Effect of a Peer-Led Behavioral Intervention for Emergency Department Patients at High Risk of Fatal Opioid Overdose

Key Points Question Can peer-led behavioral interventions in the emergency department for patients who have had a recent opioid overdose increase short-term treatment engagement after the emergency department visit? Findings In this randomized clinical trial of 648 emergency department patients at high risk of opioid overdose, there was no difference in treatment engagement within 30 days of the visit for participants who received a peer-led intervention vs those who received a standard behavioral intervention by a clinical social worker (32% vs 30%). Meaning An emergency department–based behavioral intervention is likely effective in promoting treatment engagement, but who delivers the intervention may be less influential on short-term outcomes.

number of certified peer navigators by 50% starting in 2018. The Director of BHDDH, Rebecca 149 Boss, has affirmed that "if the proposed trial demonstrates peer-based ED navigation to be 150 superior to current standards of care for patients admitted to the ED for an overdose, we would 151 continue to expand this program throughout hospitals across Rhode Island" (see letter of 152 support). Finally, we will work with the Laura and John Arnold Foundation and other partners to 153 broadly disseminate our finding to other states, many of which have already adapted the Rhode 154 Island peer recovery model. 53 155 156 B. Innovation 157 B.1 Rhode Island is a national leader in developing highly innovative programs for 158 overdose prevention. The state has established national models for overdose prevention and 159 intervention, and has encouraged community-level innovation. In response to the growing 160 overdose crisis and with funding from BHDDH, in 2014, the community-based organization 161 Anchor Recovery Community Center launched a new initiative, AnchorED. Anchor Recovery 162 Community Center (ARCC) is Rhode Island's first and only community organization run by and 163 dedicated to those living in recovery from substance use disorders. Focusing on a peer-to-peer 164 support model, ARCC helps those living with substance use disorders maintain long-term 165 recovery, improve social connectedness, improve self-sufficiency, and live healthier lives. The 166 AnchorED program takes the peer model into the ED, dispatching trained and certified peer 167 navigators to EDs throughout RI to provide recovery support and services navigation to 168 individuals who have experienced an opioid overdose (see additional details in section C.3). 169 Importantly, AnchorED is the first program of its kind in the nation. 54 Despite a paucity of 170 evidence to demonstrate effectiveness and improved long-term outcomes for overdose 171 survivors, jurisdictions in New York, Connecticut, Massachusetts,New Hampshire,and New 172 Jersey are in the process of creating programs based on the Rhode Island peer model. 55 Given 173 the intense interest in peer recovery models in settings across the nation, the results of 174 our study have a strong potential to fundamentally shift clinical practice paradigms for 175 treating overdose patients in the ED. 176 B.2 A data-rich policy and public health environment to boost study rigor. In 2015, 177 Governor Gina Raimondo signed an executive order to establish the Rhode Island Overdose 178 Prevention and Intervention Task Force. This task force charged experts (including Dr. Marshall,179 Co-PI) to develop a strategic plan that would guide efforts to tackle the state's overdose crisis. 56 180 An important outcome of this plan was a comprehensive data-sharing framework between key 181 state agencies and academic researchers. 57 This data sharing agreement offers unprecedented 182 access to population-based overdose morbidity and mortality data. As a result, our team has the 183 capacity to link Navigator Trial patient data with multiple statewide administrative databases 184 (e.g., medical examiner case files, behavioral health data on treatment admissions) to ascertain 185 objective outcome data. Importantly, by linking patient data to these datasets, we will be able to 186 assess objective primary and exploratory outcomes (e.g., engagement in treatment, overdose 187 death) for all study participants. The ability to conduct robust data linkages is an important 188 methodological innovation of our proposed project, and one that overcomes the limitations 189 frequently present in other observational and experimental studies of overdose patients (e.g., 190 biases related to self-reported outcomes and loss to follow-up). Finally, management of these 191 datasets and support for the state's electronic overdose surveillance system 192 (http://www.PreventOverdoseRI.org) are funded by the CDC; thus, the proposed project 193 leverages existing infrastructure and administrative data to reduce overall study costs. 194 195 C. Approach 196 C.1. Study overview, setting, and target population. We will compare the effectiveness of 197 peer navigation versus a standard social work intervention delivered in the RIH and TMH EDs 198 following an ED visit for opioid overdose. 650 patients treated for an opioid overdose or at risk 199 for an opioid overdose (defined below) will be randomized to receive either Social Work or Peer 200 Navigator services (n=325 per arm) and followed using administrative datasets. Our primary 201 outcomes will be: (1) engagement in treatment within 30-days after the ED visit, and (2)  202 recurrent ED visit(s) for opioid overdose over the 18-month follow-up period. Additionally, we will 203 also enroll a "no intervention" self-selected control group (n=325) at both hospitals to assess the 204 impact of the two interventions from the RCT (peer navigation OR social worker intervention) 205 against no intervention. A total of 975 (325 receiving Peer Navigator services, 325 receiving 206 Social Work services, and 325 who declined receiving either of these services) subjects will be 207 enrolled in this study. We decided to allow patients to self-select into the "no intervention" 208 control group. Given that an offer of one of the two behavioral interventions has become the 209 standard of care for overdose visits in our ED, we believe that it would be unethical to randomly 210 assign individuals to a "no intervention" control condition. We considered using another  based ED as the control condition, but other EDs across RI have already adopted our program 212 of offering behavioral intervention following overdose. We will use quasi-experimental statistical 213 analyses to control for selection bias and confounding that could arise from non-random 214 assignment of the controls. The two sites where this investigation will take place are Rhode 215 Island Hospital and The Miriam Hospital ED in Providence (RIH being Dr. Beaudoin's primary 216 practice site). RIH is an academic, tertiary care, level one trauma center with >105,000 annual 217 ED visits, that serves a heterogeneous and demographically diverse patient population. The 218 RIH ED cares for the majority of opioid overdoses throughout Rhode Island. In 2017, RIH has 219 recorded an average of 85 opioid overdoses per month (~1,000/year). As such, RIH is an ideal 220 site for the proposed study. 221 We will recruit adult ED patients in both hospitals who are: (1) being treated for an opioid 222 overdose, or (2) have had an opioid overdose in the past 12-months (identified by self-report 223 during screening or in review of the EMR); or (3) are presenting with a visit related to illicit 224 injection opioid use (e.g., cutaneous injection-related infection, opioid withdrawal, endocarditis).

