Abstract
Background: This proof-of-concept study tested the feasibility and acceptability of INTEGRATE-D, an implementation support intervention for primary care clinics to improve the psychosocial care of patients with type 2 diabetes.
Methods: Cluster randomized controlled pragmatic trial, with a parallel, convergent mixed methods design. Two Intervention Clinics (ICs) were offered tailored training on American Diabetes Association (ADA)-recommended psychosocial care and facilitation to identify and support clinical change. Two Control Clinics (CCs) received no intervention. Primary outcomes: intervention acceptability, appropriateness and feasibility. Secondary outcomes: process-of-care metrics (eg, depression screening, diabetes management) and clinical outcomes measures (PHQ-9 and A1C). Qualitative data were collected to assess implementation and experience with the intervention.
Results: ICs were offered training and received 15-months of facilitation. To accommodate COVID-19-related safety restrictions, the intervention was changed to be delivered virtually (eg, remote facilitation and training sessions). Despite an adapted delivery and COVID-19 and staffing stressors, clinics exposed to INTEGRATE-D found it to be acceptable, well-aligned with clinics’ needs, and feasible. Qualitative data suggest COVID-19 stressors tempered feasibility. The effect of INTEGRATE-D on process and clinical outcome measures were mixed. Several factors, including differences in ICs and CCs not addressed in randomization and delivery of a less intensive intervention due to the pandemic, may help explain these results.
Conclusions: Given the growing number of people with type 2 diabetes and the importance of psychosocial care for these patients, INTEGRATE-D warrants further pilot-testing with a larger sample of clinics and patients, and under conditions where in-person facilitation and expanded training is possible.
- Behavioral Counseling
- Chronic Care Management
- Diabetes
- Disease Management
- Integrated Health Care Systems
- Outcome Measures
- Patient Health Questionnaire
- Primary Health Care
- Psychosocial Care
- Type 2 Diabetes Mellitus
Introduction
Over 30 million people in the United States (US) live with type 2 diabetes,1 the majority of whom receive care for this chronic condition in primary care settings.2,3 Complex environmental, social, behavioral and emotional factors, collectively called psychosocial factors, influence type 2 diabetes management. This is reflected in the American Diabetes Association (ADA) recommendations for psychosocial care for patients with diabetes (Figure 1). Helping patients manage their behavioral health needs and meet diabetes management goals reduces the risk of adverse outcomes.4–6
American Diabetes Association (ADA) recommendations for psychosocial care for patients with diabetes.1
Psychosocial care for patients with type 2 diabetes involves identifying when depression, anxiety, diabetes distress (DD), cognitive impairment, social and economic need are affecting people, in general, and diabetes self-management, in particular, and developing the pathways to help patients access appropriate services and/or treatments.7–10 Busy primary care clinics are often challenged with competing demands and priorities, may be unaware of the ADA recommendations, and struggle to adapt team-based workflows. Therefore, clinics can benefit from time, training and support from an outside expert to develop skills and adopt processes to better integrate patient care.11–14
We report the results of a mixed methods cluster randomized clinic level controlled pragmatic pilot study to test the acceptability, feasibility and appropriateness of a package of implementation strategies – called INTEGRATE-D – designed to assist primary care clinics (henceforth, clinics) with aligning care with ADA recommendations. We assessed whether intervention clinics (ICs) found INTEGRATE-D acceptable, appropriate and feasible.15 We examined if clinics reported using more quality-improving strategies to improve type 2 diabetes care (prepost) compared with control clinics (CCs), met a higher number of process-of-care metrics for type 2 diabetes, depression and diabetes distress screening (prepost) as compared with CCs, and demonstrated a positive trend (improvement) in depression symptoms, measured by the Patient Health Questionnaire 9 (PHQ-9), and diabetes management, measured by Hemoglobin A1c (A1C) scores, (prepost) as compared with CCs. We also sought to understand the context in which INTEGRATE-D was implemented, how it was implemented, and how it was experienced by clinical teams and patients. To our knowledge, INTEGRATE-D is one of the first studies to pilot-test an intervention to help primary care clinics better align care with the ADA recommendations.4
The INTEGRATE-D Intervention
INTEGRATE-D was an implementation support intervention designed to help clinics align care with ADA recommendations.4 The intervention was initially comprised of 15-months of tailored external support that included 3 evidence-informed implementation support strategies: (1) Audit and feedback, which involved assisting clinics in accessing actionable data reports to identify care gaps;16,17 (2) Skill-building, which included expert training on ADA recommended care, DD, pragmatic screening and treatment strategies, and education in the medical aspects of type 2 diabetes targeted to BHCs18–21; and (3) Facilitation–monthly, tailored support to help clinics identify and implement changes to align care with ADA recommendations using Plan Do Study Act (PDSA) cycles and the Bodenheimer Building Blocks.22–24
We modified INTEGRATE-D to accommodate strains due to COVID-19 and to align with what we learned from baseline assessments. Table 1 shows clinics received monthly facilitation remotely rather than in person. IC1 received 15 once monthly facilitation meetings; IC2 received 11 due to the delayed intervention start. Facilitation was tailored and included PDSA cycles that tested workflows to incorporate psychosocial screening; facilitators did not use audit and feedback data. We delivered 2 remote expert trainings to IC1. Trainings were recorded and shared with IC2.
