Background:

The rising prevalence of conditions such as diabetes and hypertension motivate innovations to prevent and manage chronic disease. Health care systems increasingly recognize the influence of social determinants of health (SDOH) such as housing, transportation, and access to healthy food. Lay health workers with community expertise may be a valuable resource for primary care teams to address patients' SDOH.

Concurrently, health care systems increasingly include patients in quality improvement and clinical redesign. Evidence on eliciting useful input from patients in this work is scarce. This project addressed 2 evidence gaps: the lack of guidance on incorporating patients as full partners in clinical design and the need for information on methods for integrating lay health workers into primary care and their effects on care.

Objectives:

Aims were to (1) develop, implement, and evaluate new methods to involve patients intensively in care design; (2) design and pilot a new role for primary care teams to connect patients with community resources; and (3) evaluate the design and efficacy of the role. The setting was Kaiser Permanente, a large, integrated care and coverage system in Washington.

Methods:

Twelve patients and 11 providers and clinic staff participated in workshops to design the new role, which they called a community resource specialist (CRS). Recruitment and workshops were evaluated using observation, participant surveys, and in-depth interviews with participants and health system leaders.

Implementation and impact of the CRS role was evaluated using electronic health records (EHRs) and administrative data, staff interviews, patient focus groups and surveys, and clinic site visits.

Results:

In design workshops, participants developed job specifications and other details about the CRS position. Including patients in the workshops was well received by patients and health system representatives. Having at least half the workshop participants be patients empowered patients to lead and contribute. One challenge to including patients in the clinical design was the use of specialized and technical terms. Based on the model developed at the design workshops, the CRS role was implemented at 3 clinics. Implementation challenges included EHR development issues and integrating a new role into primary care teams and workflow.

CRS evaluations showed high patient satisfaction. Primary care teams reported benefits such as eased workload and perceptions of providing better care. In focus groups, patients who used CRSs reported behavior changes and improved health, although no changes were detected from EHR or patient survey data.

Key learnings from this proof-of-concept implementation of a new clinical role were importance of clearly defining responsibilities; locating the CRS in a place visible to clinical staff; allowing time for training and developing relationships with primary care staff and community resources; personally introducing patients to the new team member; and enlisting a strong clinic champion, preferably a provider, to increase physician uptake.

Conclusions:

We demonstrated that patients can be influential codesigners of clinical processes. Lay primary care team members with expertise in community resources, if well integrated into primary care teams can improve patient perceptions of their health and well-being and produce high levels of patient satisfaction.

Research Aims

The overall goals for our Learning to Integrate Neighborhoods and Clinical Care (LINCC) project were to engage patients in developing a new clinical role to augment primary care teams and help address SDOH by linking patients with community resources. Our approach was based on the elements of the Expanded CCM and promoted clinic-community linkages to help prevent and treat chronic illness. We involved patients in the design process for the new primary care role. The setting was Kaiser Permanente Washington (formerly Group Health), an integrated care and coverage system in Washington state with more than 600 000 members. Our specific aims were the following:

• Aim 1: Develop, implement, and evaluate new methods to involve patients intensively in care design. We created processes and tools to select and train patients and staff to effectively participate in a care design process based on the Lean management approach (which emphasizes maximizing customer value, minimizing waste, and using specific management and continuous improvement tools based on those developed by Toyota).^30-32 During the design process, the role was given the official job title of community resource specialist (CRS). We documented our approach's benefits and challenges using a mixed methods evaluation.
• Aim 2: Pilot a new role for primary care teams to help connect patients with community resources. Using the results from the patient-centered care design process, we piloted this new CRS role at 3 primary care clinics (fully at 2 clinics and for part of the project at a third).
• Aim 3: Evaluate the design and efficacy of the new role. We conducted both formative and outcome evaluations of the new CRS role. Specifically, we set out to assess the extent to which the CRS role was implemented as designed and helped connect patients to community resources. We also explored the impact of CRSs on the experiences of patients and primary care staff and patient-centered outcomes such as care coordination, quality of care, overall patient satisfaction and well-being, and efficient use of patient and health care resources.

