Addiction Consultation Services for Opioid Use Disorder Treatment Initiation and Engagement

Key Points

Question

Do interprofessional hospital addiction consultation services increase initiation and engagement in medication for opioid use disorder (MOUD) treatment after discharge?

Findings

In this stepped-wedge cluster randomized clinical trial of 2315 adults with opioid use disorder with hospitalizations in 6 New York, New York, public hospitals, rates of postdischarge MOUD were compared for patients admitted to hospitals providing an addiction consultation intervention vs usual care. Patients hospitalized during the intervention period had significantly higher odds of MOUD initiation (7.96 times higher) and 30-day MOUD treatment engagement (6.90 times higher).

Meaning

The results of this trial found that interprofessional hospital addiction consultation services increased postdischarge MOUD use among patients with opioid use disorder who were not previously receiving MOUD treatment.

Abstract

Importance

Medications for opioid use disorder (MOUD) are highly effective, but only 22% of individuals in the US with opioid use disorder receive them. Hospitalization potentially provides an opportunity to initiate MOUD and link patients to ongoing treatment.

Objective

To study the effectiveness of interprofessional hospital addiction consultation services in increasing MOUD treatment initiation and engagement.

Design, Setting, and Participants

This pragmatic stepped-wedge cluster randomized implementation and effectiveness (hybrid type 1) trial was conducted in 6 public hospitals in New York, New York, and included 2315 adults with hospitalizations identified in Medicaid claims data between October 2017 and January 2021. Data analysis was conducted in December 2023. Hospitals were randomized to an intervention start date, and outcomes were compared during treatment as usual (TAU) and intervention conditions. Bayesian analysis accounted for the clustering of patients within hospitals and open cohort nature of the study. The addiction consultation service intervention was compared with TAU using posterior probabilities of model parameters from hierarchical logistic regression models that were adjusted for age, sex, and study period. Eligible participants had an admission or discharge diagnosis of opioid use disorder or opioid poisoning/adverse effects, were hospitalized at least 1 night in a medical/surgical inpatient unit, and were not receiving MOUD before hospitalization.

Interventions

Hospitals implemented an addiction consultation service that provided inpatient specialty care for substance use disorders. Consultation teams comprised a medical clinician, social worker or addiction counselor, and peer counselor.

Main Outcomes and Measures

The dual primary outcomes were (1) MOUD treatment initiation during the first 14 days after hospital discharge and (2) MOUD engagement for the 30 days following initiation.

Results

Of 2315 adults, 628 (27.1%) were female, and the mean (SD) age was 47.0 (12.4) years. Initiation of MOUD was 11.0% in the Consult for Addiction Treatment and Care in Hospitals (CATCH) program vs 6.7% in TAU, engagement was 7.4% vs 5.3%, respectively, and continuation for 6 months was 3.2% vs 2.4%. Patients hospitalized during CATCH had 7.96 times higher odds of initiating MOUD (log-odds ratio, 2.07; 95% credible interval, 0.51-4.00) and 6.90 times higher odds of MOUD engagement (log-odds ratio, 1.93; 95% credible interval, 0.09-4.18).

Conclusions

This randomized clinical trial found that interprofessional addiction consultation services significantly increased postdischarge MOUD initiation and engagement among patients with opioid use disorder. However, the observed rates of MOUD initiation and engagement were still low; further efforts are still needed to improve hospital-based and community-based services for MOUD treatment.

Trial Registration

ClinicalTrials.gov Identifier: NCT03611335

This randomized clinical trial examines the effectiveness of interprofessional hospital addiction consultation services in increasing medication for opioid use disorder treatment initiation and engagement.