225
We are specifically targeting patients with a current or recent opioid overdose and those who 226 inject opioids illicitly, as they are at highest risk for opioid overdose and death. 4,58 227 C.2. A highly experienced research team will ensure this project's success. This proposal 228 leverages thousands of hours of work by the investigative team in conducting clinical trials and 229 observational studies (see biosketches). For example, co-PI Dr. Beaudoin has an outstanding 230 track record of leading studies that recruit patients in a busy ED environment, overseeing 231 screening evaluations and assessments for thousands of patients, and successfully conducting 232 follow-up in a challenging population of patients with addiction. Dr. Marshall is a nationally 233 recognized expert in substance use epidemiology and, as the director of the state's overdose 234 surveillance system, www.PreventOverdoseRI.org, brings a wealth of experience in the 235 management and analysis of overdose-related patient data. In sum, Drs. Marshall and Beaudoin 236 have the complementary expertise necessary to achieve the aims of this project and are 237 surrounded by a team of multi-disciplinary co-investigators and consultants with expertise in 238 addiction, behavioral psychology, and peer navigation 239

C.3. Overview of behavioral interventions available at the Rhode Island Hospital and The 240
Miriam Hospital ED. Currently, standard of care is to receive a behavioral intervention, either 241 peer navigation or social work. Whether a patient sees a social worker or peer navigator varies 242 and can depend on the time of day, availability, and preferences or biases of the provider or 243 patient. Once funded, we have permission and explicit support from the peer and social work 244 programs to randomize patients to one of the two interventions (see letters of support). 245 Peer Navigators: The Peer Navigators of the Anchor Recovery Community Center (ARCC) of 246 Rhode Island will deliver the peer navigation arm of the intervention. They are available 24 247 hours a day, 7 days a week to provide recovery support, referrals, and ongoing engagement for 248 ED patients after discharge. They arrive in the ED within 30 minutes of consultation, assess 249 individuals for readiness to seek treatment, provide linkage to treatment, and educate on 250 overdose prevention and response, including naloxone administration. Following the ED visit, 251 peer navigators follow up with patients within 24-48 hours and maintain contact and services 252 navigation for at least 90 days following ED discharge, on a weekly basis and more frequently 253 as needed.

254
To become a peer navigator, applicants must be in long-term recovery (≥2 years) and 255 undergo a rigorous training program. The peer navigators learn motivational interviewing 256 techniques and the transtheoretical model of behavior, also known as the "stages of change". 59 257 Motivational interviewing is considered to be a patient-centered, non-judgmental approach that 258 aims to enhance the intervention recipient's intrinsic motivation to change and building on their 259 self-identified goals and strengths. 60-62 The "stages of change" framework is widely applied to 260 substance use disorders, and proposes that individuals move through discrete steps when 261 making behavior change: Pre-contemplation, Contemplation, Preparation, Action, and 262 Maintenance .  practice models are strategies that the social worker can incorporate into their interventions in 278 order to help people meet their goals (e.g., task-centered practice, 69 cognitive behavioral 279 therapy, 70 the crisis intervention model 71 ). For the purposes of the study, social workers will also 280 receive a refresher course in motivational interviewing and a stages of change framework as 281 described above. Although this intervention and theoretical framework is similar in some 282 respects to the peer navigation arm, the social work intervention is a single brief intervention 283 with referral to treatment that may also rely on social work theory and practice. The peer 284 navigation arm is distinguished by many unique aspects as outlined in sections above, and 285 highlights that the person delivering the intervention may be as or more important than the 286 intervention itself. 287 C.4. Participant eligibility criteria, recruitment, and enrollment. A consecutive sample of ED 288 patients presenting to the RIH and TMH EDs will be assessed for eligibility. The RIH ED has 11 289 full-time research assistants (RAs) available to recruit 24 hours per day, 7-days per week. At 290 TMH ED, there are 3 full-time RAs available to recruit participants from 7 AM-11 PM, 7 days a 291 week. Participants at TMH ED will be recruited during the hours in which social workers and 292 RAs are available as described above. Participants will be identified by screening the ED's 293 electronic medical records (EMR) or by referrals from ED treating providers. Patients who meet 294 the initial eligibility screen will undergo a further in-person assessment by a study RA.