Adjustments To The Project Plans

This project was funded early in PCORI's award process. As such, it was not a traditional comparative effectiveness project but was planned instead as a proof-of-concept pilot. We developed our proposal in response to ideas that originated with delivery system leaders. When funding was awarded, however, the delivery system leadership was undergoing a major shift and the leaders who initiated this work had left. These changes affected our project, including our budget. Funding for the CRS role was originally intended to be an in-kind contribution from the organization. However, when leadership changed, and funding was eliminated, we modified our PCORI budget to cover the CRSs' salaries and benefits. The delivery system continued to provide indirect support such as space, leadership and supervisor time, human resources and electronic health record (EHR) support, and other operational assistance.

During the LINCC project, other organizational issues, such as hiring freezes, limited human resources capacity, and a backlog in workload queues for EHR changes, had direct and indirect effects on our work. The time frame of our project overlapped with an integrated group practice reorganization that resulted in leadership and staff turnover in key departments such as clinic management, communications, health information technology (IT), marketing, and human resources. Nonetheless, our scientific team remained stable. One coinvestigator took a job elsewhere but other coinvestigators with similar expertise filled his role. Throughout the project, we proactively reached out to our PCORI program officers to inform them of changes in our health care delivery system to consult and get approval for proposed modifications to the study protocol.

Given the complexity of our project, we took some liberties with the prescribed final report outline to ensure clarity for readers. First, we added a separate section for patient and stakeholder engagement that contains all the engagement strategies we used. Next, we divided our methods and findings into separate sections for each aim. Our intent is to help readers understand the distinction between the design process and its evaluation (aim 1), and the CRS intervention implementation (aim 2) and evaluation of the intervention (aim 3).

Aim 1: Methods

Involving Patients in Care Design

Overview

Our goal was to give patients a majority voice in a multifaceted care design process. Therefore, we developed a unique patient recruitment strategy and developed 2 design workshops: a 4-day initial workshop and a 3-day check/adjust workshop aimed at giving patients a strong voice in developing a new clinical role in primary care.

Leadership involvement in development and planning of design workshops

Before the design workshops, the research team worked with delivery system leaders on a clearly stated vision for the new clinic role. Because of the impact on clinics, delivery system leaders set boundaries and human resources specifications within which the role could be designed. At the same time, we also secured clinic champions at each site, who were physicians who volunteered to participate in the design workshops and promote use of the new role. At Kaiser Permanente Washington, design workshops were based on the Lean approach to continuous improvement. Kaiser Permanente Washington Lean consultants worked closely with the research team to ensure that PDPs would be involved as equals with providers and clinic staff.

Design workshops were structured to include a mix of large-group and small-group activities to engage participants (Appendix B). We had 5 facilitators and added activities to the standard Lean process to promote effective communication, such as defining medical jargon, and to ensure all workshop participants understood the PDPs' role and expertise.

Pilot clinic identification

We chose 2 pilot clinics from the 25 in the Kaiser Permanente Washington system during project proposal development based on patient demographics that were associated with higher rates of chronic disease. After meeting with clinic leadership at each of the 2 proposed pilot clinics, both agreed to participate in the study. At each clinic, the medical clinic manager and a physician leader agreed to provide leadership and support for the project. Later, due to leadership changes at one of the pilot clinics, we selected a third pilot clinic to participate in place of one of the original sites. All primary care team members at pilot sites were included in study activities. Providers at the clinics represented internal medicine, family practice, osteopathy, and pediatrics.

PDP recruitment and orientation

We recruited PDPs from patients with recent experience at the pilot clinics, which we defined as continuous Kaiser Permanente Washington enrollment plus at least 2 face-to-face visits in the year prior. We excluded people with certain mental health conditions (delusion, psychosis, or schizotypal disorders) or with dementia, people who were very sick (defined as Adjusted Clinical Group Resource Utilization Bands score = 5),³⁶ and Kaiser Permanente Washington employees. We set targets so that (1) the 2 pilot clinics would be equally represented and patients would be approximately half of the design workshop participants, leading to a recruiting goal of 5 to 6 PDPs per clinic; (2) at least 1 patient would have each of the chronic conditions of asthma, diabetes, and hypertension; (3) age, sex, and race would be broadly represented; and (4) at least 3 patients would have Medicaid insurance.

