Study setting {9}
The IBH-PC study takes place in 45 primary care practices across the United States (Figure 1). Practices represent a diverse distribution of geographic regions, population densities (urban vs. rural), patient population size, specialty (family vs. internal medicine), community health centers, federally qualified health centers, nonprofit and for-profit organizations, resident training sites, and ownership (hospital or health system, academic, private).
Eligibility criteria {10}
Study Practices
Primary care practices are eligible to participate if they:
- have at least one PCP and at least one BHP onsite (co-located),
- have at least 0.5 full-time equivalent BHPs licensed to practice independently,
- commit to maintaining onsite BHP for the duration of the study,
- provide the research team with access to electronic medical records (EMRs) to identify patients with specific medical and behavioral health conditions for recruitment,
- are willing to complete survey instruments periodically throughout the study, and
- are willing to be randomized to either the active or control arm.
For practices with more than 10% of their revenue generated by Medicare, at least one BHP per practice must be eligible to bill Medicare. Practices are ineligible if they are already undergoing, or plan to undergo, another QI initiative similar to the IBH-PC intervention or if they already have an advanced level of BH integration as evidenced by a total Practice Integration Profile score (PIP, described below) above 75.
Study Patient Participants
Patients are eligible to participate if they:
- are at least 18 years old,
- are an active patient of a participating study practice as evidenced by at least two visits in a period of 24 months for any purpose, including at least one in the most recent 6 months,
- are willing to complete three surveys over two years, and
- have both an eligible chronic medical condition and an eligible chronic behavioral health condition, or at least three eligible chronic medical conditions. Eligible chronic medical conditions include arthritis; obstructive lung disease including emphysema, chronic bronchitis or asthma; non-gestational diabetes; and heart disease manifested as heart failure or hypertension. Eligible behavioral health conditions include mood disorder (anxiety or depression), chronic pain (including headache, migraine, neuralgia, fibromyalgia, or chronic musculoskeletal pain), insomnia, irritable bowel syndrome, and substance misuse (substance use disorder, tobacco use, or problem drinking).
To determine patient eligibility, medical records are reviewed for a period of 24 months with the most recent date being within 12 months of the study start at each practice. Within this period, evidence of age, minimum number of visits, and medical and behavioral health conditions will be determined.
Evidence of medical conditions may take the form of a specific diagnosis on a problem list. Evidence of diabetes may also be indicated by 3 months of insulin or other diabetes medications (excepting metformin which is not specific to diabetes) or any hemoglobin A1C value greater than 6.4%. Evidence of heart disease may also be indicated by 3 months of cardiac medications specific to hypertension or heart failure (excepting beta-blockers and other medications with broader indications), or 3 sequential blood pressure measurements with mean systolic pressure >140 mmHg or mean diastolic pressure >95 mmHg.
Evidence of a behavioral health condition may take the form of a specific diagnosis on a patient’s problem list, 3 months of certain medications used for behavioral conditions (antidepressants, anxiolytics, opioids, antineuropathic agents, agents for alcohol use disorder or smoking cessation, etc.), or persistent failure to attain physiologic control of a medical problem evidenced as systolic blood pressure > 165 mmHg for 3 months or more or A1C > 9% for 6 months or more.
Who will take informed consent? {26a}
Eligible patient participants are identified by review of EMRs in each participating primary care practice (26). Acting as trusted third party, the DARTNet Institute (27) reviews the EMR data, identifies potential patient participants, and offers participation.
Eligible patients are mailed a letter describing the purpose of the study and that, if they choose to participate, they are requested to complete a brief set of surveys at three timepoints via web, paper, or phone. The letter also describes the process to opt-out as well as provides a phone number to call for more information on the study. The DARTNet Institute follows-up the letters with phone calls to eligible patients. Consent is administered and documented by phone, paper, or via web at the completion of the baseline set of surveys. Participating patients provide consent to link their EMRs to survey data and may discontinue participation at any time. All study materials are available in both Spanish and English.