Introduction

People in the US are dying of opioid-related overdose at unprecedented rates, with more than 82 000 such deaths in 2022.¹ Opioid use disorder (OUD) can be treated with highly effective, lifesaving medications, but only an estimated 22% of people with OUD receive medication for OUD treatment (MOUD).² OUD treatment rates are low among racial and ethnic minority and socioeconomically disadvantaged populations,^3,4,5,6,7,8 who are also disproportionately affected by the current overdose crisis.^9,10,11 This treatment gap persists despite substantial federal, state, and local investments in expanding MOUD treatment access.^12,13

Hospitals are a touchpoint for reaching people with OUD who have elevated risk of death and other poor outcomes.^14,15,16 Individuals with untreated OUD have high rates of hospital admissions,¹⁷ which can be an opportunity for initiating MOUD treatment and improving engagement with medical and harm reduction services.^18,19,20 Inpatient medical clinicians can provide MOUD in the hospital, prescribe postdischarge buprenorphine, and discharge patients to ongoing treatment. However, rates of treatment following hospitalization remain low,²¹ largely due to structural and social barriers.^{16,20,21,22,23,24} The needs of hospitalized patients with OUD, many of whom have complex medical and psychiatric conditions, experience polysubstance use, and face structural challenges, such as poverty and criminal-legal involvement, warrant having skilled and specialized clinicians.^{18,20,24,25,26,27} This has led to the increasing adoption of hospital addiction consultation services.^28,29,30

A growing body of evidence from observational studies suggests that treating OUD during hospitalization can be associated with better hospital care, increased use of MOUD, lessened addiction severity, improved trust in clinicians, and reduced mortality after discharge.^{20,28,31,32,33,34,35,36,37} However, other than a small pilot study,³⁸ to our knowledge, there have been no prospective randomized clinical trials of the effectiveness of hospital addiction consultation services compared with a control condition. Our pragmatic implementation-effectiveness trial, which was conducted in a large public hospital system, tested the hypothesis that an interprofessional addiction consultation service can increase postdischarge MOUD initiation and engagement.

This study represents a research collaboration with New York City Health + Hospitals (H+H), the largest municipal public health care system in the US. H+H provides essential medical services to more than 1 million individuals annually and is the primary source of medical and substance use disorder (SUD) care for low-income individuals in New York, New York.³⁹ Beginning in 2017, funding from a mayoral initiative to reduce opioid overdose deaths³⁹ provided H+H with resources to implement a new Consult for Addiction Treatment and Care in Hospitals (CATCH) program in 6 hospitals. The CATCH providers, which include clinicians and peer counselors, work as an interdisciplinary addiction consultation team²⁹ to diagnose SUD, make treatment recommendations, offer harm reduction education and support, and link patients to postdischarge addiction treatment.

Methods

Study Design

The primary objective of this hybrid type 1 randomized clinical trial⁴⁰ was to test the effectiveness of the CATCH intervention for increasing postdischarge initiation and engagement in MOUD treatment (dual primary outcome). Treatment initiation and engagement are key quality indicators for substance use services.^41,42 Details of the implementation process will be reported separately. The protocol was previously described,⁴³ and the trial was registered. The protocol (Supplement 1 and Supplement 2) was approved by New York University School of Medicine institutional review board with a waiver of consent, and there was no data and safety monitoring board. Reporting of results followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guidelines for cluster-randomized trials,⁴⁴ modified for a stepped-wedge design.⁴⁵

A stepped-wedge cluster randomized clinical trial compared outcomes for patients admitted during treatment as usual (TAU) and CATCH intervention periods. Parallel randomization was not acceptable to the H+H system, and the stepped-wedge design provided rigor while still allowing the timely entry of all hospitals into the CATCH condition.⁴³ Hospitals were required to be at a predetermined level of readiness, with at least 2 fully staffed CATCH teams, before randomization. Hospitals were randomized into 2 groups by the study biostatistician using a computer-generated list of random numbers. The first group of 3 hospitals was randomized August 1, 2018, and the second group was randomized May 31, 2019. Hospitals were informed of their start date at the time of randomization and implemented CATCH sequentially at assigned dates, separated by approximately 3 months.