We identified a random sample of eligible patients and mailed them patient outreach materials with an invitation to call if they were interested in participating. Based on the results of this first round of recruitment, we identified a second random sample of eligible patients who represented the target groups that were missing or underrepresented in our first recruitment round. Patients indicated interest in participating by phone call. One PDP was referred by clinic leadership but went through a screening process similar to the other PDPs (see Table 1 for PDP demographics). PDPs were compensated up to $2000 based on participation at each stage.

Table 1

Patient Design Partner Characteristics (N = 12).

To promote meaningful patient engagement, we gave significant attention to PDP orientation. An interactive orientation session was held at each pilot clinic 1-2 weeks before the initial design workshop to provide PDPs with information about the clinic and the primary care team. A clinic tour familiarized them with clinic activities, including from the staff perspective. The orientation included an overview of their role as a PDP, time with the other participating patients, and an introduction to some of the providers and staff who would be participating in the design workshop.

Provider and clinic staff participants and workshop facilitators

Pilot clinic leaders chose 10 provider and staff workshop participants. In addition, clinic leadership recommended a delivery system patient engagement leader, who was also was selected to participate. These participants, according to standard Kaiser Permanente Washington practice for care design, attended workshops in lieu of regular duties. Also present at the initial design workshop were 5 event facilitators (2 Lean consultants, 2 other delivery system leaders, and 1 research team leader [CH]).

Design workshops

The initial 4-day design workshop, held in February 2014, included 23 participants—PDPs (n = 12), Primary care providers (n = 2), and other clinic personnel (n = 9)—who worked collaboratively to design the new role. The workshop started by explaining if and how clinic patients were currently connected to community resources, the problem that needed to be addressed, and key principles of Lean design. After this introduction, the group was guided through a series of role-playing activities to simulate how patients were currently linked to community resources and envision how a new role could facilitate these connections. Once the group developed the basic components of the role, the large group split into work groups for specific issues, for example, drafting a job description and generating the CRS title. Other work groups focused on developing EHR tools to assist the new primary care role and building relationships with community organizations. The 4 days concluded with a presentation to delivery system leaders such as the medical director of primary care and the vice president of quality improvement. After 15 months, in April 2015, participants returned for a 3-day check/adjust workshop with group discussion (day 1) and in-clinic testing of the newly designed EHR tools and processes to facilitate CRS work (days 2-3; see Appendix B for a summary of the design workshop). This workshop included 8 of the original 12 PDPs (10 were scheduled but 2 could not attend because of illness).

Methods to Evaluate the Design Process

Participant and leader interviews

We conducted structured interviews with design workshop participants (N = 27, with research team leader CH excluded from interviews) after the first design workshop, with a second round of interviews with the PDPs who participated in the check/adjust workshop 15 months later (n = 8; see Table 2). We interviewed 5 delivery system leaders in fall 2015 (the last phase of the project) to ask about the perceived value of including patients in the workshops and care design, in addition to their overall assessment of the project (see Table 3).

Table 2

Design Workshop Participants Who Were Interviewed After the Workshops.

Table 3

Participants in Formal Interviews.

Observation

Two research team members observed the design workshops using an observation guide to capture information on content, group collaboration, facilitation, and level of PDP engagement and input. To ensure quality and shared understanding of the observed phenomena, observers compared notes.

Survey data

All design workshop participants completed a short, deidentified survey assessing experience, satisfaction, challenges, and perceptions for each day of workshops.

