) and a 7.2-point (95% CI, 3.2-11.1; P < .001) higher warmth rating. Participants exposed to the combined visual campaign and patient vignette about the value of medication

(continued)

Key Points Question Are specific communication

strategies effective in reducing stigma

toward people with opioid use disorder

(OUD) among health care professionals?

Findings In this randomized clinical trial

involving a national sample of 1842

health care professionals, exposure to

visual campaigns combined with short

narrative vignettes told from the

perspective of a patient with OUD that

emphasized the harm of stigmatizing

language or the effectiveness of

medications for treating OUD was

associated with reduced levels

of stigma.

Meaning The findings of this

randomized clinical trial suggest that

carefully designed communication

campaigns may reduce OUD-related

stigma among health care professionals.

+ Invited Commentary

+ Supplemental content

Author affiliations and article information are listed at the end of this article.

Open Access. This is an open access article distributed under the terms of the CC-BY License.

JAMA Network Open. 2022;5(2):e2146971. doi:10.1001/jamanetworkopen.2021.46971 (Reprinted) February 4, 2022 1/14

Downloaded From: https://jamanetwork.com/ on 08/07/2022

Abstract (continued)

treatment for OUD also had significantly lower levels of stigma compared with those in the nonexposed control group (eg, unwillingness to have a person with OUD as a neighbor: difference, −15.3 percentage points; 95% CI, −24.6 to −6.0; P = .001).

CONCLUSIONS AND RELEVANCE In this study, messages about nonstigmatizing language and effective medication for OUD reduced stigma among health care professionals. Stigma-reduction efforts targeting health care professionals may improve health care system capacity to serve people with OUD.

TRIAL REGISTRATION ClinicalTrials.gov Identifier: NCT05127707

JAMA Network Open. 2022;5(2):e2146971. doi:10.1001/jamanetworkopen.2021.46971

Introduction

Overdose mortality rates have worsened during the SARS-CoV-2 pandemic.1,2 Stigma toward people with opioid use disorder (OUD) has been an intractable challenge to ameliorating this problem.3-5

Defined through a sociological perspective, stigma occurs when the processes of labeling, stereotyping, status loss, and discrimination are enforced through power differentials.6 Stigma toward people with OUD and other substance use disorders is rooted in racism, classism, and other systems of oppression.7,8 In the health care system, stigma manifests at multiple levels.9 Expectation of stigma may reduce treatment seeking and engagement with care.10,11 Stigma enacted by clinicians may result in worse quality of care and exacerbate patient mistrust.4,5,12,13 On a structural level, stigma contributes to discriminatory policies and reduced investment in systems that support people with OUD.5,14 Although interest in stigma-related issues has grown,3 the evidence base for stigma- reduction strategies remains limited.4,15

Previous studies have reported a connection between language and stigma. Terms such as addict and substance abuser have been associated with greater stigma relative to person-centered language, such as person with a substance use disorder.16-18 Clinician use of stigmatizing language may translate to worse quality of care.13 However, stigmatizing language remains common in clinical and public discourse.19,20

In addition to discouraging use of stigmatizing language,21,22 communicating about solutions to OUD and overdose, including effective treatment, may mitigate stigma.4,23-25 Most health care professionals receive minimal training in addiction and its treatment.26 Health care professionals may be more aware that they are interacting with patients with OUD when these individuals display OUD symptoms and may be unaware that they are interacting with patients with OUD when these individuals are in recovery,12 which could distort the practitioner’s sense of the potential for recovery.

Despite the robust evidence base in favor of medication treatment, stigma has impeded broad availability of highly effective medications approved by the US Food and Drug Administration to treat OUD.27 Stigma toward the opioid agonists methadone and buprenorphine is related to the misperception that these medications replace 1 drug for another.28,29 Among clinicians, higher levels of stigma have been associated with less interest in prescribing medication for the treatment of OUD.26,30,31 Sympathetic narratives that highlight solutions such as effective treatment can help audiences empathize with a person with OUD.4,15,23,25,32 Communication strategies focusing on the effectiveness of medication in facilitating recovery may be particularly persuasive to an audience of health care professionals.