295
Participants will be eligible if they are: (1) English speaking, (2) 18 years of age or older, and (3) 296 are being treated for an opioid overdose or identified as having had an opioid overdose in the 297 past 12 months or are being treated for a visit related to illicit opioid use (e.g., abscess, opioid 298 withdrawal), (4) are identified as having an alcohol OR other drug use disorder (excluding 299 marijuana) PLUS illicit opioid use in the past 6 months, (5) and are able to provide informed 300 consent. Participants are ineligible if they are critically ill or injured, are previously enrolled in the 301 trial, in police custody or incarcerated, pregnant, or live outside of Rhode Island or Southeastern 302 Massachusetts where they primarily receive their care (patients must live in state OR within 303 Southeastern Massachusetts but primarily receive care in Rhode Island to link to administrative 304 database). Patients who are critically ill will be eligible once cleared by their physician. 305 C.4.1. After screening, if the patient is eligible and willing to participate, then full written informed 306 consent will be obtained. After obtaining consent, the RA will randomly assign that patient (1:1 307 allocation) to the peer navigation arm or the social work arm using sealed envelopes. The 308 randomization schedule will consist of permuted block sizes stratified on gender and age. Both 309 gender and age may be important determinants of the effectiveness of treatment, and will be 310 examined as moderators in our exploratory analysis. This schedule will be maintained by the 311 study analyst; neither the study PIs, nor the recruiting staff will be aware of the schedule. We 312 anticipate the intervention will begin within 30 minutes of randomization. 313 314 315 C.4.2. Patients who decline to be randomly assigned to one of the two treatment groups as part 316 of the study are also eligible to be part of the "no intervention" control group. Such patients will 317 be offered the opportunity to be assessed by either a peer navigator, social worker, or both. If 318 they decline to be assessed by either a peer navigator or a social worker, they will be eligible to 319 be in the control group. Eligible control patients will be asked for their permission to participate 320 in follow-up assessments and other data collection procedures throughout the study period. 321 322 C.4.3. Study assessments. Baseline assessments in the EDs will collect information about 323 socio-demographic characteristics, medical history, medication use, substance use, prior 324 addiction treatment, pain symptoms, and depression. The survey forms may be RA or self-325 administered. Sections on more sensitive topics will be self-administered unless the participant 326 requests that the research assistant administer it. Patients in the randomized group will receive 327 a $40 gift card on the day of enrollment for completing the baseline survey while those in the 328 control group will receive a $25 gift card for completing their baseline survey. See appendix for 329 instruments.

330
Qualitative interviews will be conducted by telephone or in-person 2-10 days following 331 enrollment in the ED. These audio-recorded interviews will take about 45 minutes to an hour to 332 complete and participants will receive a $50 gift card for their time. In-person interviews will be 333 held at a Lifespan facility or a public community space that has been agreed upon by RA(s) and 334 participant. Transportation in the form of cab voucher and RIPTA bus-pass will be provided to 335 those requiring it. 336 337 338 C.4.4. Qualitative-Interview Procedures: The protocol will apply the same recruitment 339 methodology as the Navigator study and patients will be approached after they have completed 340 their assigned Navigator treatment arm, or after assessments for the control group. Navigator 341 participants will be eligible for participation in the interviews if they have completed the baseline 342 assessment and have been assigned to one of the intervention arms, and received the 343 intervention in the ED, or control arms of the study, have a working telephone number or are 344 prepared to return a Lifespan facility to participate in the patient interview protocol. Qualitative 345 interview audio recordings will be transcribed by a HIPAA compliant transcription service. 346 347 C.4.6 Qualitative Data Analysis 348 After the audio-recordings have been transcribed and cleaned by the RA, the investigators will 350 to read the transcripts and meet on several occasion to conduct an iterative content analysis, 351 based on the approach of the immersion-crystallization method of qualitative analysis. This 352 qualitative approach involves individual followed by a larger group determination of emerging 353 themes around the content of the interviews discussing the data as a group to determine 354 emerging themes, salient across all groups, and also those themes reflected by the subgroups 355 sampled. From this thematic analysis a code book will be developed and NVivo qualitative data 356 coding software will be employed to manage and sort the data. Coding discrepancies emerging 357 during investigators meeting will be discussed and resolved; to collectively determine the final 358 code. Following NVivo analysis the investigators will meet to complete the interpretation of the 359 themes. 360 361 C.5. Definition and operationalization of primary outcomes. This RCT will have two primary 362 endpoints: (1) 30-day treatment engagement, and (2) recurrent ED visits for an overdose. We 363 will obtain objective assessments of these outcomes through the use of statewide administrative 364 database as outlined below: 365 30-day treatment engagement (endpoint 1). The first primary outcome, engagement in addiction 366 treatment, will be defined as the proportion who are admitted to a formal addiction treatment 367 program within thirty days following the initial ED visit. This outcome was chosen because a 368 key short-term goal of the ED behavioral intervention is to promote early treatment engagement. 369 This outcome will be assessed using BHDDH and Prescription Drug Monitoring Program 370 (PDMP) records. The BHDDH database contains information on all admissions to publicly 371 funded substance abuse treatment programs in the state. We will define treatment engagement 372 as admission to any of the program types licensed by BHDDH, including inpatient detoxification, 373 day treatment programs, residential treatment, intensive outpatient services, and opioid 374 treatment programs (i.e., methadone). Second, we will query the participant's RI PDMP records 375 in order to identify enrollment in office-based buprenorphine therapy. The RI PDMP manages a 376 database that contains information on all prescriptions for schedule II-IV substances filled in the 377 state. The database is updated daily; all pharmacies are required to report prescriptions within 378 48-hours of the fill date. The database includes information on the patient (e.g., name, sex, birth 379 date, address including zip code), the prescription filled (e.g., quantity, days supply, national 380 drug code number), and prescriber/pharmacy data. Pharmacies are required by law to report 381 prescriptions to the PDMP regardless of payment type. All records will be linked 382 deterministically to participant data using identifiable information (e.g., name, social security 383 number) within the Stronghold computing environment, a HIPAA-compliant server maintained 384 by Dr. Marshall's team at Brown University. Our research team has experience extracting and 385 analyzing PMDP and BHDDH data for statewide surveillance purposes. 72 386 Recurrent ED visits for overdose (endpoint 2). The second primary outcome, recurrent ED visit 387 for overdose, will be defined as the proportion of participants who are treated in any Rhode 388 Island ED for an opioid overdose at any time during the 18-month follow-up period following the 389 initial ED visit. Recurrent ED visits for overdose were chosen as the second primary outcome 390 as a long-term goal of the ED behavioral interventions is reduce fatal and non-fatal overdose. 391 Two data sources will be used to assess this outcome. First, we will access the electronic 392 medical records (EMRs) of the 12 EDs in Rhode Island through the Rhode Island Quality 393 Institute Statewide Health Information Exchange. This data source will be made accessible 394 through Brown's Advance-CTR Unified Research Data Sharing Access (URSA) infrastructure.