Analysis

We analyzed interview data using a thematic analysis approach.³⁷ We derived themes a priori and inductively. One team member (EH) developed an initial code list based on themes from review of 2 transcripts, including a patient and a staff transcript. Three team members (CH, EH, JM) applied the initial code list to 2 transcripts. We used differences and observations that emerged in the coding and reconciliation process to clarify definitions and revise codes and code definitions. Two rounds of additional coding comparisons finalized the code book and confirmed shared understanding and consistent application of codes. We codified transcripts in Atlas.ti ([computer program] version 7.5.2. Scientific Software Development GmbH). After coding, we organized data by specific codes and reviewed them. Based on review of coded data, 1 team member (EH) drafted a coding memo, including key findings with example quotes, with feedback from the coding team (JM, CH) and input from research team members. We summarized observation data and compared them to the findings from the interview coding memos to triangulate findings and understand similarities and differences in observer data and interview findings. We discussed variations and specific examples with the qualitative team to reach a shared understanding of themes.

Aim 1: Results

We recruited 2 pilot clinics. One (Clinic A) was located in a suburban community and served a diverse, primarily working-class population. The other (Clinic B) was located in an urban community that is one of the most diverse zip codes in the United States. When Clinic A experienced a leadership gap, we decided to move the pilot to another clinic. The new pilot clinic (Clinic C) was within 15 miles of Clinic A but was more rural and less diverse (see Table 4 for more specific information about the size and ethnic representation at these clinics). We recruited 12 PDPs who represented a cross-section of the population likely to be served by the new primary care CRS role (Table 1). We mailed invitations to participate in our design workshops to 200 eligible patients stratified by targets described in Aim 1: Methods. Based on the results of this first round of recruitment, we identified a second random sample of 149 eligible patients from target groups that had no members or were underrepresented after our first recruitment round. Of the 23 total patients who called in response to the recruitment letters and indicated an interest in participating, and the 1 who was referred by clinic leadership, we secured commitment from 12 patients to serve as PDPs.

Table 4

Clinic Descriptions and Key Data Descriptors by Clinic.

The design workshop participants, including PDPs, produced a suite of materials that guided implementation of CRSs at pilot clinics. These materials included (1) a general vision for the job (which included that CRSs would provide personalized resource referrals, motivational coaching to assist with use of the referral, follow-up to ensure the resources were appropriate and accessible, relationship-building with local resources, and education of the primary care team regarding resources); (2) a job title; (3) a job description, formatted for submission to the human resources department; (4) a CRS training schedule and list of training activities; (5) a draft brochure to advertise the position; and (6) HER templates to integrate the role into the EHRs, including a patient action plan form, which is a tool to elicit and document next steps for patients.

Several themes emerged from analysis of the qualitative interview and observation data. Overall, the design workshop successfully engaged both clinic staff and PDPs as a design team. Workshop participants felt that their views were always or almost always respected (98.9% per daily evaluation surveys) and that the participants always or almost always worked well together (98.9%).

Workshop participants and facilitators viewed PDPs as a valuable addition. Most participants stated that involving patients in the design workshop influenced the design of the new CRS role, but had difficulty pointing to concrete examples. As one provider stated, “It's hard for me to say one thing because I think they had their fingerprints on everything.” Nearly all participants described how active patient participation changed the discussion and activities toward a more patient-centered design. Patient impact was more visible at the check/adjust workshop, for example when PDPs reemphasized the importance of warm handoffs (ie, in-person introductions) in the referral process. All participants described effective collaboration between PDPs and provider/staff participants.

Many linked this result to the egalitarian and democratic atmosphere.

It is astounding to see a doctor working along with a patient. The degree of separation between the 2 is tremendous, even in our society, and yet here they were working side by side, giving ideas and exchanging ideas, accepting ideas, and it wasn't a case of them telling us, “Oh, this is a lot better to do it this way than this way.” (patient design partner)

I heard a couple of people comment like, “Oh, I forgot that person was a patient” or “I thought they were an employee.” I thought it was pretty brilliant to create an atmosphere where … you're both drawing on that individualized expertise that only a patient can offer, because they've had that experience, or only a staff person can offer because they've worked in the clinic, but they never became opposing or competitive or divided. It really seemed everybody wanted to learn from each other and was on the same page. (patient design partner)

PDPs appreciated the participatory aspects of the workshop, such as role plays and small groups. Provider/staff participants perceived the specific patients recruited as appropriate for helping design the CRS role based on their experiences. Most of the PDPs (75%) had experience with significant or chronic conditions and shared personal health care experiences at key moments to bring the patient experience to life. More than half of the PDPs (58%) had direct experience working or volunteering in community organizations that would interact with the new clinic CRS role, sometimes with job duties similar to the CRS role being created. Participants described how PDPs brought a different kind of expertise than clinicians. They saw this as critical.

