Depression, anxiety, pain and chronic opioid management in primary care: Type II effectiveness-implementation hybrid stepped wedge cluster randomized trial

Abstract

Even though current prescribing trends reveal that high-dose opioid prescribing and opioid prescribing in general has decreased, sustained efforts are needed to help providers adopt and maintain safe prescribing behaviors. The purpose of this four-year type 2 effectiveness-implementation hybrid stepped wedge cluster randomized trial is to: (1) compare the clinical and cost effectiveness of electronic medical record-based clinical decision support [EMR-CDS] versus additional integrated, collaborative behavioral health [EMR-CDS + BHI-CCM] for opioid management of patients with co-morbid chronic non-cancer pain with depression or anxiety; and (2) examine facilitators and barriers to implementing these interventions within 35 primary care clinics in a integrated delivery health system. The EMR-CDS alerts providers to employ opioid risk mitigation and safe prescribing practices at the point of care. The BHI-CCM consists of primary care embedded community health workers for case management; licensed clinical social workers for cognitive behavioral therapy, and a clinical pharmacist for medication management who provide care management via telemedicine (virtual video or audio only visits) under the guidance of a consulting psychiatrist. The primary outcome is reduction in the percentage of patients with average daily opioid dose ≥50 mg morphine equivalent. Secondary outcomes include changes in service utilization, patient reported outcomes and processes of care. The investigators anticipate that study results will elucidate the role of technology versus care team optimization in changing opioid prescribing behaviors. The investigators further anticipate that integrated mental/behavioral health care will increase value-based care and the efficiency with which guideline concordant care is delivered.

Introduction

In 2016, the Center for Disease Control (CDC) released evidence-based practice guidelines to help providers identify risky opioid user behaviors, heighten provider awareness of prescribing practices and promote use of risk mitigation strategies [1]. Even though prescribing trends between 2006 and 2017 have shown that high-dose opioid prescribing and opioid prescribing in general has decreased through 2017 [2], sustained efforts are needed to help providers adopt and maintain safe prescribing behaviors.

We must find ways to accelerate practice guideline adoption in primary care in the face of the opioid crisis. Successful implementation of guidelines requires surmounting barriers related to: (1) provider personal factors (e.g. knowledge); (2) guideline factors (e.g. plausibility, accessibility); and (3) external factors (e.g. organizational resources, social/clinical norms) [3]. Guidelines that are evidence-based, plausible, goal-oriented, user-friendly and easily accessed at the point of care have a greater likelihood of being adopted quickly. Organization level use of computerized decision support systems, reminders and standing orders as well as standardization of processes, procedures and protocols within the context of quality management improves practice adherence. Finally, inter-professional, multidisciplinary collaborations among health care professionals can facilitate guideline implementation and adherence [3].

Given the high prevalence of major depression co-occurring with chronic opioid use [4–6], risk mitigation must incorporate proactive diagnosis and management of psychiatric conditions, which can complicate management of chronic pain. Several studies demonstrate that collaborative interdisciplinary care is effective for improving management of depression, anxiety, pain and substance abuse in primary care [7–14]. Collaborative care improves physical and mental health outcomes, quality of life, functional status, satisfaction with care, and can provide good economic value with lower costs.

Integration of collaborative mental/behavioral health services has the potential to accelerate implementation of safe opioid management in primary care. The aims of this type 2 effectiveness-implementation hybrid stepped wedge cluster randomized trial are to: (1) compare the clinical effectiveness of CDC guideline concordant electronic medical record clinical decision support for safe pain management [EMR CDS guided care] among patients with co-morbid depression or anxiety and on chronic opioid therapy for non-cancer pain versus additional integrated behavioral health collaborative care management [BHI-CCM]; (2) examine the cost-effectiveness of EMR CDS guided care versus additional BHI-CCM; and (3) qualitatively examine facilitators and barriers to implementing the multi-component interventions using the Consolidated Framework for Implementation Research (CFIR) [15].

Methods

This study was approved by the Ochsner Health Institutional Review Board and is registered at clinicaltrial.gov (NCT03889418).