395
This unified data system provides access to EMR data from all major health systems in Rhode 396 Island. Thus, we will capture repeat visits for an opioid overdose that occur in all 12 EDs in 397 Rhode Island. We will define an ED visit for an opioid overdose based on CDC guidelines for all 398 opioid poisonings (which includes illicit opioids) and utilizes International Classification of 399 Disease (ICD) coding. 73 Second, we will query the RI Department of Health (RIDOH) Opioid 48-400 Hour Overdose Surveillance System. The RIDOH mandates all suspected opioid overdose 401 cases presenting to an RI hospital be reported to the department within 48 hours. 74 This data 402 source will capture recurrent overdoses not identified by ICD codes in the unified EMR data 403 system, and also contains additional fields of interest (e.g., pre-existing risk factors for 404 overdose). 405 C.6. Statistical analyses. For all analyses, routine procedures will first be conducted to ensure 406 data accuracy/adequacy. We will use an intention-to-treat (ITT) approach in all analyses to 407 address potential problems inherent in following only intervention completers; a sensitivity 408 analysis ("per protocol") will be conducted among only those that complete the ED intervention. 409 Prior to examining intervention effects, we will first assess the success of randomization on pre-410 intervention characteristics using analyses of variance (ANOVA), with intervention group as the 411 predictor variable. If there are differences in baseline characteristics, we will include these 412 covariates in the primary outcome analyses, as described below. Given use of administrative 413 data sources we anticipate minimal missingness in our final dataset. However, missing 414 covariate data will be handled using multiple imputation performed in two stages using chained 415 equations that specify the conditional models for all of the variables with missing values. 75-77 416 C.6.1. Effectiveness, 30-day engagement in treatment (primary endpoint 1a). We will 417 compare the effectiveness of the peer navigation versus social work intervention on increasing 418 engagement in formal addiction treatment within 30 days of the initial ED visit. As the primary 419 analysis, we will compare the proportion who are admitted to a licensed addiction treatment 420 program (using chi-square analysis) between the two groups. Next, logistic regression models 421 will be used to determine the independent effect of the intervention arm on 30-day treatment 422 admission, adjusting for any baseline covariates as described above. 423