Data were collected from October 2017 to January 2021. All hospitals were evaluated for 12 months preimplementation (TAU period) and a minimum of 12 months postimplementation of the CATCH intervention (intervention period) (Figure 1). Following the first 12 months of the intervention period, hospitals continued offering the same level of CATCH services and were followed up throughout a maintenance period that lasted until the end of the study. During the maintenance period, technical assistance decreased, but hospitals continued CATCH. The first month following implementation at each hospital was considered transitional, and intervention period measures began 1 month later.

Figure 1. Stepped-Wedge Study Design.

The study condition is demonstrated by hospital during each study period. Dates represent the start of the Consult for Addiction Treatment and Care in Hospitals (CATCH) program at each hospital. Each colored box represents 3 months. White boxes represent the 1-month transition period during which the CATCH program was introduced at each hospital. TAU indicates treatment as usual.

Setting and Study Population

The trial was conducted in 6 of the 11 H+H hospitals. Participating hospitals were selected by H+H to include those with the highest prevalence of patients with SUD, and with consideration for their geographic location. CATCH was offered as part of standard clinical care at the participating hospitals.

Patients were identified using New York State Medicaid claims (ie, Medicaid data) as the primary data source and electronic health records as a secondary data source. Individuals enrolled in Medicaid and Medicare (dual-eligible) were excluded because their service utilization could not be comprehensively captured in Medicaid data. A supplemental analysis with inclusion of this group has been provided (eTables 3 and 4 and eFigure 3 in Supplement 3). Continuous Medicaid coverage during the 30 days pre- and posthospitalization was not required. Eligible hospitalizations were of adults (18 years and older) with an admission or discharge diagnosis of OUD, opioid poisoning, or opioid adverse effects based on International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) diagnostic codes (eTable 1 in Supplement 3) and hospitalized for at least 1 night. Because the CATCH program did not serve dedicated inpatient psychiatric, detoxification, or prison units, patients hospitalized only in these departments (as identified using electronic health record data) were excluded. If a patient had a Medicaid claim for MOUD (prescription for buprenorphine or extended-release naltrexone, or a visit to an opioid treatment program) during the 30 days before hospitalization, the admission was excluded. Individuals could have multiple eligible admissions during the study.

Interventions

CATCH was designed as an interprofessional addiction consultation service.²⁹ Each hospital received funding for 3 teams who worked collaboratively, led by a physician with addiction expertise. Each team consisted of 3 providers: (1) addiction physician or nurse practitioner; (2) social worker or addiction counselor; and (3) peer counselor (Figure 2). CATCH services were provided at the request of the patient’s medical care team and available for any patients with alcohol or drug use, regardless of insurance, although the study population was limited to Medicaid patients with opioid diagnoses. CATCH team members assessed patients at the bedside, diagnosed SUD, made treatment recommendations, and provided peer support and harm reduction resources (including take-home naloxone). A key function was to link patients to postdischarge community-based addiction treatment. If a treatment linkage could not be established before discharge, patients may be referred to short-term bridge clinics staffed by CATCH providers.

Figure 2. Description of the Consult for Addiction Treatment and Care in Hospitals (CATCH) Program Model.

All 3 members of the CATCH team seek to meet with the patient during hospitalization. MOUD indicates medication for opioid use disorder.

Data Collection and Outcome Measures

Medicaid data were the source for all descriptive and outcome measures. Demographic characteristics were self-reported at the time of Medicaid application and described using the first eligible admission during the study period (called the index admission). To further characterize the population, past acute care utilization and diagnoses of medical and psychiatric conditions were described for 3 years before the index admission. Chronic conditions were measured using the Chronic Condition Indicator, version 2017.1, and the Clinical Classification Software, version 2017.1.⁴⁶ Characteristics of the hospitalization, including length of stay and diagnoses, were described for each eligible admission. Medicaid data identified 1 primary and up to 19 secondary discharge diagnoses for each hospitalization using ICD-10 codes. Substance use diagnoses were classified for opioids (OUD, opioid-related poisoning, opioid adverse events), alcohol (alcohol use disorder, alcohol-related poisoning, alcohol adverse events), and other drugs not including tobacco or prescription medications (other drug use disorder, related poisoning, and adverse events); diagnostic codes included in each category are described in eTable 1 in Supplement 3.