Study Hypotheses

The main study hypothesis is that a higher proportion of patients who receive additional BHI-CCM compared to EMR CDS guided care only will have greater decreases in the rate of high-dose opioid prescriptions and higher rates of receiving opioid risk mitigation care (e.g. urine drug screens, naloxone, functional assessments specialty care). The second hypothesis is that patients who participate in BHI-CCM will receive more cost-effective care compared to patients who only receive EMR CDS guided care and have greater reductions in inpatient hospitalizations and/or emergency department utilization. The third hypothesis is that intervention adaptability to local contexts (e.g. geographic location), clinic readiness for implementation (leadership engagement; available resources; care team knowledge about clinical guidelines), and physician champions will influence uptake of the intervention components in primary care.

Study Setting

Ochsner Health, based in New Orleans, is the largest integrated delivery health system within the state of Louisiana with 40 owned, managed and affiliated hospitals, more than 100 health centers and urgent care centers, approximately 25,000 employees, and over 1,300 employed physicians in more than 90 specialties and subspecialties [16].

Study Design

This study meets the criteria for type 2 effectiveness-implementation hybrid study design [17] and employs a modification of the stepped wedge cluster randomized clinical trial design. By the time this clinical trial commenced in April 2019, Ochsner Health had already implemented the EMR CDS systemwide by October 2017 as the standard of care (usual care). The BHI-CCM intervention is an adaptation of the Improving Mood-Promoting Access to Collaborative Treatment (IMPACT) model [7] that employs a team of community health workers, licensed clinical social workers, a clinical pharmacist and consulting psychiatrist.

For pragmatic reasons, this study employed the stepped wedge cluster randomized clinical trial design to scale up the BHI-CCM intervention in a stepwise fashion across the healthcare system. A total of 35 primary care clinics within Ochsner Health across five geographic regions of Southeast Louisiana are participating in the study. All clinics within a given region operate under the same management team, have joint provider/staff meetings, and share clinical resources. It was not practical to individually randomize phased implementation of opioid BH-CCM on the clinic/provider/patient level. Therefore, the regions were randomized using a computer-generated random number. Implementation of the BHI-CCM intervention required 15 months to scale up in 3 to 4-month intervals across the health system (Figure 1) between April 2019 and July 2020. Within each region, the first 3–4 months of BHI-CCM implementation constituted a transition phase during which the principal investigator and behavioral health intervention team conducted, as part of routine clinic staff meetings, 20-minute in-person and virtual education sessions on BHI-CCM for depression and anxiety and problem solved the logistics of referring patients to the BHI-CCM team.

Figure 1.

Randomization of phased implementation of collaborative care model in 35 primary care clinics across 5 geographic regions of the health system

*Electronic medical record clinical decision support (EMR CDS)

†Collaborative care model (CCM)

Provider and Patient Eligibility

The study interventions target primary care providers (Internal Medicine or Family Medicine) who are practicing in the targeted clinics. Patients who meet the following inclusion criteria are eligible for the study: (1) age 18 and older, (2) have a primary care physician at any of the study clinics, (3) receive opioid prescriptions (long-acting opioids or >3 months short-acting opioids) for chronic non-cancer pain, and (4) have a diagnosis of depression or anxiety. Patient exclusion criteria include: (1) age <18 years, (2) severe mental illness (e.g. schizophrenia; bipolar disorder, addiction disorder) that would interfere with participation in the study, (3) active cancer or undergoing cancer treatment, (4) chronic cancer-related pain, (5) having a terminal illness or (5) receiving hospice care.

Patient Identification and Eligibility Screening

The chronic opioid health maintenance registry within Epic electronic health record system is the primary source for patient identification [18]. This registry is linked to a best practice alert (BPA) for PHQ-4 depression/anxiety screening [19], which had already been incorporated into the Medical Assistant patient rooming standard in primary care prior to this study. The PHQ-4 BPA triggers if there is no documentation of screening recorded within the prior 6-months. A PHQ-4 score of ≥3 triggers a BPA which prompts primary care physicians (PCP) to consider placing a BHI-CCM referral order if deemed clinically appropriate. PCPs can also place referral orders to the team without the BPA. Patients on the registry also receive brief communications about the study either via postal mail or MyOchser patient portal and can contact the study research coordinator for more information. The BHI-CCM team clinical pharmacist first verifies that referred patients meet the eligibility criteria for chronic opioid therapy. The community health worker then conducts a telephone interview to screen for study eligibility using the PHQ-9 and GAD-7 questionnaires [20–21] if not already documented by the primary care team within 2-weeks prior to referral. Patients must have moderate to severe symptoms of depression or anxiety defined as a score ≥10 on the PHQ-9/GAD-7 questionnaires in order to be eligible for study enrollment.