I really liked having the patients present. I think it kept us focused on the patient. Before we would say we were [focused on the patient] and then still be doctor focused or still be schedule-focused or still be staff focused. [At the design workshop] we couldn't let ourselves get away from being patient focused. (Kaiser Permanente Washington staff/leadership)

Having as many PDPs as clinical representatives in the workshops was seen as a positive factor that brought a variety of viewpoints and increased patient comfort in participating.

It felt like patients, because they knew there were so many other patients there, they felt empowered and were very free to share opinions, versus the couple of patients I've seen come, where [at other workshops] they're the only one or they're the only 2 in the room who listened more and threw out an opinion if they were asked directly. (Kaiser Permanente Washington staff/leadership)

Some PDPs described a sense of personal growth or satisfaction from participating, and a more positive view of the delivery system.

Few challenges were reported. Those most often mentioned were PDPs' challenges with the more technical aspects of the workshop, particularly terminology related to medical and health IT, although this did not bar patients from participating in discussions, due in part to provider/staff participant efforts to explain terminology and support patient understanding.

Aim 2: Methods

Based on the above process, each of the pilot clinics hired 1 full-time CRS (see Figure 2). The CRS job description (see Appendix C and below) did not require a college degree or clinical experience but required experience working in community settings or service industries. Clinic managers conducted the search and hiring process. CRSs were hired by the delivery system (rather than the research team) and were supervised by pilot clinic managers. Using specifications developed during the design workshops, the research team collaborated with delivery system partners to finalize EHR tools to support the CRSs, including forms to guide CRSs through patient intake, action planning, and follow-up, and a patient registry for CRSs to track services provided and when follow-up was needed. CRSs received 1 month of training and shadowing before seeing patients. Training included overview of general clinic flow and staff roles; detailed review of the CRS role, responsibilities, and tasks (including how the role fit into the clinic flow); orientation to social work and how best to collaborate with social work colleagues; introduction to health coaching, goal setting, and motivational interviewing (ie, collaborative guidance in making health changes)³⁸; introduction to EHR usage and documentation; orientation of community resources databases and available resources in the community; and orientation of key internal delivery systems resources. Because the role was new to the health care system and clinic managers had limited training time and expertise in the content specific to this role (such as motivational interviewing), the research team hired an individual to work closely with the new CRSs as their coach. The CRS coach had extensive experience training clinicians and clinical staff to use motivational interviewing techniques and conduct health coaching. The CRS coach was available for 2 to 8 hours a week for approximately 9 months and met with CRSs weekly to discuss questions, review cases, and problem-solve challenging aspects of their role.

Figure 2

Timeline of Key Implementation Events.

The CRS role and work processes were based on the job description developed in the design workshops (aim 1; see Appendix C for complete job description), with adjustments as the role stabilized. CRSs were instructed to work closely with the primary care teams at their clinic. The key activities of the CRSs were (1) working directly with patients to identify community resources and, using coaching and motivational interviewing techniques, set goals regarding patients' health and accessing health-related resources; (2) researching and becoming familiar with community resources; (3) increasing the primary care team's knowledge of community resources; and (4) administrative tasks. Patients could be referred to the CRS by anyone in the clinic through a warm handoff (ie, in-person introduction) or the EHR referral process. Patients could also self-refer. If referred electronically (or after the CRS's first contact with the patient), patients were automatically entered into an EHR-based CRS registry. Patient visits with the CRS could occur by phone or in person. CRSs had a series of EHR-based tools, based on specifications developed during the initial design workshop, including templates for intake, action planning, and follow-up, to help them work with patients (see Appendix D). Each CRS had a work station at the clinic and full access to the EHR.