Additional consent provisions for collection and use of participant data and biological specimens {26b}
The study consent process includes permission for additional analyses of collected data. No biological specimens are collected.
Interventions
Explanation of the choice of comparators {6b}
The active comparator (the intervention) is the use of the IBH-PC toolkit to support a practice-level change process. Use of the IBH-PC toolkit is intended to increase the delivery of highly integrated behavioral health and primary care, which is hypothesized to result in improved patient outcomes. The control comparator is co-location of a behavioral health provider within or adjacent to the primary care practice (at the same street address), without additional integration (12). Co-location was chosen as the control comparator as it is highly prevalent in the field and has emerged as a standard of care in this domain (10).
Intervention description {11a}
The intervention, the IBH-PC toolkit, is a set of implementation strategies consisting of four components:
1) an online educational curriculum,
2) a structured, team-based, practice redesign and implementation workbook,
3) remote QI coaching services for internal QI facilitators, and
4) an online learning community.
The purpose of the IBH-PC toolkit is to guide and support primary care practices through a practice change process as they work to increase their degree of behavioral health integration through, for example, improvement of screening, case identification, management, or follow-up. Progress through the toolkit and use of its components are expected to be variable among participating practices, taking from 9 to 24 months.
Online Education Curriculum
The IBH-PC toolkit includes an asynchronous online curriculum on evidence-based concepts of IBH and methods of applying them. The curriculum includes one interprofessional course for all practice member roles and seven individual courses, targeting different practice member roles (practice manager, BHP, PCP, nurse, staff, IBH-PC facilitator) Courses take approximately 4 – 14 hours to complete, depending on the practice member role. The online curriculum is housed at the Arizona State University and is made available through an online learning management platform (28). It is intended, but not required, that all practice members complete the courses appropriate for their roles.
Practice Redesign and Implementation Workbook
The redesign and implementation process is available as a workbook format (.pdf format) that is posted in an online, shared workspace. Practices are encouraged to establish interdisciplinary project teams, with a practice member serving as project leader and an on-site, internal facilitator. The workbook directs the project team through four stages: study start up and leadership engagement, planning the scope and boundaries of workflow redesign using a Lean Management approach (29-31), redesigning workflow with recommended tactics, and implementing those changes in the practice. It is intended, but not required, that the project teams complete all stages sequentially.
Remote Quality Improvement Coaching
To support the project teams through the redesign process, external QI coaches work with individual practices. Each study practice is assigned a team of two coaches who meet with them via phone or web-meeting throughout the practice redesign process (up to 24 months). The remote QI coaches are part of the research team based at the University of Vermont (UVM). It is intended, but not required, that all project teams meet with their coaches on a regular basis.
Online Learning Community
The IBH-PC toolkit includes an online forum and learning community created using an open source bulletin board system (32). It provides a platform for practice members and other IBH-PC stakeholders to communicate with one another. Discussion topics vary and are primarily focused on practice member needs and interests. It is intended, but not required, that all project teams engage in the online learning community through regular participation by at least one team member.
Criteria for discontinuing or modifying allocated interventions {11b}
Study practices randomly assigned to the intervention are encouraged to use the IBH-PC toolkit completely and sequentially, as prescribed. However, practices may use the toolkit out of sequence, modify sections, or skip them completely. Given the low risk of this practice-level intervention, there are no early stopping rules.
Strategies to improve adherence to interventions {11c}
The IBH-PC toolkit is designed to encourage practice adherence to the intervention through engagement with a remote QI coaching team, access to continuing education credits as part of the online education curriculum, and creating a sense of community with standards of normal behavior through the online learning community.
Relevant concomitant care permitted or prohibited during the trial {11d}
Participating practices may not participate in other practice change projects similar to the IBH-PC toolkit during the trial. However, they may use QI and other strategies to improve any aspect of their practice they see fit, including improving BH services.