Consent

The Ochsner Institutional Review Board (IRB) provides oversight and approved the trial procedures. We did not obtain explicit consent from the PCPs. Their receipt of the study information along with no request to opt out constituted informed consent and is consistent with the goals and methods of pragmatic studies. The IRB determined that the study posed minimal risk as intervention activities involve clinical services already available to Medicare patients (CPT 99492, 99493, 99494, 99484). A waiver of signed informed consent and the Health Insurance Portability and Accountability Act authorization (HIPAA) was granted. Verbal consent and authorization are documented within the patients’ medical record as part of their screening for study eligibility. Requirements for study participation includes patients’ express understanding that the BHI-CCM team consists of health professionals working with the primary care team, agreement to participate in the counseling intervention via telephone or video and willingness to activate patient portal accounts.

Interventions

EMR CDS Guided Care for Opioid Management (Usual Care).

The EMR CDS guided care component went live across the health system as the standard of practice in October 2017. EMR CDS is based on Epic electronic health record system’s Healthy Planet patient registry tool and related population health maintenance workflows already employed by Ochsner primary care. Details of the intervention and baseline 12-month evaluation of its clinical effectiveness are described in previous publications [18, 22]. In brief, the health maintenance tool displays whether patients are up to date on chronic opioid therapy risk mitigation per CDC guidelines. Providers are prompted to complete the Opioid Risk Tool (ORT), PEG 3-Item Pain Scale and depression/anxiety screening (PHQ-4) [19, 23–24]. The ORT score stratifies patients for risk of opioid misuse/abuse [23]. The PEG-3 assesses pain intensity and interference with enjoyment of life and general activity [24]. The ORT score, morphine equivalent of the opioid dose prescribed, and hyperlinks to the Louisiana pharmacy drug monitoring program data and current pain management agreement are visible in the medication order composer. An Epic banner appears in the charts of patients with medium to high risk (as per the ORT) to alert providers of existing opioid management agreements.

Opioid Behavioral Health Collaborative Care Management (opioid BHI-CCM):

The opioid BHI-CCM intervention is an adaptation of the Improving Mood-Promoting Access to Collaborative Treatment (IMPACT) model based on the study team’s prior work in collaborative care management of depression and anxiety in primary care [7, 25]. The core components of integrated collaborative care includes routine screening of patients for psychiatric conditions; patient education and self-management support; medication management; clinical monitoring of response to treatment; psychotherapy; standardized follow up; formal stepped care for systematic adjustment of care plans until treatment goals are achieved; and physician supervision [11]. BHI-CCM is an additional component of the overall opioid management strategy delivered via telemedicine (virtual video through MyOchsner patient portal) or telephone. All BHI-CCM team care management activities are documented in the electronic medical record.

The opioid BHI-CCM team is composed of community health workers (CHW), licensed clinical social workers (LCSW), a clinical pharmacist (PharmD), and a consulting psychiatrist. The CHW conducts baseline assessments to determine study eligibility. The PharmD reviews and reconciles active medication lists through Epic Care Everywhere; and assesses medication side effects, drug interactions and adverse events prior to the LCSW baseline assessment. The LCSW conducts assessments to ascertain whether patients meet the DSM criteria for depression/anxiety or more complex conditions. The CHW-PharmD-LCSW team meets once a week with the psychiatry consultant to present new study participants’ baseline assessments, discuss follow-up cases, and review treatment plans. Patients with complex mental health conditions that are not routinely managed solely by primary care (e.g. bipolar, schizophrenia, substance abuse) are referred to specialty care for therapeutic management if they are currently not under the care of a psychiatrist or an addiction specialist at the time of referral to the program. The BHI-CCM team provides general recommendations about antidepressant medication titration, risks of opioid misuse/abuse, drug-drug interactions and counseling interventions to the participants’ primary care provider. Decisions about prescribing, changing or discontinuing opioids or other controlled substances are made per the discretion of the prescriber.