C.6.2 Effectiveness, any behavioral intervention versus "no intervention" treatment 424
(primary endpoint 1b). We will conduct analyses similar to those described above to determine 425 whether participants who receive any intervention have improved outcomes compared to those 426 who refuse to receive any behavioral intervention. The primary independent variable of interest 427 for Aim 1b is treatment versus control group membership. However, unlike Aim 1a (in which two 428 randomized interventions are compared), patients non-random assignment (self-selection) into 429 the control group. As such, we will need to account for factors that may be associated with 430 refusing a behavioral intervention in the ED. We will use inverse probability of treatment weight 431 (IPTW) techniques to account for: (1) baseline risk factors that predict control group 432 membership, and (2) post-randomization confounding. 433 434 C.6.3. Effectiveness, recurrent ED visit for an opioid overdose (primary endpoint 2). We 435 will compare the effectiveness of the peer navigation versus the social work intervention on 436 preventing subsequent ED visits for opioid overdose. As the primary analysis, we will compare 437 the overall proportion of patients experiencing a subsequent opioid overdose over the 18-month 438 follow-up period between the intervention groups (using chi-square analysis). Next, logistic 439 regression models will be used to determine the independent effect of the intervention arm on 440 recurrent ED visits for opioid overdoses, adjusting for baseline covariates. 441 In exploratory analyses, we will also examine a number of other outcomes, including: 442 overdose rates, overdose death (all overdose deaths in RI are analyzed by Dr. Marshall's team 443 and can be linked deterministically to patient data), and successful completion or retention in 444 addiction treatment. Successful completion and/or retention in addiction treatment will be 445 defined based on discharge data collected in BHOLD and prescription refill data in the PDMP 446 (e.g., on MAT for 6 months). We will examine the time to ED visit for an opioid overdose using 447 a Kaplan-Meier analysis. Patients will be censored at the end of the 18-month follow-up period, 448 considered the last point of contact. We will use Breslow's method to test if the time to 449 subsequent opioid overdose rates differs between the groups. Next, Cox proportional hazards 450 modeling will be used to estimate hazard ratios (HRs) and corresponding 95% confidence 451 intervals (CIs) for occurrence of repeat overdose between groups. HRs will be adjusted for 452 clinical and demographic characteristics believed to predict the outcome of opioid overdose in 453 order to adjust for possible residual confounding and treatment-factor interactions. Finally, since 454 participants may experience multiple opioid overdoses during follow-up, we will also conduct 455 recurrent-event survival analyses. 78 These models extend the Cox model approach and allow 456 for estimation of hazard ratios pooled across repeated periods at risk. 79 Finally, we will examine 457 if heterogeneity of intervention effect is modified by age, sex, race, pre-existing chronic pain, 458 past treatment history, and reason for presentation to the ED. We will perform stratified 459 subgroup analyses to determine if treatment effects vary between groups of individuals. 460 C.7. Feasibility and Sample Size Calculation. First, we assume that ~2,000 patients will be 461 treated at the RIH or TMH ED for an opioid overdose over the 24-month recruitment period; this 462 is based on data from the RI Opioid Overdose Surveillance System and is a conservative 463 estimate that does not include other eligibility criteria (e.g., recent overdose). Next, we assume 464 that 65% (n=1,300) of these patients will be willing to be screened (prior studies of behavioral 465 interventions for drug use at the RIH ED have had screening rates > 80%). 80 Third, we assume 466 that ~50% (n=650) of these patients will be eligible and randomized to an intervention arm 467 (n=325 per arm). Thus, we estimate that 24 months will be required to recruit 650 participants. 468 Given our use of objective outcome data from administrative datasets, we do not expect dropout 469 to significantly impact our statistical power, but our power calculations conservatively reflect a 470 10% loss to follow-up rate. Will attempt to also enroll similar numbers in the control arm (n=325). 471 For our sample size calculation, we assumed that 7% of participants in the social work arm 472 will enroll in a formal treatment program within 30 days of ED discharge (based on preliminary 473 data from the state and RIH). Given this, we have >80% power to detect a two-fold increase 474 (i.e., >7% absolute increase) in the rate of 30-day treatment engagement between the two arms 475 ( Figure 2A); this increase has been deemed a bench-mark by key state stakeholders. For 476 primary endpoint 2 (recurrent ED visit for an opioid overdose), we assumed that 15% of patients 477 in the social work arm will have a recurrent ED visit for an opioid overdose within 18-months of 478 their first visit. This estimate is based on a chart review of 374 patients after program 479 implementation of the RIH ED behavioral intervention program and in recently published data by 480 Banta-Green et al. 81 The latter study from Washington State found around a 20% incidence of 481 overdose within 18-months following of an initial ED visit. Our assumption of a 15% incidence of 482 repeat ED visit for overdose is conservative compared to this finding, particularly in light of the 483 fact the Rhode Island has nearly twice as many overdose deaths per capita than Washington 484 State. 82 We will have >80% power to detect a 50% relative reduction (7.5% absolute reduction) 485 in the risk of recurrent overdose within 18 months of their ED visit ( Figure 2B) that most individuals who will 497 Figure 2A: Power for Endpoint, 30-day treatment engagement experience another overdose will do so within the first 18 months. 81 Second, given the urgent 498 need to have an evidence-based evaluation of peer-led behavioral interventions for OUDs, a 499 shorter length of follow-up would allow us to disseminate our study findings sooner. However, 500 we recognize that there is potential value in having a longer follow-up period both from an 501 impact standpoint and also in terms of statistical power to detect a difference between the 502 treatment groups. Figure 2b compares the difference in power between 18 and 30 month study 503 endpoints (> 80% to detect a RR=0.5 for 18 months and RR=0.58 for 30 months; absolute rate 504 of overdose in the intervention arm of ~7.5% at 18 months and ~11.5% at 30 months, note that 505 this assumes an 20% incidence of overdose at 30 months). Therefore, as a contingency plan, 506 we could extend the length of follow-up by one year (30 months total) based on 18 month 507 outcome analyses. This is possible because of the use of administrative data for the outcomes 508 assessment, but would require additional cost (approximately $120,000). We propose to make 509 this decision in conjunction with the LJA Foundation should this proposal be approved for 510 funding.  Purpose: The purpose of the patient interviews is to understand the experiences of ED 527 patients who have agreed to participate in the study after meeting study eligibility 528 criteria. These will be semi-structured interviews to allow for patient directed reflections 529 on the interactions they experienced as a patient before they were recruited into the 530 study, the decision making that led them to agree to participate in the study, their 531 experience with the peer navigator or social worker they were randomized to be 532 Figure 2B: Power for Endpoint 2, Recurrent ED visit for overdose exposed to, and for those patients in the control group (declining either social worker or 533 peer navigator), their reasons for not wanting to see either treatment option. The key 534 research questions to be addressed in the qualitative portion of the navigator study are 535 around the ED research experience , how this could be improved, and the effect that the 536 Navigator study has on their motivation to engage in treatment, in the ED and after 537 discharge, for their opioid use. Below are specific themes that the patient interview 538 protocol will address: 539 1. What influences the patient to be part of the Navigator study-were their reasons 540 apart from compensation?  Screening and recruitment. The protocol will apply the same recruitment methodology 584 as the Navigator study and patients will be approached after they have completed their 585 assigned Navigator treatment arm, or after assessments for the control group. Navigator 586 participants will be eligible for participation in the interviews if they have completed the 587 baseline assessment and have been assigned to one of the intervention arms, and 588 received the intervention in the ED, or control arms of the study, have a working 589 telephone number or are prepared to return to a LifeSpan affiliated facility or public 590 community space to participate in the patient interview protocol. We will use a selective 591 sampling approach, where the RA will approach patients eligible to participate in the 592 interviews from the recruitment schema below. At the time of consent in the ED into the 593 navigator study the patient will be asked to consent to possibly be contacted to take 594 part in the qualitive interview components of the study. Within 2-10 days after the ED 595 visit, which will be sufficiently recent to remember the ED visit, and, for the peer 596 navigator group, sufficient time will have passed to ensure that there has been at least 597 one contact with the per navigator.

598
The RA will call a selected participant and ask if they w oud like to participate in the 599 semi-structured interview will be conducted by phone or in-person and will be 600 audiotaped for transcription and later analysis. Verbal consent will be obtained from the 601 participant. The RA will give the option of completing the interview by phone or in 602 person at a Lifespan facility or a public community space that has been agreed upon by 603 RA(s) and participant. If the participant agrees to a telephone interview the RA will 604 remind the participant that the interview will be audio-recorded, switch on the audio 605 recorder and proceed with the verbal consent statement (see Procedures below).