Intervention Components, Description, Frequency, and Timeline
Methods
We used a cluster randomized controlled pragmatic trial, with a parallel, convergent mixed methods design to evaluate INTEGRATE-D. Clinics were enrolled in March 2020. The intervention period was from November 2020 to June 2022. Data collection was completed in December 2022. This study was approved by the Oregon Health & Science University Institutional Review Board (STUDY00020783) and was registered as a Clinical Trial (NCT04461405).
Practice Recruitment
Oregon Rural Practice-based Research Network (ORPRN) practices, a statewide practice-based research network, led recruitment. ORPRN sent an e-mail blast to clinics with information about the study and a brief response about level of interest. Outreach was followed with personalized e-mails and phone calls. Four clinics agreed to participate. These clinics all employed at least one behavioral health clinician (BHC) (or intended to use a BHC), varied on their location (rural and urban), size (number of clinicians), and ownership (health-system owned, independent federally qualified health center). Clinical team members were invited by e-mail to participate in interviews and surveys and received up to 2 reminder e-mails. Clinics contacted identified patients; the study team contacted patients willing to consider participation, explained the study, completed informed consent and scheduled an interview.
Clinic Randomization
Consistent with aims of pilot studies, we sought to test and apply a randomization procedure that would be ideal in a larger sample size trial. Clinics were pair-matched and randomized using a covariate-constrained randomization procedure25,26 incorporating each clinic’s number of clinicians, ownership type, clinic specialty (Family Medicine and/or Internal Medicine), and rurality using 2010 RUCA codes.27 One clinic from each pair was randomized to receive the intervention and the other to the control group, which received no intervention. Following randomization, one clinic dropped out before the intervention start due to COVID-19-related organizational strain. We replaced this clinic with one that had similar characteristics to its matched control.
Sample
Clinic staff participants included at least one clinic leader (eg, Medical Director, Practice Manager) and staff (eg, clinicians, medical assistants, BHCs) who worked with a facilitator. Before the intervention, we purposively selected and invited clinical staff from a range of roles to participate in interviews. Following the intervention, clinic staff who were exposed to the intervention were invited to complete a postintervention interview and survey.
Two samples of patients were randomly selected for chart abstraction: (1) patients with a diagnosis of type 2 diabetes and (2) patients with a diagnosis of type 2 diabetes and depression symptoms (PHQ-9 > 9)28 or elevated diabetes distress (DD ≥2).29 The first sample of patients assessed the impact of screening for depression and DD; the second assessed changes in care delivery among patients who screened positive for depression and DD. Chart audits did not contain patient health information and did not require patient consent.
We identified 5 patients across the 2 intervention clinics (ICs) that had a diagnosis of type 2 diabetes and either depression symptoms or elevated DD to participate in an interview. These patients were most likely to have been exposed INTEGRATE-D-related clinic changes.
Data Sources and Measures
Table 2 shows the data sources and measures used to assess the feasibility and effect of INTEGRATE-D.
Study Measures, Variables and Data Sources
Data Collection
We evaluated quantitative data 12 months before the first facilitation visit (preintervention) to 6-months after the last facilitation visit (postintervention). Preintervention interviews and survey data were collected between November 2020 and July 2021. We delayed the intervention start for IC2 to allow them to prioritize their COVID response, adjusting data collection for IC2 and its matched control clinic 2 (CC2). Monthly check-in calls with ICs started in the month following the first intervention visit (December and June, respectively) and continued for the duration of the intervention. Postintervention survey and interviews were conducted between March and June 2022 for IC1 and between June and July 2022 for IC2.