Follow-up plans for opioid BHI-CCM care management for depression/anxiety/chronic pain are tailored to individual patient needs. The CHW provides case management, care coordination and peer support under the supervision of the LCSW who meets daily with the CHW. The LCSW provides cognitive behavioral therapy, acceptance and commitment therapy and/or mindfulness training. These therapy models have been proven effective for managing depression, anxiety and chronic pain [26–28]. The LCSW also assesses adherence to treatment plans; reviews symptom progress; reviews personal goals and helps patients problem solve if goals were not met; updates assets/barriers/supports to recovery; as well as monitors for pain control and addiction behaviors. The pharmacist re-enforces or adjusts antidepressant treatment plans and/or recommendations for naloxone (if indicated). Acute care treatment recommendations are sent to PCPs via carbon copying notes documented in the electronic medical record with orders attached for co-signing if needed.

During the acute phase of active symptom management, patients receive counseling with the LCSW weekly or at least a minimum of every four weeks (via virtual video or audio visits).. The frequency with which patients are seen is tailored to individual patient preferences as agreed upon between the patient and the LCSW. The protocol requires that patients be seen at a minimum frequency of every 4 weeks because the PHQ9/GAD7 is only measured every 4 weeks. The total number of counseling sessions for an individual patient depends on the time it takes him/her to achieve a “stable” status of mild symptoms of depression/anxiety (PHQ-9/GAD-7 score <10); or, more ideally, complete resolution of symptoms (score <5) – defined as 3 scores each measured at least 4 weeks apart. This measurement cycle translates into at least 12 weeks or 3 months of stable symptoms. Thereafter, if clinically appropriate, patients are transitioned from the acute phase of management (frequent contact) into a maintenance phase of remote monitoring (less frequent contact) whereby the CHW assesses PHQ-9/GAD-7 scores every three months for an additional nine months. If patients have PHQ9/GAD7 scores >=10 during the maintenance phase, the LCSW assesses whether patients should transition back into acute symptom management versus remain in the maintenance phase of remote monitoring.

Recruitment and Retention:

We will employ several approaches to keep PCPs engaged in the study: (1) structured EMR documentation of patient care progress; (2) engagement of physician champions and leaders to encourage their peers to participate; (3) periodic program updates at provider meetings; and (4) educational 1-page newsletters disseminated quarterly via email. Medical Assistants are the front-end care team members who screen for depression and anxiety symptoms with the PHQ4 as part of their rooming standard for Medicare patients and patients on the chronic opioid therapy registry. The study team will submit employee recognitions to Human Resources for Medical Assistants who most consistently complete the PHQ4.

Given the sensitive nature of chronic pain and co-morbid psychiatric conditions, we anticipate that the most effective form of recruitment in the opioid BHI-CCM arm of the study will be through physician referral. Patients may also self-refer by opting-in (i.e., contacting the research staff to express their interest in participating). To introduce the program to potential patients, the research team sends mailings of the program flyer (i.e. BHI Team Flyer) and/or MyOchnser notifications to patients who have an upcoming primary care appointment within the next 60 days. Individuals who receive the flyer or MyOchsner notification must contact the research team for more information. If the patient responds that they are not interested, the EMR logs the response, and they are filtered out of further portal research communications for the study.

All clinical services rendered are incorporated into routine clinical operations. Patients receive automated telephone reminders of upcoming visits with the LCSWs, complete routine clinic pre-visit questionnaires and registration processes. Patients are not charged for services rendered by the opioid BHI-CCM team. The health system primary care leadership opted not to charge patients receiving care from the BHI-CCM team for the duration of this study because: (1) only Medicare covers this service and there is a clear need that extends other patient populations; and (2) the information gleaned from this study will help better quantify the cost of program implementation and potential return on investment as well as assess implications for BHI-CCM as value-based care model. A variety of approaches are being used to retain patients including offering convenient options of telephonic and virtual visits. We are also very clear from the initial screening session onward about what we expect from patients and what they can expect from the BHI-CCM team (including co-management and communication of care plans with their PCPs via the EMR). Keeping patients well informed enhances adherence to the BHI-CCM intervention.