606
Alternatively an appointment for the in-person interview will be scheduled and the same 607 verbal consent procedure will be conducted.. We anticipate hearing distinctly different reports of the overall ED treatment experience 627 and reflections on the Navigator research and treatment experience in the three groups, 628 regardless of eligibility criteria. We anticipate that the interviews will take between 45 629 minutes to an hour to complete, and to reflect this additional request on Navigator 630 participants' time we will offer participants a $50 gift card for participation. Participants 631 will be offered transportation to bring them to the interview if they chose the in-person 632 option. The participant data will be confidential and will not be linked with any other 633 research data they provide as part of the main Navigator research study.

634
Procedure. At the start of the telephone or in-person interview the person will be 635 reminded that the interview will be audio recorded and asked to consent to this-this 636 consent will be repeated when the audio-recorder is activated. The following preamble 637 will be recited to the participant: you about your experiences of taking part in the Navigator study when you were treated in the 646 emergency department at (RIH/TMH) on -------(date when participant was recruited). We would 647 also like to ask you about your overall experience when you were treated in the emergency 648 department at the time of your recruitment into this study. 649 We will audio recorded this interview and have what you said transcribed just as you said it but 650 without any information that could identify you or others to protect your confidentiality. This 651 audio recording will be securely stored for data analysis purposes only and destroyed as soon 652 as is possible after the analysis is complete. At any time, you can refuse to answer any question 653 or chose to end the interview. Do you agree to be interviewed by me today?" 654 Completing this interview will probably take 45 minutes to an hour of your time. There are a 655 number of questions we would like you to answer y. There are no right or wrong answers, this 656 is about your experiences. 657 There are no questions that should cause you any discomfort. Your taking part in this research 658 interview is completely voluntary. You are free to choose not to complete or take part in this 659 interview. 660 Your completion of this interview may not benefit you personally. We are hoping these 661 completed interviews will provide information to help us to understand how the Navigator study 662 could be improved to help others 663 The interviews from this study will be kept confidential. None of the information you provide will 664 have your name or any number on it that will identify your personally. 665 If you have any questions about this interview or the research study itself, please feel free to ask 666 the research assistant providing you with this information. Or you can call us at 444-4444. 667 If you have any questions about your rights as a research subject please feel free to call our 668 Research Protections Office Director, Janice Muratori, at 444-6246. 669 After the participant has agreed to continue with the interview the researcher will 670 implement the semi-structured interview guide (see Appendices 2 and 3). When the 671 interview is completed the participant will then be compensated and thanked for their 672 involvement; study staff will upload the audio file for transcription, and complete the staff  the Navigator research team that will be conducting this interview with you today. We 684 want to ask you about your experiences of taking part in the Navigator study when you 685 were treated in the emergency department at (RIH/TMH) on -------(date when participant 686 was recruited). We would also like to ask you about your overall experience when you 687 were treated in the emergency department at the time of your recruitment into this 688 study. 689 We will audio recorded this interview and have what you said transcribed just as you 690 said it but without any information that could identify you or others to protect your 691 confidentiality. This audio recording will be securely stored for data analysis purposes

698
There are no questions that should cause you any discomfort. Your taking part in this 699 research interview is completely voluntary. You are free to choose not to complete or 700 take part in this interview.

701
Your completion of this interview may not benefit you personally. We are hoping these 702 completed interviews will provide information to help us to understand how the 703 Navigator study could be improved to help others 704 The interviews from this study will be kept confidential. None of the information you 705 provide will have your name or any number on it that will identify your personally.

711
No one outside of this study will have access to these recordings and they will be 712 destroyed after our final report is written.    After you were randomized you were told that a peer navigator (or term Probes: Describe the discussion you had with (probe who this was).

847
How did you feel about that discussion? How did your experience of stigma impact services provided to you for your opioid use?

912
How did this impact whether you wanted to accept services offered? Navigator research team that will be conducting this interview with you today. We want to ask 955 you about your experiences of taking part in the Navigator study when you were treated in the 956 emergency department at (RIH/TMH) on -------(date when participant was recruited). We would 957 also like to ask you about your overall experience when you were treated in the emergency 958 department at the time of your recruitment into this study. 959

We will audio recorded this interview and have what you said transcribed just as you said it but 960
without any information that could identify you or others to protect your confidentiality. This 961 audio recording will be securely stored for data analysis purposes only and destroyed as soon 962 as is possible after the analysis is complete. At any time, you can refuse to answer any question 963 or chose to end the interview. Do you agree to be interviewed by me today?" 964 Completing this interview will probably take 45 minutes to an hour of your. There are a number 965 of questions we would like you to answer verbally. There are no right or wrong answers, this is 966 about your experiences. 967 There are no questions that should cause you any discomfort. Your taking part in this research 968 interview is completely voluntary. You are free to choose not to complete or take part in this 969 interview. 970 Your completion of this interview may not benefit you personally. We are hoping these 971 completed interviews will provide information to help us to understand how the Navigator study 972 could be improved to help others 973 The interviews from this study will be kept confidential. None of the information you provide will 974 have your name or any number on it that will identify your personally. 975 If you have any questions about this interview or the research study itself, please feel free to ask 976 the research assistant providing you with this information. Or you can call us at 444-4444. 977 If you have any questions about your rights as a research subject please feel free to call our 978 Research Protections Office Director, Janice Muratori, at 444-6246. 979 No one outside of this study will have access to these recordings and they will be destroyed 980 after our final report is written. The purpose of this study is to determine the effectiveness of peer navigation versus a 3 standard behavioral intervention delivered in the emergency department (ED) to overdose 4 patients and those at risk of recurrent opioid overdose. A total of 650 ED patients will be 5 recruited from two emergency departments in a single health care system in Providence, Rhode 6 Island into a two-arm randomized trial with 18 months of follow-up post-randomization. Eligible 7 participants will be randomly assigned (1:1) in the ED to receive a behavioral intervention from 8 a certified peer recovery support specialist or a licensed clinical social worker (LCSW).