Provisions for post-trial care {30}
Not applicable. The study represents a 24 month practice level-intervention, there is no need for post-trial care.
Outcomes {12}
Research Question 1: Does using the IBH-PC toolkit, a practice-level intervention, affect patient-centered outcomes in adult patients with multiple chronic medical and behavioral health conditions?
Primary Outcomes
- Change in mean patient Patient-Reported Outcomes Measurement Information System (33) (PROMIS-29) scores from baseline to follow-up for eight domains:
- Physical function
- Anxiety
- Depression
- Fatigue
- Sleep disturbance
- Social functioning
- Pain intensity
- Pain interference
Secondary Outcomes
- Change in other mean patient-reported measures of health and quality of care from baseline to follow-up.
- Empathy (34)
- Medication adherence (35)
- Healthcare utilization (36)
- Time lost due to disability (37)
- Physical function (38)
- Patient centeredness
- Change in mean measures of patient disease control from baseline to follow-up for specific subgroups identified from EMRs.
- Depression (39)
- Anxiety (40)
- Asthma symptoms (41, 42)
- Substance use disorder (43)
- Problem drinking (44)
Research Question 2: Does using the IBH-PC toolkit affect the degree of practice-level behavioral health care integration in primary care practices?
Secondary Outcomes
- Change in median PIP total score from baseline to follow-up
- Change in median PIP domain scores from baseline to follow-up
- Practice Workflow
- Clinical Services
- Integration Methods
- Case Identification
- Patient Engagement
- Workspace Arrangement and Infrastructure
Research Question 3: What factors support or impede successful integration of behavioral health care into primary care practices?
Secondary Outcomes
- Identification of practice characteristics and external factors supporting or impeding integration as measured by changes in PIP scores.
Research Question 4: What are the costs of implementing the IBH-PC toolkit?
Secondary Outcomes
- Total estimated costs of using the IBH-PC toolkit over a period of 24 months, excluding the costs and revenues associated with ongoing clinical operations.
Participant timeline {13}
Practices are screened on a rolling basis for study eligibility and will therefore have varying randomization and study start dates. Moreover, it is anticipated that practices will move through the IBH-PC toolkit at different rates. Therefore, the timeframes represented in Figure 2 are anticipated averages across all study practices.
Sample size {14}
Sample size calculations were adjusted for the cluster randomized design (45). To achieve a nominal significance level of 0.05 for the main hypothesis that any one or more of the eight PROMIS primary outcomes achieves significance, we will apply the Bonferroni correction (46) requiring P<0.05/8=0.00625. All PROMIS domain scales are normalized to a standard deviation of 10 points (33). We assume that the within-practice correlations of the PROMIS scales are 0.03 (similar to the SF-12 (47)). Although minimally important differences (48) are not fully specified, they range from 2 to 8 points for other instruments in the PROMIS series (49, 50).
An analysis of 75 subjects from each of 40 practices gives the study 3,000 subjects and 90% power to detect a difference in any of the 8 PROMIS scales of 2.5 points or more.
Recruitment {15}
Recruitment of primary care practices to participate in the study leverages the professional networks of the study’s leadership and co-investigator teams. Study leadership calls potential practices, screening for interest and eligibility.
To recruit study patient participants, the DARTNet Institute identifies potential eligible patients using electronic medical record data from each practice. The data extract contains information on patient visits, medications, and diagnoses. These data retrospectively span at least 24 months and include dates within 12 months of the study start date. Eligible patients are contacted by mail and/or phone with an invitation to participate in the study.