Primary and Secondary Outcome Measures

All outcome measures, data sources, and schedule of data collection are detailed in Table 1. The primary outcome is the reduction in the percentage of patients with average daily opioid dose ≥50 mg morphine equivalent daily dose (MEDD) between study groups (EMR CDS guided care vs. EMR CDS+ opioid BHI-CCM). All other outcomes described in Table 1 are secondary outcomes for the purposes of this analysis. Responses to patient reported outcome measures are documented via Epic’s flowsheet tool to facilitate capturing this information as structured data. The CHW administers the PHQ-9 [20], GAD-7 [21], PROMIS-10 [29] and Epic’s social determinants of health questionnaire. The LCSW administers the PEG-3 (Pain, Enjoyment, General Activity) [30–31] to assess therapeutic response of chronic pain management and the Current Opioid Misuse Measure (COMM-9) [32] to monitor for signs of aberrant behaviors. Psychometric properties and domains measured are summarized in Table 2.

Table 1.

Primary and secondary outcome measures

	Outcome Measure	Data Source	Frequency Administered (Study Months)	Frequency EMR Data Extraction
Primary Outcomes
Morphine Equivalent Daily Dose	% prescribed high dose (≥50 mg)	EMR†	…	Quarterly
	% prescribed very high dose (≥90 mg)	EMR	…	Quarterly
Secondary Outcomes
Service Utilization	Inpatient hospitalization	EMR	…	Quarterly
	Emergency department use	EMR	…	Quarterly
Process of care documented	% Specialty referrals	EMR	…	Quarterly
	% Pain agreements	EMR	…	Quarterly
	% Urine drug screening	EMR	…	Quarterly
	% Naloxone documented	EMR	…	Quarterly
	% Non-opioid prescriptions	EMR	…	Quarterly
Patient Reported Outcomes*	PHQ-9 depression screen	EMR	0,1,2,3,4,5,6,9,12,16	Quarterly
	GAD-7 anxiety screen	EMR	0,1,2,3,4,5,6,9,12,16	Quarterly
	PROMIS-10 Global Health	EMR	0, 3, 6, 12	Quarterly
	Social Determinants of Health	EMR	0 (and as needed)	Quarterly
	PEG-3	EMR	0, 1, 2, 3, 4, 5, 6	Quarterly
	COMM-9	EMR	0, 1, 2, 3, 4, 5, 6	Quarterly
Provider Survey	Provider Experience with Mental Health Care Management	Redcap survey	0, 12, 24, 36	Annually

^*Patient reported outcome measures are administered by the community health worker(PHQ-9, GAD-7, PROMIS-10; social determinants) and licensed clinical social worker (PEG-3; COMM-9) on a schedule that begins with study enrollment. The schedule assumes that patients will be in the acute phase of treatment for 6-months followed by 9 months of remote monitoring.

^†Electronic medical record (EMR)

Table 2.

Patient Reported Outcome Measures

Questionnaire	Number of items	Reliability	Validity	Domains
PHQ-9	9	Cronbach alpha 0.89	Construct Criterion	DSM IV criteria for depression
GAD-7	7	Cronbach alpha 0.92; ICC* 0.83	Construct Criterion	DSM IV criteria for generalized anxiety disorder
PROMIS-10	10		Content	Physical function, fatigue, pain, emotional distress, social health, quality of life
PEG-3	3	Cronbach alpha 0.73 to 0.89	Construct	Pain, enjoyment, general activity
COMM-9	9	ICC 0.82	Criterion	Negative affect, current aberrant drug behvaior
Social Determinants of Health	14	…	…	Food insecurity, housing, financial resource strain, social connections, physical activity, stress, transportation needs

^*ICC = intraclass correlation

Data Safety and Monitoring Plans

Due to low risk posed by the study intervention, the data safety and monitoring plan relies on periodic monitoring by the principal investigator in conjunction with the investigative team and the Ochsner IRB. Reviews of hospitalizations, emergency department visits and deaths among active subjects are conducted by a psychiatrist every month for the duration of the study.