9
Effectiveness will be measured objectively through linkage to administrative statewide 10 databases, with two primary endpoints: (1) engagement in formal addiction treatment (e.g., 11 inpatient services, outpatient services, medication-assisted treatment [MAT]) from a licensed 12 substance abuse treatment provider within 30 days following the index ED visit; and (2), 13 recurrent ED visit for an opioid overdose within 18-months following the index ED visit.
14 Exploratory outcomes of interest are: overdose fatality, repeat ED visits related to opioids, and 15 successful completion of an addiction treatment program and/or long-term retention in MAT.

20
This study addresses the following research questions: 2. If peer navigation is found to be more effective than standard of care, is there 28 heterogeneity of treatment effect related to key patient characteristics (e.g., sex, race, 29 type of opioid used, and history of comorbid chronic pain, depression or posttraumatic 30 stress disorder)? 31 32 33

35
A total of 650 patients treated for an opioid overdose or at risk for an opioid overdose 36 (defined below) will be recruited (n=325 per arm) from two emergency departments in a single 37 health care system in Providence, Rhode Island and followed prospectively using administrative 38 datasets. Our primary outcomes will be: (1) engagement in treatment within 30-days after the ED 39 visit, and (2) recurrent ED visit for opioid overdose over the 18-month follow-up period.

41
We will recruit adult ED patients who are: (1) being treated for an opioid overdose, or (2) 42 have had an opioid overdose in the past 12 months (identified by self-report during screening or 43 in review of the EMR); or (3) are presenting with a visit related to illicit injection opioid use 44 (e.g., cutaneous injection-related infection, opioid withdrawal, endocarditis). We are specifically targeting patients with a current or recent opioid overdose and those who inject opioids illicitly, 46 as they are at highest risk for opioid overdose and death.

48
All participants will be recruited from two EDs-Level 1 and Level 2 trauma centers-49 located in the state's capital of Providence. Together, these two EDs receive over 175,000 adult 50 visits each year. Between 2017 and 2018, the two EDs reported a total of 1,446 visits for 51 suspected opioid overdoses, representing 45% of all ED visits for suspected opioid overdoses 52 reported to the Rhode Island Department of Health (n = 3,239).

54
A consecutive sample of ED patients will be assessed for eligibility by one of eleven full-55 time research assistants employed in the two EDs who can recruit participants 24 hours per day, 56 seven days a week. Potential participants will be identified by the research assistants by 57 screening electronic medical records (EMR) or by referrals from treating providers in the ED.

58
Patients who meet the initial eligibility screen will undergo a further in-person assessment by a 59 study RA.

61
Participants will be eligible if they are: (1) English-speaking, (2) 18 years of age or older, and 62 Recurrent ED visits for overdose (primary endpoint 2): The second primary outcome, 137 recurrent ED visit for overdose, will be defined as the proportion of participants who are treated 138 in any Rhode Island ED for an opioid overdose at any time during the 18-month follow-up 139 period following the initial ED visit. Recurrent ED visits for opioid overdose were chosen as the 140 second primary outcome as a long-term goal of the ED behavioral interventions is to reduce fatal 141 and non-fatal overdose. Two data sources will be used to assess this outcome. First, we will 142 access the electronic medical records (EMRs) of the 12 EDs in Rhode Island through the Rhode 143 Island Quality Institute Statewide Health Information Exchange. This data source will be made 144 accessible through Brown's Advance-CTR Unified Research Data Sharing Access (URSA) 145 infrastructure. This unified data system provides access to EMR data from all major health 146 systems in Rhode Island. Thus, we will capture repeat visits for an opioid overdose that occur in 147 all 12 EDs in Rhode Island. We will define an ED visit for an opioid overdose based on CDC 148 guidelines for all opioid poisonings (which includes illicit opioids) and utilizes International

152
This data source will capture recurrent overdoses not identified by ICD codes in the unified 153 EMR data system, and also contains additional fields of interest (e.g., pre-existing risk factors for 154 overdose).

156
Finally, we will determine mortality outcomes by requesting data from the National Death 157 Index (NDI), which is a centralized database of death record information on file in state vital 158 statistics offices. We will use the NDI data in conjunction with the Rhode Island Department of 159 Health medical examiner data to determine whether or not a participant in the study has died 160 during follow-up, and if so, the cause of death (including overdose).

165
For the two primary outcomes (engagement in formal SUD treatment within 30 days of the 166 initial ED visit and occurrence of a subsequent ED visit for opioid overdose within 18 months of 167 the initial ED visit), we will use separate logistic regression models with indicators for treatment 168 allocation and study site, as well as term representing the interaction of treatment allocation with 169 study site. Second, we will conduct subgroup analyses to understand potential heterogeneity of 170 treatment effects by age and gender.

172
For all analyses, routine procedures will first be conducted to ensure data accuracy/adequacy. 173 We will use an intention-to-treat (ITT) approach in all analyses to address potential problems 174 inherent in following only intervention completers; a sensitivity analysis ("per protocol") will be 175 conducted among only those that complete the ED intervention. Given use of administrative data 176 sources we anticipate minimal missingness in our final dataset. However, missing outcome and 177 covariate data will be handled using case-wise deletion. Additionally, we will perform a 178 sensitivity analysis to determine the potential impact of missing data on treatment effect. This 179 sensitivity analysis will use multiple imputation performed using chained equations that specify 180 the conditional models for all of the variables with missing values.