Health care professionals are frequently exposed to messages from patients, other clinicians, and health care system administrators. A health care professional may better relate to communication from a fellow health care professional but be more emotionally engaged when messages are delivered by patients. To our knowledge, no study has evaluated the effectiveness of

JAMA Network Open | Substance Use and Addiction Effect of Visual and Narrative Campaigns on Addiction Stigma in Health Care Professionals

JAMA Network Open. 2022;5(2):e2146971. doi:10.1001/jamanetworkopen.2021.46971 (Reprinted) February 4, 2022 2/14

Downloaded From: https://jamanetwork.com/ on 08/07/2022

different stigma-reduction messages or messengers among health care professionals. Understanding the ways in which different messages and messengers affect stigma and related attitudes among health care professionals is important to informing the design of effective stigma-reduction communication campaigns in health care settings. In this study, we enrolled a large national sample of health care professionals to assess the impact of 2 message framing strategies communicated through a visual campaign with and without an accompanying written narrative vignette from the perspective of a patient with OUD, a clinician, or a health care system administrator.

Methods

From November 13 to 30, 2020, we conducted a parallel-group randomized clinical trial involving 1842 health care professionals recruited from a blended sample of 2 online survey panels (Ipsos KnowledgePanel and SurveyHealthcareGlobus). The study was approved by the institutional review board of Johns Hopkins Bloomberg School of Public Health. All participants were members of 1 of the 2 online panels and provided consent to participate in surveys at the time of panel enrollment (with incentives for participation provided by the panel administrators). Therefore, this study was deemed exempt from the need for additional informed consent. The trial protocol was registered retrospectively because the original conception was a survey experiment (Supplement 1). This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline for randomized clinical trials.

Data The study sample included 1242 Ipsos KnowledgePanel members who self-reported their current employment in a health care profession. Ipsos uses probability- and address-based sampling to create an online panel comprising 60 000 members representative of the US population.33 Among 2703 KnowledgePanel members invited to participate in the study, 1484 completed the online survey, and 1242 qualified for study inclusion based on current employment in a health care profession (Figure).

To ensure greater representation of physicians and nurses, we supplemented the KnowledgePanel sample with 600 members of SurveyHealthcareGlobus, a health care market research firm with an opt-in online panel of physicians and nurses.34 Among the 111 105 SurveyHealthcareGlobus panelists invited to participate, 10 265 completed the online survey, and 600 qualified for study inclusion. Our sampling approach was informed by an effort to enroll a large diverse group of health care professionals from across the US. We did not focus on clinicians dedicated to treating OUD because stigma affects quality of care for patients with OUD across all health care settings.

Study Design Ipsos used a probability proportional to size sampling approach to select random samples corresponding to 8 exposure groups and 1 nonexposure control group. Each of the 9 groups included approximately 200 participants, providing adequate statistical power to detect medium-sized differences between groups. The 8 groups were exposed to 2 message frames, Words Matter and Medication Treatment Works, which were communicated through a visual campaign alone or together with a written narrative vignette from the perspective of 1 of 3 messengers: a simulated patient with OUD, a clinician, or a health care system administrator (eTable 1 in Supplement 2). Thus, for each message frame, exposed participants were randomized to 1 of 4 groups: visual campaign only, visual campaign plus patient vignette, visual campaign plus clinician vignette, or visual campaign plus administrator vignette.

The Words Matter message frame discouraged use of stigmatizing language related to substance use (including OUD) and encouraged use of nonstigmatizing language. The Medication Treatment Works message frame highlighted the value of US Food and Drug Administration–

JAMA Network Open | Substance Use and Addiction Effect of Visual and Narrative Campaigns on Addiction Stigma in Health Care Professionals

JAMA Network Open. 2022;5(2):e2146971. doi:10.1001/jamanetworkopen.2021.46971 (Reprinted) February 4, 2022 3/14

Downloaded From: https://jamanetwork.com/ on 08/07/2022

approved medications in facilitating recovery. These message frames were developed during a 2-day communication design workshop held in September 2019 involving 25 invited participants, including individuals receiving treatment for OUD, peer recovery coaches, nurses, physicians, administrators, communication experts, and researchers.

The visual campaigns for the Words Matter and Medication Treatment Works message frames both displayed headline text stating, “What we say and do matters for patients with substance use disorder,” with images of clinicians interacting with patients (eFigure in Supplement 2). The Words Matter visual campaign also displayed stigmatizing terms to avoid along with preferred alternative nonstigmatizing terms and invited participants to take a pledge committing to the use of nonstigmatizing language. The Medication Treatment Works visual campaign included text that stated, “Medications help patients recover and live full lives. Don’t let misperceptions get in the way. Learn more about methadone, buprenorphine (also called Suboxone or Subutex), and injectable extended-release naltrexone (also called Vivitrol).”