We cannot sample from the entire practice panel of potentially eligible patients because, in most cases, the practices have many more patients with multiple chronic medical and behavioral health conditions than their BHPs can reasonably support, which would limit our ability to capture the effects of each practice’s case finding and clinical management changes on study outcomes. Nor can we allow each practice to know exactly who the consented patient participants are because the practice may tend to concentrate behavioral health resources on those patients in a way not usual in real world practices, introducing bias. Therefore, we developed a “Community Panel” for each practice consisting of a random subset of the entire practice’s patient population of active adult patients (Group A in Figure 3). Within the Community Panel, some patients will have the target medical and behavioral conditions and be eligible for to participate in the study (Group B). The study subjects (Group C) are recruited from Group B. Because practice change efforts to integrate behavioral health care usually involve increasing the workload of BHPs, the Community Panel size is proportional to the number of BHPs in the practice, approximately 1,000 adults per 1.0 full-time equivalent BHP. The Community Panel allows us to conduct a pragmatic trial within the staffing constraints imposed by real-world conditions.
All practices receive the names of the patients who comprise their Community Panel during the baseline period of the study, but they are not told which ones are eligible for the study or which consented to participate. Control practices may choose to use the Community Panel in the course of their usual care. Active practices are encouraged to focus their redesign efforts (pertaining to screening, case identification, management, follow-up, etc.) on the patients in the Community Panel.
Approximately 8 participating practices will be purposefully selected for inclusion in a qualitatively-driven collective case study of factors that support or impede successful integration of behavioral health and primary care (research question 3). We will identify tiers of “high change” and “low change” practices, where change is defined as the difference between baseline and midpoint median PIP total scores. Practices will be recruited on a rolling basis as midpoint PIP scores become available. Site selection will be informed by emerging findings from both the qualitative data and successive phases of quantitative data that are slated to continue throughout the project period (51).
To estimate the costs of using the IBH-PC toolkit (research question 4), approximately 8 active practices will be recruited using a combination of purposeful and convenience sampling. Selection will be based on practice type, size, location, and capacity to participate in data collection activities. Recruitment and data collection will begin while practices are actively engaged with the IBH-PC toolkit. (These sites need not be those selected for the qualitative analysis of research question 3.)
Assignment of interventions: allocation
Sequence generation {16a}
Eligible practices are assigned to the active or control arm using a stratified, randomized, approach, in blocks of 4. Randomization blocks are developed using R (52). Practices are stratified based on degree of behavioral health integration at baseline (total PIP score < 50th percentile vs. total PIP score >=50th percentile) and geographical area.
Concealment mechanism {16b}
Randomization cards are a dark color, in order to prevent transparency, and individually sealed in opaque manila envelopes. Randomization envelopes are shuffled within block, numbered sequentially within strata and stored in a locked file cabinet.
Implementation {16c}
When a practice is deemed eligible and ready to be randomized, a data analyst determines which stratum they belong to and opens the next randomization envelope within that stratum. She writes the practice’s name and date of randomization on the card, and then initials the card before filing it. The site's randomization status is communicated to the Principal Investigator (PI) who sends an email to the practice and notifies the study Practice Change team.
Assignment of interventions: Blinding
Who will be blinded {17a}
Once a study practice has been randomized, their assigned study arm is not blinded. Practices are blinded as to who the patient participants are within their practice.
Procedure for unblinding if needed {17b}
Not applicable. This study is an unblinded, practice-level intervention.
Data collection and management
Plans for assessment and collection of outcomes {18a}
The main source of outcomes data and descriptors are the patients and practices themselves. On-line surveys are collected and managed using the Research Electronic Data Capture (REDCap) system, a secure, web-based application designed to support data capture for research studies (53). We also offer patients the option of completing the forms by telephone with a trained interviewer using a scripted interview, or via paper using a mail-in option. Outcome measures are collected at three timepoints: baseline, midpoint, and follow-up. The Data and Analysis Team reviews all records for potential errors.
Study Instruments
Research Question 1: Does using the IBH-PC toolkit, a practice-level intervention, affect patient-centered outcomes in adults with multiple chronic medical and behavioral health conditions?