Anticipated Effect Size, Sample Size Justification and Planned Statistical Analysis

The primary outcome proposed is the reduction in average daily opioid dose ≥50 mg morphine equivalent dose following enrollment in the intervention (EMR CDS + opioid BHI-CCM program). A power analysis was carried out based on the type 2 effectiveness-hybrid clinical trial to determine sample size required to detect a difference in outcome between the intervention and usual care groups (EMR CDS only). Notably, the stepped wedge design is only being used to pragmatically implement the intervention. A parallel control group from the usual care arm of the study is being constructed for comparison of outcomes with the intervention group. Control patients are selected based on study inclusion/exclusion criteria and are matched to enrolled patients by select characteristics. Controls are being selected from patients seen within participating clinics in the same time period as the intervention group.

In a prior study carried out in the Ochsner Health System on the clinical effectiveness of EMR CDS for opioid prescribing, approximately 20% of patients on chronic opioid therapy were observed to have an average daily opioid dose ≥50 MEDD [22]. Using equal baseline response rate (RR) for the usual care and BHI-CCM groups, a total sample size of 372 patients are required to detect an odds ratio of 2.25, corresponding to a 50% decrease in the BHI-CCM group (RR=10% vs 20% in usual care), with 80% power. The total sample size amounts to an average of 11 patients in each study group within each of the 35 clinics. In accordance with the cluster randomized design, the variance inflation factor 1 + (m − 1)ρ is incorporated, where m is the average cluster size (m=11) and ρ is the intra-cluster correlation coefficient (ρ≤0.03). Applying this inflation factor and rounding up to allow for equal group sizes across clusters results in a total sample size of 490 patients - 245 per group. Accounting for a potential 15% attrition rate, 578 total patients - 289 per group - are needed to attain the target sample size required to detect a 50% decrease in proportion of patients with average daily opioid dose ≥50 MEDD in the BHI-CCM group compared to the usual care group.

At end of study, generalized linear mixed models will be used to assess average daily opioid dose defined as a binary indicator of ≥50 mg morphine equivalent dose and as a continuous measure. The outcomes will be assessed as pre-enrollment to post-enrollment changes in the EMR CDS + opioid BHI-CCM enrolled patients versus those in the usual care setting. In typical fashion, covariate adjustment will be used in outcomes models to account for group imbalance in patient characteristics. We will also incorporate cluster and time, via fixed and random effects, in the multivariable analyses to account for potential effects of these design factors on the outcomes.

Aim 2 Cost Effectiveness Evaluation

All analyses will follow the guidelines on cost-effectiveness analysis conducted alongside clinical trials and economic modeling [33]. Because of the chronic, recurrent nature of opioid management, we will consider variable time horizon from trial duration to lifetime. We will develop a semi-Markov model to reflect the cycles of health status level of risk reduction (> 50 vs. <=50mg MEDD), abstinence (discontinuation of opioids), and relapse as well as death among patients treated with opioids. First, we will derive model parameters from the trial data where possible. For example, trial data will provide the transition probabilities and health utility of different health statuses. We will collect information on resource consumptions from electronic medical records which are linked with claims data. Then, we will conduct a literature review for model parameters. In addition, we will project costs and outcomes over the expected duration of treatment and its effects to ensure that all relevant differences in future outcomes of the treatment comparisons will be captured. Specifically, the cost-effectiveness analysis will use the following parameters: 1) Target population–patients on chronic opioids (long-acting or short-acting); 2) Comparators - We will compare costs for clinics with access to both the EMR CDS guided care and opioid BHI-CCM team versus clinics that only have access to the EMR CDS guided care; 3) Perspectives – Societal, health plans/payers, and health providers (Ochsner Health System); 4) Time horizon - 2 years, 5 years, and 10 years as well as the lifetime (30 years); 5) Outcomes - % reduction in patient prescribed high dose (% <=50mg MEDD), PEG-3 score, PROMIS-10 quality adjusted life year (QALY); 6) Cost measures - direct medical costs (inpatient, outpatient, pharmacy, lab, etc.) and program costs (material, training, staff unreimbursed time, etc.); 7) Discount rate - 3% (5% as sensitivity analysis); and 8) Results - costs and outcomes will be compared.