Aim 2: Results

Aim 3: Methods

Site Visits/Clinic Observations and Informal Interviews

We used clinic observations to provide formative feedback for quality improvement and to document training and implementation of the CRS role, focusing on observing successes and barriers to implementing the CRS role. One or more members of the research time visited each pilot clinic every 2 to 4 months for a full day. Site visits and observations were more frequent early in the study, with fewer visits once the position stabilized (Table 5).

Table 5

Site Visit Summary.

The research team developed a day-long site visit/clinic observation protocol (see Appendix H). The protocol mixed general observation of clinic flow and activities, shadowing of the CRS, and informal interviewing of the CRS and the clinic manager. Due to the transition from having a CRS at Clinic A to having 1 at Clinic C, the clinics received different numbers of site visits (Table 5). At each site visit, we conducted 1 informal interview with the CRS and 1 with the clinic manager. This resulted in 3 interviews with the CRS and 3 with the clinic manager at Clinic A, 6 with each at Clinic B, and 2 with each at Clinic C. Interviews were not audio-recorded, but detailed notes were taken. For analyses, we included and coded only the final informal interviews with the CRSs and clinic managers along with formal interviews of clinic providers and staff (see below for analytic methods). This regular contact allowed the research team to understand the emerging nature of the CRS role and its impact throughout implementation. Observers took detailed notes, which were transcribed and summarized, with an emphasis on recording implementation successes and challenges and suggestions for improvement. Summaries and recommendations were shared with the research team and clinic managers and CRSs to support continuous improvement of the implementation at the clinic level.

To understand how the CRSs spent their time, we analyzed work activities tracked on CRS electronic calendars using appointments to identify different tasks. Periodically between June 2014 and December 2015, CRSs exported their calendar data for a specified period into an Excel spreadsheet. A research team member coded appointment blocks and analyzed work activities by week. Graphs visualized changes in workload distribution over time and differences between clinics. We provided summaries to clinics as formative feedback.

Administrative Data

We examined health care utilization and clinical outcomes using administrative data from Kaiser Permanente Washington's EHR and health plan data systems. We analyzed data using an interrupted time series design with observations by patient month. We included anyone who saw the CRS at the 3 pilot clinics who was aged 18 years or older at the time of intake visit. Individuals contributed data until disenrollment from Kaiser Permanente Washington or the end of September 2015, whichever came first. To adjust for temporal trends at Kaiser Permanente Washington, we matched clinics with a CRS to control clinics that were geographically close to intervention sites (and therefore, had access to similar community resources), were of similar size, and had similar patient demographics including race/ethnicity distributions. Every CRS patient (as of September 30, 2015) was then matched with up to 3 individuals from the corresponding control clinic based on these criteria: (1) health care utilization in the index month (the month of their first CRS visit), (2) race/ethnicity, (3) duration of enrollment before CRS visit, (4) age, (5) gender, and (6) Adjusted Clinical Group resource utilization bands.⁴⁰ In several cases, an individual was matched with only 1 to 2 controls because appropriate controls were not available. We performed this matching process to select a group of controls that would have been eligible and likely to be referred to the CRS, if their clinic had one. We accounted for missing data in clinical outcomes using multiple imputation via chained equations with hierarchical models allowing for random intercept of each individual with 100 imputations. We combined all 3 pilot clinics into a single exposure category because we compared a single intervention with usual care. Although CRSs were given the same training, job descriptions, and tools to conduct their work, we expected some variation in this real-world setting.

We used random-effects models to analyze data, accounting for correlation between observations within the same patient. We used a linear spline to model the overall temporal trend in Kaiser Permanente Washington, with a knot placed at the time of the CRS visit and 2 additional knots at 3 and 6 months postvisit. To identify differential temporal trends between intervention and control, we included interaction terms for CRS patients with the linear spline terms. We provided CRS patients and controls separate intercepts, to adjust for different baseline rates. A full specification of the model is in Appendix I.