Our primary outcome, the PROMIS-29 (54), is a set of person-centered measures of physical, mental, emotional, and social health included in the National Institute of Health’s (NIH) measurement toolbox. Rather than a disease-focused measurement system, it provides valid measures of Physical Function, Anxiety, Depression, Fatigue, Sleep Disturbance, Social Functioning, and Pain. These aspects of functioning and well-being are relevant across many chronic conditions and broadly represent “health-related quality of life.”
Additional instruments will be used to collect secondary outcomes:
- The Consultation and Relational Empathy (34) (CARE) survey is a 10-item validated self-report measure that assesses patients’ perception of provider empathy. The CARE has been shown to have excellent reliability (Cronbach’s alpha 0.92).
- The Modified Self-reported Medication-taking Scale (35) (MMAS) is a four-item self-report measure assessing adherence to prescription medication. The MMAS has been shown to have adequate reliability (Cronbach alpha = 0.61 (55)).
- The Utilization Patient Report (36) is a 3-item self-report measure asking patients to recall their healthcare utilization in the past year. Specifically, patients are asked to report utilization of visits to the emergency room, overnights in the hospital, and outpatient appointments to a healthcare provider.
- Time loss due to disability will be measured using the Restricted Activity Days survey (56). This survey asks patients to report on the restriction of their daily lives due to illness and disability.
- Physical function will be measured using the Duke Activity Status Index (38) (DASI), a 12-item validated self-report measurement used to assess functional capacity. The DASI has been shown to have good to excellent reliability in several studies (Cronbach’s alpha ranging from 0.80 to 0.93 (57)) and correlates with peak oxygen uptake measured by exercise stress test.
- Patient-centered primary care will be measured using the Patient Centered Index (PCI), a 16-item questionnaire. This tool was developed for use in the IBH-PC study by a team of researchers, patients, and other stakeholders to assess how a patient perceives their care. Reliability and validation measures are not yet available.
- The 7-item Generalized Anxiety Disorder Scale (40) (GAD-7) is a self-report questionnaire developed and validated in a larger primary care patient sample to assess anxiety symptom severity. The GAD-7 has been shown to have excellent reliability in several studies (Cronbach’s alpha ranging from 0.79 to 0.91 (58)).
- The 9-item Patient Health Questionnaire (39) (PHQ-9) is a validated self-report measure of depression symptom severity. The PHQ-9 has shown to have good reliability (Cronbach’s alpha ranging from 0.86 to 0.89 (39)).
- The Asthma Symptom Utility Index (41, 42) (ASUI) is a 10-item validated self-report measure assessing control and quality of life as it relates to asthma symptoms. Reliability for the ASUI is considered good with a Cronbach’s alpha of 0.74 (41).
- Alcohol use will be assessed using 2-items from the Self-Report Habit Index-Alcohol (44) (SRHI-A). The SRHI-A has shown to have excellent reliability (Cronbach’s alpha 0.94 (44)).
- Substance Use Disorder will be assessed using the 5-item Substance Disorders subscale of the Global Assessment of Individual Needs-Short Screener (43) (GAIN-SS), a biopsychosocial screener for individuals presenting with substance use and mental health concerns. The GAIN-SS Substance Use subscale has been shown to have acceptable reliability (Cronbach’s alpha ranging from 0.77 to 0.84 (59)).
Research Question 2: Does using the IBH-PC toolkit affect the degree of practice-level behavioral health care integration in primary care practices?
The Practice Integration Profile (PIP) is a survey of 30 items completed by primary care providers and staff about their own practice (60, 61). It provides a total integration score and 6 domain scores: practice workflow, clinical services, integration methods, case identification, patient engagement, and workspace arrangement and infrastructure.
Research Question 3: What factors support or impede successful integration of behavioral health care into primary care practices?
A cross-case comparative analysis (51, 62, 63) will be conducted in selected practices. Researchers will collect data contained in: field notes, such as observations of the physical layout of the practice, workflow, and practice meetings; transcripts of semi-structured interviews with coaches, cluster leaders, and practice staff; and relevant site documents.