Survey Data
We conducted 2 surveys. One was completed by the office manager or clinical leader from each clinic (n = 4). This survey collected clinic demographic data and 14 items from the Change Process Capability Questionnaire (CPCQ), which was modified to assess a medical group’s use of quality improvement strategies aligned with improvement in type 2 diabetes care. Demographic data were collected once, preintervention. The CPCQ was completed pre and postintervention (see Appendix 2).30 ICs and CCs completed this survey with a 100% response rate. The second survey assessed the feasibility, acceptability and appropriateness of INTEGRATE-D using 3 psychometrically assessed measures developed by Weiner et al.15 This survey was distributed to clinical team members exposed to the intervention (n = 17) after the intervention ended, with an 82% response rate. Surveys were conducted using REDCap.
EHR Data
EHR data were abstracted through a combination of manual and automated reporting, both of which followed a protocol.31,32 Research staff assisted clinic staff with creating a list of patients with type 2 diabetes seen in the clinic at least once in the 15 months before the intervention start and once during the intervention. Using a random number table, we selected 50 patients from this list. The same procedure was used to generate a list of 30 patients who had depression symptoms (PHQ-9 > 9) or elevated DD, using the 2-question or long screener.33 Of those patients, chart auditors further assessed behavioral health outcome measures. The A1C measure was assessed for these 2 groups. Staff conducting chart audits were trained to determine which individuals to include in numerators and denominators, which clinical data to include and appropriate parameters to record.
Interviews
Experienced qualitative researchers conducted semistructured interviews (see Appendix 3) and monthly checks with clinics. Preintervention clinic member interviews (n = 19) explored experiences with delivering psychosocial care to patients with type 2 diabetes. Postintervention interviews (n = 5) and one e-mail interaction complemented postintervention surveys. Interviews were conducted with those exposed to the intervention and explored their experience with the intervention and the changes they implemented. Monthly phone check-ins were conducted with one person from each IC to monitor progress; notes were prepared from these conversations. Patients from the 2 ICs (n = 5) were interviewed to explore their experiences of integrated psychosocial and diabetes care and how it may have changed.
Practice Facilitator Data
The facilitator completed notes following each session documenting what they worked on with the clinic, resources shared, progress, planned quality improvement cycles, successes and challenges. Monthly study team check-ins with the facilitator complemented these notes and allowed for monitoring progress and fidelity. Notes were developed to document these conversations.
Clinic member and patient interviews generally lasted 30 to 45 minutes, were conducted via telephone or web platform, audio-recorded with permission, professionally transcribed, and reviewed for accuracy. Qualitative data were deidentified and organized into Atlas.ti7 for management and analysis.
Analysis
Feasibility, acceptability and appropriateness of INTEGRATE-D were assessed postintervention among ICs. For the other quantitative variables, we compared differences in values preand postintervention among ICs and CCs. For the CPCQ as well as for feasibility, acceptability and appropriateness, we performed clinic-level analyses. For process-of-care measures, we performed patient-level analyses, with results summarized by clinic. When testing the effect of INTEGRATE-D on these outcomes, the independent variables were exposure to the intervention (whether the patient was associated with a clinic randomized to the intervention) and time period where baseline represented the closest measure available for the patient before the start of the intervention and follow-up represented the closest measure after the end of intervention.
To assess the effect of the intervention on A1C and PHQ-9 outcomes, we compared ICs and CCs performing a patient-level analysis using linear mixed effects models. We modeled the outcome of interest as a function of the indicator for period (post vs pre), indicator for intervention arm (IC vs CC), and the interaction between period and intervention, using random effects to account for repeated measures within the same patient over time. We adjusted for potential confounding using a range of patient-level characteristics (see Table 2). Statistical tests were 2-side (α = 0.05) and performed in R (version 4.2.0).
Qualitative researchers with expertise in primary care, integrated care and implementation science conducted analyses. We used a group process to analyze qualitative data, tagging text to assign codes that were aligned with emerging patterns/findings. When codes were clearly defined (ie, we had a codebook) and used consistently among the team, we transitioned to individual data analysis. The team continued to meet to review work and discuss emerging findings. Tagged data were analyzed a second time to examine the similarities and differences across the 2 ICs. Qualitative findings were summarized and connected with quantitative findings to explain the study results.
Results
Table 3 shows the clinic characteristics. Clinics employed at least one BHC. IC and CC arms were well-balanced on ownership, rurality and number of clinicians.