Participants randomized to receive exposure to the visual campaign in combination with a vignette viewed the visual campaign first, followed by a screen displaying the written narrative vignette. The vignettes were approximately the same length (165 words) and communicated the message frame from the perspectives of the 3 different messengers (patient with OUD, clinician, or health care system administrator) (eTable 2 in Supplement 2). For instance, in the patient vignette for

Figure. CONSORT Flow Diagram

113 808 Individuals assessed for eligibility 111 105 From SurveyHealthcareGlobus

2703 From Ipsos KnowledgePanel

819 Randomized to receive words matter message 819 Received words matter message as

randomized

1842 Included in the final analysis

822 Randomized to receive medication treatment works message 822 Received medication treatment works

message as randomized

201 Randomized to receive no message 201 Received no message as randomized

111 966 Excluded 110 505 From SurveyHealthcareGlobusa

100 840 Did not complete screening questions or survey 9665 Did not meet health care profession inclusion

criteria 1461 From Ipsos KnowledgePanelb

1219 Did not complete screening questions or survey 242 Did not meet health care profession inclusion

criteria

1842 Randomizedc

617 Received visual plus written narrative format 208 Received system

leader vignette 205 Received patient

vignette 204 Received clinician

vignette

205 Received visual format only

613 Received visual plus written narrative format 205 Received clinician

vignette 205 Received system

leader vignette 203 Received patient

vignette

206 Received visual format only

a Excluded if not working in health care profession at time of study. Nurses, physicians categorized as general practitioners or family doctors, and physicians categorized as specialists were included to achieve 600 participants, with approximately one-third represented from each category.

b Excluded if not working in health care profession at time of study or if working in a nonqualifying health care profession, such as dentist, dental hygienist, dental assistant, optometrist, veterinarian, veterinary assistant, or massage therapist.

c Ipsos KnowledgePanel used a probability proportional to size sampling approach to select 9 random samples.

JAMA Network Open | Substance Use and Addiction Effect of Visual and Narrative Campaigns on Addiction Stigma in Health Care Professionals

JAMA Network Open. 2022;5(2):e2146971. doi:10.1001/jamanetworkopen.2021.46971 (Reprinted) February 4, 2022 4/14

Downloaded From: https://jamanetwork.com/ on 08/07/2022

the Words Matter message frame, the patient described interacting with clinicians using language such as addict and feeling as though the clinicians “didn’t see me as a person” and “all they could see was my addiction.” In the clinician vignette, the clinician described using terms such as addict and learning that this language made patients feel as though “I didn’t see them as people” and “all I could see was their addiction.” In each narrative vignette, the patient, clinician, and health care system administrator evolved in their understanding of the harm of stigmatizing language, and the vignettes ended by noting that “health care professionals can be role models.” The Medication Treatment Works vignettes included similar content and language across the 3 different messenger narratives and shared themes with the Words Matter message frames, including an emphasis on health care professionals as role models. The Medication Treatment Works patient vignette featured a person recovering from OUD with the help of medication; the clinician and administrator vignettes described positive experiences with providing medication for the treatment of patients with OUD.

Measures After exposure to the visual campaign with or without an accompanying narrative vignette, all participants answered questions designed to measure stigma toward people with OUD (domain A), the primary set of outcomes. Participants in the control group answered the questions with no preceding exposure to either message frame. After completing questions for domain A, participants randomized to 1 of the 4 Words Matter groups answered questions about the appropriateness of various terms (domain B). Participants randomized to 1 of the 4 Medication Treatment Works groups answered questions about OUD medication treatment (domain C). Participants randomized to the control group answered questions in all 3 domains. For all participants, domain A questions appeared first; for the control group, the order of domains B and C was randomized. The order of questions within each domain was randomized for all groups to reduce risk of bias due to priming.

Domain A included items adapted from the General Social Survey stigma module35,36 and from published scales measuring stigma.14,30 These items assessed preferences for social distance (such as unwillingness to have a person with OUD marry into their family or be a neighbor),14,30,36,37

perspectives on the causes of OUD,14,30,36 support for increased governmental spending on OUD programs, and level of warmth felt toward people with OUD.38 Dimensions of stigma were measured using 5-point Likert scales, and levels of warmth were measured by a feeling thermometer (range, 0-100 points).

Domain B questions asked participants to rate on a 5-point Likert scale the extent to which they agreed or disagreed that language was appropriate to use in clinical settings. The language included 5 stigmatizing terms (addict, substance abuse, dirty and clean [in reference to drug test results], and addicted baby) and 5 alternative nonstigmatizing terms (person with substance use disorder, substance use, negative and positive [in reference to drug test results], and baby born with neonatal opioid withdrawal syndrome). We selected terms based on previous work suggesting that these terms elicited different levels of stigma.16-18 We created a stigmatizing terms scale (Cronbach α = .74) by summing the Likert scale responses to the stigmatizing terms and dividing by the total number of items. We used the same procedure for the alternative terms to create a nonstigmatizing terms scale (Cronbach α = .72). Participants in the Words Matter and control groups reported whether they would sign a Words Matter pledge committing to the use of nonstigmatizing language.