The dual primary outcomes were MOUD initiation and engagement postdischarge and did not include MOUD receipt during the hospital stay. Initiation (aim 1a) was defined as having at least 1 outpatient MOUD event (filled a prescription for buprenorphine or long-acting naltrexone, or had an encounter in an opioid treatment program) within 14 days following hospital discharge. Engagement (aim 1b) was defined as having, during the first 30 days following treatment initiation, either 2 encounters in an outpatient opioid treatment program (providing methadone and/or buprenorphine) or filling additional prescriptions for buprenorphine or long-acting naltrexone (≥2 prescription fills or 1 prescription that covers at least 28 days). A prespecified secondary outcome (aim 2) was the rate of continuous retention in treatment for 6 months, defined as MOUD program visits and/or filling MOUD prescriptions for more than 80% of days after initiation.

Statistical Analysis

The statistical approach accounted for clustering of patients within hospitals and the open cohort nature of the study by modeling repeated hospitalizations. Patients could have multiple eligible admissions during the TAU, CATCH, and maintenance periods, and these admissions could occur at multiple study hospitals. The models used 2 random effects (a hospital cluster-specific random effect and a patient-level random intercept effect). An autoregressive, AR(1) dependency structure was used to model between-period site-level correlation over time. Due to these study complexities, with hierarchies induced from clustering within patients and hospitals, a bayesian analysis was conducted. The prior distributions used for the model coefficients, random-effect variances, and the AR(1) correlation were normal, exponential, and uniform, respectively. Attempts to use frequentist modeling approaches encountered convergence difficulties due to model complexities. The eMethods in Supplement 3 provides details and explanations on the bayesian modeling approach for the analysis. The results focused on the posterior probabilities of the CATCH intervention effect coefficient in the logistic regression models. Analyses were performed using R, version 4.1.2 (R Foundation), and Stan, version 2.21.0.⁴⁷ Binary (0-1) treatment initiation and engagement indicators were each modeled using a hierarchical logistic regression model adjusted for age, sex, and a period factor (13 periods, each 3 months long). Eligible admitted patients following CATCH implementation (intervention and maintenance periods) were considered to be in the CATCH condition, while those in 12 months before CATCH implementation were considered to be in TAU. 95% Credible intervals (CrIs) and posterior probabilities for the CATCH intervention model coefficient obtained from the posterior distributions of the model parameters were used to compare CATCH with TAU. A similar model was planned for analysis of the secondary outcome of treatment retention. However, this outcome is reported descriptively only because of insufficient cases to allow for stable model fitting.

Results

There were 2315 unique patients eligible for inclusion (Figure 3). These patients had 3225 eligible admissions. The Table describes the characteristics of the study population. Most participants were male (1687 [72.9%]), and the mean (SD) age was 47.0 (12.4) years. There were 729 African American/Black patients (31.5%), 58 Asian patients (2.5%), 753 Hispanic patients (32.5%), 61 Native American or American Indian patients (2.6%), 810 White patients (35.0%), and 91 multiracial patients (3.9%). During the 3 years before the index admission, 1568 (67.7%) had 3 or more chronic medical conditions, and 1487 (64.2%) had at least 1 diagnosis of serious mental illness. During the same period, 1172 (50.6%) had 3 or more hospitalizations, and 1352 (58.4%) had 3 or more emergency department (ED) visits. Fewer than half (1039 [44.9%]) had received any MOUD during the 3 years before the index admission.

Figure 3. Flow Diagram of Patients Eligible for Inclusion in the Analysis.

CATCH indicates Consult for Addiction Treatment and Care in Hospitals; TAU, treatment as usual.

^aHospitalizations in dedicated psychiatric/detoxification units or jail/prison units were not eligible.