Research Question 4: What are the costs of implementing the IBH-PC toolkit?
A dedicated staff member from each of selected active study sites will periodically complete an in-depth program implementation Cost Assessment Tool (CAT). The CAT is a role-based time-effort evaluation instrument listing the number and type of practice staff participating in each toolkit activity. We will total the time various practice members engage with each component of the IBH-PC toolkit over a 24-month period and calculate the time-effort and cost in U.S. dollars using median total compensation rates appropriate for the practice’s ZIP code and the actors’ professional roles. We will also include any supply, equipment, capital or other non-personnel costs, including training time and coaching time associated with toolkit implementation. Time-effort data collected from participating practices by each participant and subtask will be analyzed for completeness, range and “face validity” of time estimates, and appropriateness of data to determine econometric outcomes. Key variables will then be estimated from global and sub-group averages, including total time and corresponding cost of program implementation steps, time/cost of the planning phase, and those of all implementation phases.
Plans to promote participant retention and complete follow-up {18b}
First, each practice receives financial support for a clinical leader and a practice manager to attend to data collection and other non-clinical (research-related) aspects of the project. Second, practice sites are organized into clusters of about 5-7 practices. Each cluster has a designated local Cluster PI (a behavioral or medical provider or researcher). The Cluster PI serves as a member of the research team and coordinates research activities, engages upper practice management, and provides liaison with the investigators. Third, all control practices have access to the intervention (the IBH-PC toolkit) upon completion of the follow-up assessments.
To encourage patient participants to remain engaged and complete the baseline and follow-up surveys, we include a midpoint survey data collection time point to minimize the time in between research contacts and pay patient participants $30 upon completion of each survey ($90 total).
Data management {19}
The research team developed a written data governance policy to guide appropriate storage and use of study data. The written policy is available upon request from the study PI.
Confidentiality {27}
Data from EMRs used to recruit patient subjects are extracted, transformed, loaded and curated from clinical practice sites by the DARTNet Institute. DARTNet uses a HIPAA-compliant sFTP cloud transfer system. This includes encryption of all data in transit and deletion of the data as soon as the transfer is complete. Data files transferred to DARTNet are immediately moved behind a second firewall where access is internet protocol address (IP)-to-IP controlled. All file manipulations occur behind this second firewall. Once DARTNet has completed subject recruitment, they use a similar system, Egnyte (64), to transfer the list of patients on a practice’s Community Panel to a practice.
Patient names, medical record numbers and other unique identifiers are stripped and replaced with a unique study identifier before transferring EMR files to the UVM research team. Electronic medical record files contain data on healthcare visits, clinical measures including blood pressure and hemoglobin A1C, as well as medication and problem lists. Only research personnel authorized to access data for the proposed project have access to the data.
After transfer of EMR data to UVM, all source material, copies, data files, transcripts and other research materials are maintained in a secure environment with access restricted to members of the study Data and Analysis Team.
Online patient and practice surveys are collected and managed using the REDCap system (53), a secure, web-based application designed to support data capture for research studies.
Physical copies of any patient surveys or other research documents with potentially identifying information are kept under lock and key. Digital materials are secured behind password protected firewalls.
Plans for collection, laboratory evaluation and storage of biological specimens for genetic or molecular analysis in this trial/future use {33}
Not applicable. Biological specimens are not collected as part of this study.
Statistical methods
Statistical methods for primary and secondary outcomes {20a}
Research Question 1: Does using the IBH-PC toolkit, a practice-level intervention, affect patient-centered outcomes in adults with multiple chronic medical and behavioral health conditions?
For the primary outcomes, we will use generalized linear mixed (GLM) models (65) of patient health status (PROMIS-29 scales) to perform intention-to-treat analyses (66) as a function of experimental condition (active vs. control) and patient characteristics, with patients clustered within practices. The parameters of interest are the central tendency (mean), statistical significance (P values) and 95% confidence intervals (CI) of the adjusted change in PROMIS-29 domain from baseline to follow-up. Each of the 8 outcome domains in the PROMIS-29 will be modeled individually as 8 separate hypotheses with adjustment for multiple comparisons and for clustering within practices.