Domain C began by asking participants whether a treatment for OUD existed that was effective over a long period.39 This question was followed by a short explanation of the Food and Drug Administration–approved medications for the treatment of OUD. Participants then responded to adapted versions of several questions from domain A (substituting “a person taking medication to treat OUD” for “a person with OUD”) and questions about perceived effectiveness of OUD medication treatment26,30,31 (all survey questions are available in eTable 3 in Supplement 2).

JAMA Network Open | Substance Use and Addiction Effect of Visual and Narrative Campaigns on Addiction Stigma in Health Care Professionals

JAMA Network Open. 2022;5(2):e2146971. doi:10.1001/jamanetworkopen.2021.46971 (Reprinted) February 4, 2022 5/14

Downloaded From: https://jamanetwork.com/ on 08/07/2022

Statistical Analysis We used unpaired 2-sided t tests and χ2 tests to assess differences in observed characteristics across study groups. We dichotomized Likert scale responses (eg, agree and strongly agree were coded as 1, and other responses were coded as 0) and estimated logistic and linear regression models (for the feeling thermometer item and scales) to assess differences in responses between the exposure groups and the control group. In the models, we controlled for participant age, sex, prescriber status (physician, nurse practitioner, or physician assistant), and race and ethnicity (non-Hispanic White or non-Hispanic other race). We calculated predicted margins and marginal effects to display results for binary outcomes as predicted probabilities and changes in predicted probabilities. To correct for multiple hypothesis testing, we generated sharpened false discovery rate q values and identified estimates with q values less than .05 as statistically significant (with 2-sided P values reported to aid interpretation of significance).40 To estimate the marginal effect of the narrative vignettes on outcomes, we also assessed differences between groups exposed to both the visual campaign and a narrative vignette vs those exposed to the visual campaign only. In a sensitivity analysis, we estimated ordered logistic regression models using the nondichotomized Likert scale responses. Data were analyzed using Stata software, version 14 (StataCorp LLC).

Results

Participants Among 1842 participants, 1324 individuals (71.9%) were female and 518 (28.1%) were male. With regard to race and ethnicity, 145 participants (7.9%) were Hispanic, 140 (7.6%) were non-Hispanic Black, 1344 (73.0%) were non-Hispanic White, and 213 (11.6%) did not self-report race as Black or White and did not self-report ethnicity as Hispanic. Most participants were nurses (505 individuals [27.4%]) or physicians (467 individuals [25.4%]) (Table 1).

A total of 201 participants were randomized to the nonexposed control group, 206 to the Words Matter visual campaign–only group, 203 to the Words Matter visual campaign plus patient vignette group, 205 to the Words Matter visual campaign plus clinician vignette group, and 205 to the Words Matter visual campaign plus administrator vignette group. A total of 205 participants were randomized to the Medication Treatment Works visual campaign–only group, 205 to the Medication Treatment Works visual campaign plus patient vignette group, 204 to the Medication Treatment Works visual campaign plus clinician vignette group, and 208 to the Medication Treatment Works visual campaign plus administrator vignette group (Figure). We observed no significant differences in participant characteristics across study groups (eg, 133 participants [66.2%] in the control group, 585 participants [71.4%] in the Words Matter group, and 604 participants [73.5%] in the Medication Treatment Works group were female; 64 participants [31.8%] in the control group, 270 participants [33.0%] in the Words Matter group, and 249 participants [30.3%] in the Medication Treatment Works group were prescribing clinicians) (eTable 4 in Supplement 2).

Attitudes Toward People With Opioid Use Disorder In the control group, 142 participants (70.6%) reported unwillingness to have a person with OUD marry into their family, and 90 participants (44.8%) reported unwillingness to have a person with OUD as a neighbor (Table 2). Most participants in the control group disagreed that people with OUD had only themselves to blame for the problem (118 individuals [58.7%]) and favored increased spending on OUD treatment (131 individuals [65.2%]). The mean (SD) level of warmth toward people with OUD was 50.7 (19.1) points on the feeling thermometer scale.

Exposure to the Words Matter and Medication Treatment Works visual campaigns alone was not associated with statistically significant differences from the control group (eg, unwilling to have a person with OUD marry into the family: difference, 0.3 percentage points [95% CI, −8.5 to 9.1; P = .94] in the Words Matter group and −5.3 percentage points [95% CI, −14.3 to 3.7; P = .25] in the Medication Treatment Works group). However