^bA total of 81 individuals (84 admissions) had a gap in Medicaid enrollment during the first 30 days postdischarge.

Table. Demographic Characteristics, Chronic Conditions, and Care Utilization Before Index Hospitalization for 2315 Patients.

Characteristic	No. (%)
Demographic characteristics
Age, y
Mean (SD)	47.0 (12.4)
Median (range)	48.8 (18.6-96.7)
Sex
Female	628 (27.1)
Male	1687 (72.9)
Race
African American/Black	729 (31.5)
Asian	58 (2.5)
Native American or American Indian	61 (2.6)
White	810 (35.0)
Multiracial	91 (3.9)
Unknown	566 (24.4)
Ethnicity
Hispanic	753 (32.5)
Non-Hispanic	1119 (48.3)
Unknown	443 (19.1)
Prior medical conditions and care utilization during the 3 y before hospitalization
Any chronic medical condition, No.a
0	167 (7.2)
1	288 (12.4)
2	292 (12.6)
≥3	1568 (67.7)
Any chronic mental illnessb
No	538 (23.2)
Yes	1777 (76.8)
Any serious mental illnessb
No	828 (35.8)
Yes	1487 (64.2)
Acute care utilization
Hospitalization admissionsc
Median (range)	3.0 (0-127.0)
Mean (SD)	7.2 (12.7)
0	562 (24.3)
1	329 (14.2)
2	252 (10.9)
≥3	1172 (50.6)
ED visitsc
Median (range)	4.0 (0-1274.0)
Mean (SD)	12.8 (46.8)
0	469 (20.3)
1	270 (11.7)
2	224 (9.7)
≥3	1352 (58.4)
Any MOUDd
No	1276 (55.1)
Yes	1039 (44.9)

Abbreviations: ED, emergency department; MOUD, medication for opioid use disorder.

^a Chronic conditions were measured using the Chronic Condition Indicator, version 2017.1, and the Clinical Classification Software, version 2017.1.⁴⁶

^b Chronic mental illness included mood disorders, bipolar disorders, schizophrenia, and other mental illness conditions. Serious mental illness was limited to manic episodes, bipolar disorder, major depressive disorder, schizophrenia, and other psychotic disorders.

^c There may have been overlap between hospital and ED admissions that could not be identified within the Medicaid claims data.

^d MOUD included: prescription for buprenorphine (including extended-release injectable), prescription for extended-release naltrexone, or a visit to an opioid treatment program.

eTable 2 in Supplement 3 presents characteristics of eligible patients during the study in the TAU, CATCH intervention, and maintenance periods. An opioid condition was the primary diagnosis in 10.2% of admissions. The median length of stay was 4 days during CATCH, and 5 days during TAU. Among the study population, 39.0% had admissions only during TAU and 54.0% had admissions only during the intervention (eFigure 1 in Supplement 3).

Primary Outcomes: Postdischarge MOUD Initiation and Engagement

In the TAU condition, for 89 eligible admissions (6.7%), patients initiated MOUD within 14 days following hospital discharge, and 70 (5.3%; 78.7% of those initiating treatment) were engaged in MOUD for 1 month. In the CATCH condition, 209 eligible patients (11.0%) initiated MOUD following discharge, and 141 (7.4%; 67.45% of those initiating treatment) were engaged in MOUD for 1 month. The rates of MOUD initiation, engagement, and retention are shown for each study period in Figure 4.

Figure 4. Proportion of Admissions With Treatment Initiation, Engagement, and Retention by Study Period Among the 3225 Eligible Admissions Included in the Analysis.

CATCH indicates Consult for Addiction Treatment and Care in Hospitals; TAU, treatment as usual.