Patient characteristics will be measured at baseline and will include age, gender, race, ethnicity, marital status, employment, education, and income.
For secondary patient outcomes, including change in other patient-reported measures of health and quality of care, and measures of patient disease control from baseline to follow-up, we will use similar GLM models as above but will not adjust for multiple comparisons.
Research question 2: Does using the IBH-PC toolkit affect the degree of practice-level behavioral health care integration in primary care practices?
We will use GLM models of practice integration level (total PIP and PIP domains) to perform intention-to-treat analyses as a function of experimental condition (active vs. control) and practice characteristics. The parameters of interest are the central tendency (median), statistical significance (P values) and 95% CI of the change in total PIP and PIP domains from baseline to follow-up.
Practice characteristics will be measured at baseline and will include ownership type, specialty, training status, non-profit status, size, region, census-tract based variables, primary care medical home (PCMH) status, and % Medicare patients.
Research question 3: What factors support or impede successful integration of behavioral health care into primary care practices?
A cross-case comparative analysis (51, 62, 63) will be conducted for selected case study practices. Practices will be selected from the extremes of change in the main outcome variable, and additional practices maybe selected for confirming/disconfirming analyses (67). Data collected from the individual case study sites will be analyzed to identify site-specific contextual factors that support or impede integration. Through multiple rounds of coding the data, researchers will organize and categorize related evidence on a topic from multiple data sources, assisting with pattern identification in preparation for thematic analysis (63, 68).
After analysis at the individual practice level, a cross-case study analysis will be conducted to identify prominent themes. This process allows for the in-depth examination of the individual practice sites to be followed by a cross-cutting analysis of integration processes. Findings will include themes and insights regarding the factors that supported or impeded successful integration of behavioral health care across the selected practices.
Research question 4: What are the costs of implementing the IBH-PC toolkit?
Descriptive statistical and basic econometrical (i.e. time/effort-cost estimation) methods will be used to summarize findings. Costs of IBH-PC toolkit implementation will be measured only, while research costs (identifying eligible patients, collecting and analyzing research data, etc.) and changes in operational costs and revenues after completion of the intervention will not be included.
Interim analyses {21b}
Not applicable. Interim analyses and stopping guidelines are not planned for this study due to the low risk nature of the practice-level intervention.
Methods for additional analyses (e.g. subgroup analyses) {20b}
Subgroup Analyses
Subgroup analyses will be conducted to understand if the IBH-PC toolkit is more or less effective for various patient and practice groups. Specific patient subgroups to be tested include gender, age in quartiles, race (white vs. other), education (less than high school graduation, high school diploma, college degree), degree of multi-morbidity (2 vs. >2 qualifying conditions), and each of the qualifying conditions.
Specific practice subgroups will be analyzed based on practice size, staffing, geography, and financial structure.
These subgroup analyses will follow the same plan as the primary analysis except that an additional characteristic (representing subgroup membership) will be included as well as its interaction with the experimental condition.
Exploratory analyses
We will replace the binary predictor variable (active vs. control) with continuous variables (the PIP total score and its subscales) to answer the question: What is the relationship between a practice’s degree of behavioral health care integration and patient-centered outcomes?
Methods in analysis to handle protocol non-adherence and any statistical methods to handle missing data {20c}
We will employ the “intention-to-treat” principle; analyses will be conducted according to a practice’s randomization assignment. The final analysis population will be fully described and differences from the enrolled population will be presented. An advantage of GLM models with time of measurement as an explicit covariate is that equal numbers of data points and equally spaced intervals between data points are not required (69), so subjects without complete data still contribute to the analysis. Cases that are lost may remain in the analysis, although they will provide less information and reduce the variance of estimated effects less than cases that are present for all measurement times.