Clinic Characteristics
Acceptability, Appropriateness and Feasibility of INTEGRATE-D
INTEGRATE-D, as modified, was deemed to be acceptable (Mean = 3.60, S.D. = 0.50; Range=[3, 4.25]), appropriate (Mean = 3.79, S.D. = 0.45, Range=[3, 4.25]), and feasible (Mean = 3.50, S.D. = 0.43; Range=[3, 4]), with participants at IC1 having slightly more favorable responses (most participants chose “agree” or “strongly agree”) as compared with IC2, whose responses tended toward “agree” (corresponding to a score of 4) and “neutral” (corresponding to a score of 3), with potential responses ranging from strongly disagree (corresponding to a score of 1) to strongly agree (corresponding to a score of 5). Data not shown.
Qualitative findings were aligned with survey results. ICs described the work with the facilitator as acceptable and valuable:
It seemed really useful to get a discussion going between members of the Quality team, Medical team, and Behavioral Health team. [The facilitator] would ask questions that would get us thinking about how we did things, what types of things we weren’t doing that needed to be done, etc., and having all three departments represented was crucial to planning and problem solving. – Quality Analyst Email, IC 1
Clinic members reported that INTEGRATE-D improved patient care: “I felt like [INTEGRATE-D] improved my care of my diabetic patients,” (Physician Assistant, Interview, IC1) because the intervention provided a more holistic approach to caring for patients with type 2 diabetes. Clinical staff that used the DD tool generally found that it fostered meaningful conversations with patients about managing their diabetes. The intervention also met clinical teams’ needs: “We now have a workflow to screen patients, and we also have a workflow to support those patients who have scored higher on the screenings. Yeah, I think that is a win because we did not really have anything like that at all before.” (Pharmacist, Interview, IC2.) The facilitator kept clinics focused on making changes that were aligned with their needs and navigated external challenges.
Participants described the intervention as mostly feasible. Limited workforce capacity and turnover – challenges exacerbated by the pandemic – affected experience. As the Care Coordinator in IC2 reported, due to the pandemic, the clinic had improvements that “…were shelved, that could help our patients; we just did not have the resources to do more.” ICs kept the changes small and doable.
With our behavioral health department in a bit of disarray throughout the course of the project, we weren’t in a great position to deep dive into new material… I think more implementation probably took place than I realize or that I can see from my perspective, but due to staffing constraints, etc., I think [what we did] was also more limited than what we’d initially hoped for. – Quality Analyst, Email, IC1
Facilitators focused on implementing the DD screening because this was worthwhile and achievable work that primary care clinicians could do without BHC input, since clinics experienced turnover in this role (Facilitator Notes).
Changes in QI Strategies
CPCQ scores increased for ICs and CCs, as shown in Table 4. The CPCQ questions that showed most positive change (average of at least 1 point increase, on a -2 to 2 scale) align with the training and facilitation ICs received, and the types of changes they reported making in interviews.
Change in CPCQ Measure
Changes in Clinical Process of Care Measures
Other than the outcome of screening for DD, which we observed increased among ICs in the postperiod among patients who screened positive for depression symptoms, we found no differences between ICs and CCs on process of care measures. The results for both groups were mixed (Table 5). For patients with type 2 diabetes, IC2 improved its screening for depression and DD and IC1 had a drop. This patterned persisted with the other process of care measures, with IC2 either maintaining or improving its process measures, and IC1 seeing a small drop. A similar pattern emerged in CCs.
Change in Process of Care Measures
Among patients who screened positive for depression symptoms or elevated DD, IC1 improved its rate of rescreening patients, and IC2 did not have adequate patients in the chart audit sample for assessment. Interviewed patients did recall experiencing a range of screenings by the clinical team, but they could not describe these screenings. Patients did not report having had discussions with their clinicians about DD.
CCs show both improvement (CC2) and worsening (CC1) of rescreening rates, but both CCs identified and rescreened patients with previously reported depression symptoms more than was observed at ICs. Among this group of patients, up-to-date screening rates for cholesterol and nephropathy was slightly higher in ICs.
Change in Clinical Outcomes (PHQ-9 Scores and A1C Levels)
Among patients with type 2 diabetes and those who screened positive for depression symptoms, we did not observe meaningful differences between ICs and CCs in change in A1C or depression symptoms (PHQ-9) after adjusting for patient-level demographic characteristics (Table 6).
Adjusted* Change in PHQ-9 and A1C, Means and 95% Confidence Intervals (CIs)
Discussion
INTEGRATE-D was an implementation support intervention designed to help clinics improve psychosocial care for patients with type 2 diabetes. INTEGRATE-D was an acceptable, feasible and fitting intervention. ICs reported implementing more type 2 diabetes quality improvement strategies after the intervention than before as compared with CCs, as measured by the CPCQ; ICs made modest to large improvements, on average in the CPCQ (prepost) and CCs made small to medium changes in this measure.34 Regarding improvements in clinic process of care and outcome measures (diabetes, depression and DD management), the results of this pilot were mixed.