In the bayesian analysis of MOUD initiation (aim 1a), a posterior distribution was generated for model parameters (eMethods in Supplement 3), and the posterior mean for the CATCH intervention coefficient was 2.074 (95% CrI, 0.514-4.002) with a posterior probability of 0.995 that the CATCH coefficient was positive. This corresponded to an odds ratio of 7.96 (95% CrI, 1.67-54.71), with the CATCH condition having higher odds of initiating MOUD. For the MOUD engagement outcome (aim 1b), the posterior mean for the MOUD engagement coefficient was 1.932 (95% CrI, 0.092-4.175), with a posterior probability of 0.979 that the MOUD engagement coefficient for CATCH was positive. This corresponded to an odds ratio of 6.90 (95% CrI, 1.10-65.04), with the CATCH condition having higher odds of MOUD engagement. eFigure 2 in Supplement 3 shows the posterior distributions of the CATCH treatment effect coefficients. There was evidence of a positive autocorrelation for the initiation and engagement outcomes across periods.

Secondary Outcome: MOUD Treatment Retention

The rate of 6-month MOUD retention was lower than rates of MOUD initiation and engagement and showed less variation by treatment condition. In TAU, 32 eligible patients (2.4%) were retained in MOUD at 6 months compared with 61 (3.2%) in the intervention.

Discussion

To our knowledge, this was the first large randomized prospective clinical trial of a hospital addiction consultation intervention, and it found that the CATCH program significantly increased postdischarge MOUD initiation and engagement in 6 public hospitals. The odds of MOUD treatment initiation were nearly 8 times higher with CATCH than in TAU, indicating a robust treatment effect. However, the rate of MOUD initiation, even during CATCH, was only 11%, and less than 8% of the population was engaged in treatment for 30 days, reflecting the challenges of reaching a non–treatment-seeking population.

Patients face multiple barriers to engaging in MOUD that hospital-based programs have limited ability to influence. Prior research on MOUD initiation and engagement in ED settings indicates that initiating buprenorphine in the ED can increase rates of postdischarge treatment.^48,49 Nonetheless, in a recent national implementation-effectiveness trial, rates of 30-day engagement remained relatively low (16.3%), even in the intervention condition.⁵⁰ While access to MOUD is better in New York, New York, than in much of the US, including for publicly insured populations,⁵¹ barriers remain, including stigmatization of medication-based treatment, financial or insurance issues, program requirements, transportation, and patient dissatisfaction with treatment options.^52,53 This is reflected in national data that show low rates of MOUD treatment, particularly among the racial and ethnic minority and financially disadvantaged populations that are well represented in our study,^3,4,5,8,54 as well as suboptimal retention among those who do engage in treatment.⁵⁵ New approaches to community-based MOUD services, such as flexible, low-barrier treatment programs⁵³ and robust patient navigation,⁵⁶ as well as improvements to the quality of care (eg, providing adequate medication dose),^57,58 may be needed to increase rates of MOUD following hospitalization.

Patients in our sample had a high burden of co-occurring medical and psychiatric illness. Perhaps reflecting the complexity and severity of these comorbidities, acute care utilization, including ED visits and hospital admissions, was high. Prior research indicates high rates of death following acute care episodes among OUD populations,^14,59,60 and future analyses of our sample will examine postdischarge overdose-related mortality.

While MOUD initiation and engagement are key indicators of effectiveness, there are other potential benefits of hospital addiction consultation services. These include more effectively treating acute pain and withdrawal,^61,62 avoiding patient-directed discharges,^63,64 integrating harm reduction and patient-centered care,^23,65,66,67 combating stigma,⁶⁸ and improving hospital systems for substance use care.²⁹ Our study of OUD treatment focuses on a subset of hospitalized patients with substance use. Alcohol use disorder in particular is prevalent in hospitalized patients and is associated with high morbidity and mortality.^19,69,70 Stimulant-related and polysubstance overdose deaths have skyrocketed in recent years,^71,72 and polysubstance use is common among hospitalized patients,³⁵ including our study population. Addiction consultation services may be effective for increasing postdischarge treatment initiation for nonopioid SUDs,^35,73,74 but this needs further study.