182
Analysis Plan for the Navigator Trial 08/07/2020 Page 5 of 7 Effectiveness, 30-day engagement in treatment (primary endpoint 1): We will compare the 183 effectiveness of the peer navigation versus social work intervention on increasing engagement in 184 formal addiction treatment within 30 days of the initial ED visit. As the primary analysis, we will 185 compare the proportion who are admitted to a licensed addiction treatment program (using chi-186 square analysis) between the two groups. In the next stage of our primary analysis, logistic 187 regression models will be used to determine the independent effect of the intervention arm on 188 30-day treatment admission, adjusting for study site and an interaction term between treatment 189 allocation and study site, as described above.

191
Effectiveness, recurrent ED visit for an opioid overdose (primary endpoint 2): We will 192 compare the effectiveness of the peer navigation versus the social work intervention on 193 preventing subsequent ED visits for opioid overdose. As the primary analysis, we will compare 194 the overall proportion of patients experiencing a subsequent opioid overdose over the 18-month 195 follow-up period between the intervention groups (using chi-square analysis). In the next stage 196 of our primary analysis, logistic regression models will be used to determine the independent 197 effect of the intervention arm on recurrent ED visits for opioid overdoses, adjusting for study 198 site, as described for primary endpoint 1.

200
In a sensitivity analysis, we will assess imbalance in key prognostic factors for the study 201 outcomes (e.g., lifetime history of overdose, lifetime treatment engagement, age, sex, race, and 202 housing status) between the two study arms. If imbalance is observed, we will include these 203 covariates in the logistic regression models to determine the robustness of the primary analysis 204 results.

206
In exploratory analyses, we will also examine a number of other outcomes, including: 207 overdose rates, overdose death, and successful completion of or retention in addiction treatment.

208
Successful completion of and/or retention in addiction treatment will be defined based on 209 discharge data collected in BHOLD and prescription refill data in the PDMP (e.g., on MAT for 210 ≥6 months). We will examine the time to ED visit for an opioid overdose using a Kaplan-Meier 211 analysis. Patients will be censored at the end of the 18-month follow-up period, considered the 212 last point of contact. We will use Breslow's method to test if the time to subsequent opioid 213 overdose rates differs between the groups. Next, Cox proportional hazards modeling will be used 214 to estimate hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) for occurrence 215 of repeat overdose between groups. HRs will be adjusted for clinical and demographic 216 characteristics believed to predict the outcome of opioid overdose in order to adjust for possible 217 residual confounding and treatment-factor interactions. Finally, since participants may 218 experience multiple opioid overdoses during follow-up, we will also conduct recurrent-event 219 survival analyses. These models extend the Cox model approach and allow for estimation of 220 hazard ratios pooled across repeated periods at risk. Finally, we will examine if heterogeneity of 221 intervention effect is modified by age, sex, race, pre-existing chronic pain, past treatment history, 222 and reason for presentation to the ED. We will perform stratified subgroup analyses to determine 223 if treatment effects vary between groups of individuals.

225
Feasibility and Sample Size Calculation. First, we assume that ~2,000 patients will be treated 226 at the two EDs for an opioid overdose over the 24-month recruitment period; this is based on 227 data from the RI Opioid Overdose Surveillance System and is a conservative estimate that does 228 Analysis Plan for the Navigator Trial 08/07/2020 Page 6 of 7 not include other eligibility criteria (e.g., recent overdose). Next, we assume that 65% (n=1,300) 229 of these patients will be willing to be screened (prior studies of behavioral interventions for drug 230 use at the two EDs have had screening rates > 80%) . Third, we 231 assume that ~50% (n=650) of these patients will be eligible and randomized to an intervention 232 arm (n=325 per arm). Thus, we estimate that 24 months will be required to recruit 650 233 participants. Given our use of objective outcome data from administrative datasets, we do not 234 expect dropout to significantly impact our statistical power, but our power calculations 235 conservatively reflect a 10% loss to follow-up rate.

237
For our sample size calculation, we assumed that 7% of participants in the social work arm 238 will enroll in a formal treatment program within 30 days of ED discharge (based on preliminary 239 data from the state and RIH). Given this, we have >80% power to detect a two-fold increase (i.e.,

240
>7 percentage point absolute increase) in the rate of 30-day treatment engagement between the 241 two arms ( Figure 2A); this increase has been deemed a bench-mark by key state stakeholders.

242
For primary endpoint 2 (recurrent ED visit for an opioid overdose), we assumed that 15% of around a 20% incidence of overdose within 18-months following of an initial ED visit. Our 248 assumption of a 15% incidence of repeat ED visit for overdose is conservative compared to this 249 finding, particularly in light of the fact the Rhode Island has nearly twice as many overdose 250 deaths per capita than Washington State. (Jiang et al., 2018) We will have >80% power to detect 251 a 50% relative reduction (7.5 percentage point absolute reduction) in the risk of recurrent 252 overdose within 18 months of their ED visit ( Figure 2B), this reduction was felt to be clinically 253 relevant and commensurate with statewide goals in reducing overdose via various strategies.

255
We chose to evaluate the outcome date within the first 18 months after the initial ED visit for 256 two main reasons. First, the risk of recurrent overdose appears to level-off by about 18 months, 257 meaning that most individuals who will experience another overdose will do so within the first 258 18 months. Second, given the urgent need to have an evidence-based evaluation of peer-led 259 behavioral interventions for OUDs, a shorter length of follow-up would allow us to disseminate 260 our study findings sooner. However, we recognize that there is potential value in having a longer 261 follow-up period both from an impact standpoint and also in terms of statistical power to detect a 262 difference between the treatment groups. Therefore, as a contingency plan, we could extend the 263 length of follow-up by one year (30 months total) based on 18-month outcome analyses. We will 264 make this decision in conjunction with the LJA Foundation.

269
The research protocol has be reviewed and approved by the Lifespan Institutional Review