If one program does not clearly dominate the others (lower costs and better outcomes), then we will calculate incremental cost effectiveness ratios (ICERs) using the EMR CDS guided care program as the base comparator. An incremental cost-effectiveness ratio (ICER) is defined as $\frac{{\Delta }_c}{{\Delta }_e}$, and is a comparison of costs and effectiveness between two strategies (S1 and S2), where the estimated components are defined as Δ_c = Costs_S1 - Costs_S2, and Δ_e = Effectivness_S1 - Effectiveness_S2. The components are point estimates and have an associated uncertainty that is described by 95% confidence intervals. Primary ICER measure will be cost per % reduction below 50mg MEDD. Secondary measures will be cost per QALY using the PROMIS-10 or PEG-3 score. A decision to implement the program is based on a threshold willingness-to-pay (WTP). The WTP threshold for a given healthcare system may be $50,000 per QALY for example [34]. Therefore, programs with an ICER below $50,000 per QALY would be desirable. We will evaluate the robustness of results with changes in model inputs, using one-way and probabilistic sensitivity analyses. One-way analysis sensitivity analysis will test key assumptions such as 5% discounting rate, effectiveness of the additional CCM program, and key cost items (rates of inpatient stay, length of stay, and emergency room visits), as well as the WTP threshold of $100,000 per QALY.

Probabilistic sensitivity analysis will be executed via Monte Carlo simulation using relevant parameter distributions (e.g., population characteristics in the proposed trial). Non-parametric bootstrapping will be used to estimate parameter uncertainty including ICER 95% CI, CE scatter-plot and CE acceptability curve, which will be used to further interpret the results.

Aim 3 Process and Formative Evaluation

The Consolidated Framework for Implementation Research (CFIR) is composed of five major domains and 37 constructs [15]: (1) intervention characteristic (e.g. evidence strength and quality; complexity; cost); (2) outer setting (e.g. patient needs and resources; external policies and mandates; peer pressure); (3) inner setting (e.g. implementation and learning climate; tension for change; leadership engagement; available resources); (4) characteristics of individuals involved (e.g. knowledge, beliefs, self-efficacy); and (5) process of implementation (e.g. champions; opinion leaders; planning, engaging, executing and evaluating).

We will gather longitudinal qualitative data on the implementation process from meeting minutes, semi-structured interviews with key stakeholders (health system leaders; primary care providers and staff; Epic support team; intervention team) and comments from provider surveys collected over the course of this study. Evaluation of the implementation process is divided into 3 phases: (1) January 2017 to October 2019 - design and implementation of the EMR CDS; (2) January 2019 to June 2020 – design and implementation of the opioid BHI-CCM; and (3) July 2020 to December 2021 – maintenance of both interventions.

We will employ methodological and analyst triangulation to produce a comprehensive understanding of facilitators/barriers to protocol adherence for chronic opioid management. We will capture data from qualitative and quantitative data sources to permit exploration of complementary and divergent concepts regarding factors influencing protocol adherence that might otherwise not be captured from single sources. Study investigators will use the CFIR construct coding guidelines and interview guide for analysis. The coding guidelines provide definitions for each construct. The non-coded minutes, emails and interview transcripts will be entered into a software package designed to handle unstructured qualitative data. Each year, two trained analysts who are blinded to implementation outcomes will independently code meeting minutes using pre-populated CFIR construct codes [15]. The analysts will then meet to reach consensus regarding differences of opinion about specific text. Unresolved issues will be brought to the larger research team for guidance and resolution. Study investigators will meet with health system stakeholders to review examples of coded statements to make sure they accurately capture diverse perspectives of patients, clinicians, administrators, and health system leaders. We will quantify the frequency with which CFIR constructs are coded. We will then organize the data according to the implementation phase during which the data was collected (pre-implementation [before phasing in collaborative care] vs. post-implementation). The CFIR constructs that are highlighted most frequently in each implementation phase will be noted. We will further classify the data as system-, practice-, provider-, or patient-level factors that facilitate or impede implementation of the multicomponent intervention.

Results

Characteristics of participating clinics

Table 3 describes the characteristics of 35 participating primary care clinics. All clinics are located within Southeastern Louisiana in the Greater New Orleans and Baton Rouge urban areas. They are staffed by a total of 196 physicians (Internal Medicine/Family Medicine), 62 nurse practitioners, and 14 physician assistants. Overall, the clinics serve 1,011,864 unique patients among which 27.2% self-identify as black, 65.3% white and 7.5% other race. Therefore, among the 578 patients for whom this study must accumulate healthcare data, the investigators anticipate that the study population will include: 143 black patients, 343 whites, and 39 other race.