As a pragmatic trial, our study may inform the effectiveness of addiction consultation services in safety-net hospital settings and is less subject to bias due to factors such as increased attention and contact with research staff.^75,76 The rollout of CATCH in a large public hospital system faced challenges, including hiring delays and staff turnover,⁷⁷ requiring hospitals to adapt in real time. The bridge clinic was envisioned as a freestanding unit, but most CATCH hospitals lacked the physical space and resources to fully implement this and instead provided MOUD through their existing addiction treatment programs. All hospitals faced major disruptions due to the COVID-19 pandemic, which had its onset during the study period and posed unprecedented challenges to hospital staffing and postdischarge treatment access.^78,79 H+H continued the CATCH program during the pandemic, suggesting they considered it valuable for patient care.

Strengths and Limitations

While our study had multiple strengths, including being a rigorous trial conducted in safety-net hospitals within clinical conditions, it had some limitations related to the study design. The analysis was limited to cases identified in Medicaid data, such that any MOUD received before and after hospitalization could be captured. Patients who were not included because they were Medicaid/Medicare dual-eligible, privately insured, or uninsured may have had different outcomes. A total of 3.5% of patients (81 of 2315) had a gap in coverage within 30 days of hospitalization, which could have led to some underestimation of the intervention effect. The sample size after applying all eligibility criteria was smaller than anticipated in the original power calculations, which may have contributed to lack of precision in estimates of the intervention effect. Medicaid data do not capture many details of care provided in the hospital (eg, receipt of CATCH services or orders for MOUD), and administrative data sources have limited and frequently missing data about patient characteristics, including race and ethnicity, that can influence MOUD treatment.^3,4,5,6,7,8 Discharge diagnosis codes may not capture all cases of OUD, although our prior analysis found that Medicaid claims had a 90% sensitivity for detecting OUD in the study hospitals.⁸⁰ The study design called for analyzing outcomes at the hospital level; thus, our sample included patients in the intervention condition who may not have received CATCH services. In a stepped-wedge design, temporal effects may be poorly measured and controlled.^81,82 While our analysis compared TAU with CATCH during each 3-month period, there may have been changes to the hospital or postdischarge environment that were unmeasured. Our analysis included a period factor that had flexibility to capture COVID-19 effects that a parametric linear time trend may have missed.⁸³ Finally, while our study benefitted from including a Medicaid population with substantial racial and ethnic diversity, it was limited to New York, New York, and the effectiveness of addiction consultation programs in different areas and patient populations may differ.

Conclusions

Hospitals have an important role to play in identifying and treating patients with SUD. This randomized clinical trial supported the effectiveness of hospital addiction consultation services for increasing initiation and initial engagement in MOUD. MOUD is highly effective, but treatment rates remain lower than what is needed to reverse the ongoing trend of increasing opioid-involved overdose deaths.² Interprofessional addiction consultation services may improve hospital-based care, but they are just one element of the SUD treatment landscape. Further efforts to improve and evaluate hospital-based and community-based service models that can improve rates of MOUD treatment and address the needs of patients who use substances other than opioids are still needed.

Supplement 1.

Trial protocol

Supplement 2.

Statistical analysis plan

Supplement 3.

eTable 1. ICD and National Drug Codes Used in the Study

eTable 2. Characteristics of eligible admissions included in the analysis by study period

eTable 3. Analyses including Medicaid-Medicare dual eligible individuals; demographic characteristics, chronic conditions, and care utilization prior to the index hospitalization

eTable 4. Analyses including Medicaid-Medicare dual eligible individuals; characteristics of eligible admissions included in the analysis, by study period

eMethods. Details on Bayesian Modeling

eFigure 1. Proportion of hospital admissions occurring during the TAU, CATCH, and Maintenance conditions

eFigure 2. Posterior distribution of the probability of MOUD a) initiation and b) engagement

eFigure 3. Analyses including Medicaid-Medicare dual eligible individuals; posterior distribution of the probability of MOUD a) initiation and b) engagement

Supplement 4.

Data sharing statement