We moved forward with INTEGRATE-D at the start of the pandemic because clinics wanted to participate. Despite this perseverance, the pandemic affected our study in numerous ways. First, we never visited clinics to understand their culture and how they delivered care, as planned. Remote facilitation did not allow for observation of clinic operations, including the informal ways teams used data and the ways that primary care and BHCs worked together. We speculate that this hampered motivation and engagement and disadvantaged the facilitator when it came to leveraging audit and feedback data to help clinics identify worthwhile operational improvements. Second, workforce challenges affected clinics’ improvement efforts. Particularly challenging was the loss of BHCs in the ICs. The facilitator worked with clinical teams to focus on improvements that did not involve BHCs, which was not ideal. Third, the pandemic likely reduced the number of PDSA cycles that ICs conducted and minimized spread clinic wide. Fewer clinic members were working in the clinic at the same time, and their focus was on the pandemic and meeting patients’ basic care needs. As a result of these factors, exposure to INTEGRATE-D was less intense than intended, and may explain the muted and mixed results that we observed.
Despite these challenges, we gained several valuable lessons from this work. BHCs in primary care clinics may know little about type 2 diabetes or their role in caring for this population. This was the case in both ICs. We identified experts to provide training to address this knowledge/experience gap. Despite being unable to interview BHCs due to turnover, we believe that more training was needed. Since the time of our study, the ADA has developed extensive training materials,35 which should be considered for future studies and quality improvement efforts.
This study also has several limitations. First, while it was not ideal to have one IC discontinue participation after randomization, we were able to replace this IC in a manner that aligned with our matching procedure. We modified our timing to ensure that matched IC and CCs received the intervention and were assessed at the same time. Second, differences in when ICs received the intervention (early or later pandemic) may have shaped their experience. Third, while ICs and CCs matched well on structural characteristics, baseline differences in A1C levels suggest differences in patient populations and unmeasured clinic characteristics. One clinic characteristic we did not assess before randomization was experience with integrating medical and psychosocial care. For example, CC1 employed 4 BHCs, and there was nearly a 2:1 ratio of primary care clinicians to BHCs. This ratio allows BHCs to work closely with primary care (eg, do warm hand-offs), and is suggestive of a more advanced clinical model36 then was present its matched IC. Future studies need a more granular understanding of clinics’ experience with integration prerandomization. Fourth, INTEGRATE-D was a clinic-level intervention, and we cannot discern if patients were not exposed to the intervention of were unaware of clinic level changes. Fifth, we conducted fewer interviews with clinical team members postintervention than anticipated; this was a result of the smaller reach of the intervention, burnout and turnover.
Conclusion
Our rigorous mixed-methods evaluation shed light on unique challenges experienced by clinics in addressing diabetes medical care, depression, and DD. Since this study was conducted, the ADA has updated its guidelines to be more specific about the need for annual screening.37 Given the growing number of people with type 2 diabetes, the importance of psychosocial care for these patients, and the continued gap between the guidelines and actual diabetes care delivery, programs like INTEGRATE-D continue to be needed and must be further explored, under nonpandemic conditions.
Acknowledgments
We are deeply grateful to the practices and their staff that participated in this study; Angela Combe, Julia Heinlein, Stephanie Hyde, and Sara Wild at ORPRN; Pam Bonsu, MA and Shuling Liu, PhD.
Appendices
Appendix 1: Acceptability, Appropriateness, and Feasibility of INTEGRATE-D
Appendix
Appendix 2: Change Process Capability Questionnaire (CPCQ) Items
Appendix
Appendix 3: Practice Member Interview Guide Pre-Intervention
Appendix
Appendix 4: Practice Member Interview Guide Post-Intervention
Notes
This article was externally peer reviewed.
Funding: This study was funded by a grant from the National Institutes of Diabetes and Digestive Kidney Diseases (NIDDK 5R34DK124146) of the National Institute of Health (NIH). REDCap is funded by a grant from the NIH (UL1TR0002369).
Conflict of interest: None.
To see this article online, please go to: http://jabfm.org/content/38/2/253.full.
- Received for publication July 15, 2024.
- Revision received October 22, 2024.
- Accepted for publication November 4, 2024.
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.