Table 3.

Characteristics of Participating Clinics in Ochsner Health (Louisiana)

Primary Care Clinics (N)		Number of Providers by Type (N)			Unique Patients by Self-Identified Race N (row %)
Location	Clinics	MD	NP	PA	Total	Black	White	Other
Region 1	11	71	13	3	455,055	127,679 (28.2%)	289,327 (63.8%)	38,049 (8.4%)
Region 2	5	33	20	6	182,019	22,027 (12.1%)	148,768 (87.1%)	11,224 (6.2%)
Region 3	4	10	2	0	44,419	11,750 (26.5%)	31,049 (69.9%)	1,620 (3.6%)
Region 4	11	61	17	3	206,907	61,332 (29.6%)	129,777 (62.7%)	15,798 (7.6%)
Region 5	4	21	10	2	123,464	52,486 (42.5%)	61,485 (49.8%)	9,493 (7.7%)
Total	35	196	62	14	1,011,864	275,274 (27.2%)	660,406 (65.3%)	76,184 (7.5%)

Discussion

This type 2 effectiveness-implementation hybrid modified stepped wedge cluster randomized trial will examine the comparative effectiveness of EMR CDS guided care versus additional BHI-CCM and the implementation strategies for each intervention.

The research team previously demonstrated that EMR CDS guided care was associated with increased rates of urine drug screening, naloxone prescribing, and referral to subspecialty care and decreased utilization of emergency departments [22]; however, the EMR CDS did not decrease rates of high dose opioid prescribing. Additionally, co-prescribing benzodiazepines was strongly associated with high dose opioid prescribing. Even though technology optimization improved some aspects of opioid risk mitigation, it did change opioid prescribing behaviors.

The BHI-CCM intervention sets clear protocols for next steps beyond opioid risk mitigation, including systematic assessments of co-occurring depression and/or anxiety that can complicate the treatment course for chronic pain management. BHI-CCM is an adaptation of the IMPACT model [7] which is included in SAMHSA’s National Registry of Evidence-based Programs and Practices. The main adaptation is the BHI-CCM team model of CHWs for health coaching and case management, LCSWs for counseling which is a required component of the intervention for the targeted population and a clinical pharmacist for medication management. The study also employs video and telephone visits to deliver the intervention and maximize scalability. There is already strong evidence that collaborative care improves outcomes for patients with depression, anxiety and/or chronic pain and has been tested in primary care settings [7–14]. Additionally, there is momentum within the study setting among health system administrators for rapid uptake of the collaborative care model for behavioral health integration. Because the study is being conducted in a real-world setting, it employs the stepped wedged cluster randomized trial design as the most pragmatic approach to implementation of the BHI-CCM arm of the study.

The study investigators anticipate several challenges with the conduct of this study related to temporal trends. First, by the time this study was initiated, health policy changes had already impacted downward trends in opioid prescribing whereby the rate for long-term prescriptions increased while the amount prescribed per person and the rate of prescribing high dose opioids (≥90 mg MEDD) and long-acting formulations decreased between 2006 and 2017 [2, 35]. Notwithstanding these trends, this study is targeting medically complex patients who are likely not to have the same outcomes if they are suffering from undiagnosed or sub-optimally treated depression or anxiety. Therefore, the study hypotheses remain the same.

Another anticipated challenge is that this study is being conducted in the middle of the Sars-COV-2 (COVID-19) pandemic. Recent national surveys substantiate rising concerns about the mental health impact of the pandemic even among people with no prior history of psychiatric conditions [36–37]. Increased stressors such as financial insecurity related to job loss or decreased salaries, grieving the death of close family and/or friends and increased social isolation remains a real concern. Under these circumstances, the CHW will be a vital team asset for helping patients navigate community resources while providing social support. Counseling sessions with the LCSW will be tailored to individual patient needs. The study intervention also utilizes strategies recommended by an international panel of experts for managing patients with chronic pain during the pandemic – specifically telemedicine and biopsychosocial management [38]. Finally, the health system where this study is being conducted has mechanisms in place to ensure continuity of care including electronic prescribing of medication and increased use of telemedicine visits.