Plans to give access to the full protocol, participant level-data and statistical code {31c}
The study is registered on ClinicalTrials.gov (NCT02868983). A complete, cleaned, de-identified copy of the final dataset used in conducting the final analyses will be made available within one year of study completion. It will include a data dictionary with response options and missing values defined as well as a complete set of survey instruments (excluding copyright protected material not licensed for transfer). The data will be available as an Stata data set or comma-separated file. We will not make data from qualitative results available because of the potential for identifying individuals.
Outside investigators who wish to use the data may request them from the PI. They will be encouraged to collaborate with the research team as appropriate. If the research question they propose is not already under analysis by a member of the research team, the data will be encrypted and sent via SFTP or other secure transfer protocol. If outside investigators wish to reproduce any of our analyses, a full description of our methods (including statistical programming scripts in Stata) as well as consultation by the PI, statistical investigators or others as appropriate, will be made available.
Outside investigators will be required to follow all the protocols for confidentiality, security, notifications to the study sponsor (PCORI), acknowledgments of funding, etc. required of the research team.
Oversight and monitoring
Composition of the coordinating centre and trial steering committee {5d}
The research team includes expertise in behavioral and medical management, research design, multicenter data collection, epidemiology and biostatistics, patient engagement, practice transformation, continuing education, quantitative outcomes assessment, and qualitative analysis of processes and outcomes. The team also includes several patients and other stakeholders from the multiple chronic conditions community who provide expertise from lived experience.
Composition of the data monitoring committee, its role and reporting structure {21a}
The Data and Safety Monitoring Board (DSMB) is an independent group that advises the IBH-PC study investigators. The members of the DSMB serve in an individual capacity and provide their expertise and recommendations. They report to the PI.
The primary responsibilities of the DSMB are to 1) periodically review and evaluate reports of potential adverse events reported by study participants, personnel or others; 2) determine the causality and severity of adverse events, and 3) make recommendations for changes in study protocols, operations, or consent procedures as a result of their findings. The DSMB considers study-specific data as well as relevant background knowledge of the patient population under study.
The chair of the DSMB is a patient. Additional members include a PCP, a Psychiatrist, and a BHP.
The DSMB meets at least annually and whenever the PI determines that a significant number of adverse events have been reported or a single adverse event report is potentially serious and warrants DSMB review.
Adverse event reporting and harms {22}
Although the intervention is low risk, it is possible that a research participant could suffer an unanticipated or adverse event related to the study. If such a concern is raised by study staff, a participant, an investigator, clinician or any other party, the event is reported to the Institutional Review Board (IRB) and reviewed by the DSMB.
The DSMB determines the causality of each adverse event as “unrelated”, “related” or “possibly related” to the IBH-PC toolkit or other aspects of the study. It determines the severity of each adverse event according to the Common Terminology Criteria for Adverse Events (70) (CTCAE). Recommendations for changes in the consent process or study protocol may be made, based on the nature of the adverse event.
Frequency and plans for auditing trial conduct {23}
There are no plans for auditing trial conduct.
Plans for communicating important protocol amendments to relevant parties (e.g. trial participants, ethical committees) {25}
Changes to the study protocol are communicated to all research team members electronically and reviewed in accordance with IRB policies. Through regular meetings with the study sponsor (PCORI), changes are discussed and tracked through a regular milestone reporting process.
Dissemination plans {31a}
Dissemination is a standing agenda item for the Investigators and the Stakeholder Advisory Group (SAG). Research team and SAG members, as well as clinicians and patients at the clinical sites, may be asked to make presentations to health care providers, accountable care organizations (ACO), primary care organizations, health-related agencies, legislative committees, and business forums (employer and payer organizations) in their communities. Patient advisors will be asked to review all materials so that findings are understandable and usable by a broad audience. It is anticipated that study findings will be reported in numerous formats such as peer-reviewed literature, conference proceedings, blog posts and policy briefs.