We estimated differences in average patient utilization of health care and patient clinical measures between cases and matched controls at 3 and 6 months postvisit, adjusting for the intervention/control variable's baseline difference at the CRS visit (Table 6). For these analyses to be valid, the pre-CRS visit trends should be similar between CRS patients and their matched controls. During our statistical review, we found that while the prior trends were comparable for utilization measures they were not comparable for patient clinical measures such as body mass index, blood pressure, hemoglobin A1c, and depression measures. Thus, we do not believe that statistical tests of differences in the post period would be valid. Hence, for completeness, we chose to display only raw rates and not conduct formal statistical tests on these measures.

Table 6

Health Care Utilization (Per 1000 Members Per Month) Over Time Among Patients Seen by a CRS (N = 420 From Administrative Data), and Matched Controls (N = 1045).

Aim 3: Results

CRS Implementation Process and Effectiveness

Patient referrals, use of services, and CRS capacity

Figure 3 shows that over the 18-month period of CRS implementation (described by clinic in Table 4), 622 patients were referred to the CRS, generating more than 2000 CRS encounters documented in the EHR. Of those 622, 67% (n = 418) had at least 1 complete visit with the CRS, 17% (n = 104) did not respond to the CRS's outreach efforts but were sent resource information via mail or secure email, and 16% (n = 100) never responded to CRS outreach or actively declined services. The distribution of first, second, and third or higher visits appears in Figure 4.

Figure 3

Referrals to CRS.

Figure 4

CRS Monthly Office and Phone Visits.

After the first 6 months of implementation (ramp-up period), a noticeable increase in phone and office visits with a CRS occurred after the check/adjust event. The CRS capacity during this 18-month time frame peaks at 55 to 60 patient visits per month, representing 40 unique patients. The drop off in visits in fall 2015 at Clinic B reflects the transition to a new CRS. CRSs reported that low numbers in late fall 2015 resulted from a busy cold and flu season and understaffing, so clinical staff did not have time to refer patients to the CRS (Figure 4).

As discussed in aims 1 and 2 descriptions, our PDPs stressed the need for warm handoffs from primary care team members to the CRS. These handoffs ranged from a quick personal introduction to the CRS with a verbal plan to follow up via phone, to an introduction followed by a full CRS intake with a person at the clinic after their visit. EHR data for patients with at least 1 CRS visit found patients who had a warm handoff required fewer outreach attempts to successfully complete their initial visit (Table 7). CRS calendar data summarized in Figure 5 shows how CRS time was spent from April to July of 2015 (the time frame with the highest volume of phone and office visits). The CRS job description projected an activity distribution of 60% to 70% patient coaching and resource connection, 20% to 30% developing community contacts/outreach, and 10% working with the primary care team. Our data showed that actual and projected distributions were close. CRS patient interaction time was 53% to 64%, community outreach time was 17% to 21%, and team interaction was 8% to 12%. Work areas not accounted for in the job description included trainings, project-/research-related meetings, and administrative time.

Table 7

Contact Attempts With and Without Warm Handoff in Patients With Documented Community Resource Specialist Encounter After May 15, 2015.

Figure 5

Community Resource Specialist Work Distribution, as Percentage of Time, May 5, 2015-August 28, 2015.

Patients who interacted with the CRS: description

We examined demographics of patients who interacted with the CRS to better understand who was served by the CRSs and the extent to which the CRSs reached the patients we hoped to serve. The CRS role was not designed to serve patients with the highest health care utilization or who were in an acute care crisis as these patients often already receive social work or care management support.

Rather, the CRS role was designed for patients with some health concerns, such as chronic disease, who need additional support and resources to improve or maintain health. We found that CRS patients were in Adjusted Clinical Group Resource Utilization Bands³⁶ corresponding to moderate health care utilization (Figure 6), an intended priority group. Most CRS patients had 1 or more chronic diseases (Figure 7): Almost 80% had 1 or more of 5 chronic conditions. CRS patients were diverse in age and ethnicity (Figures 8 and 9) and 74% were female.36

Figure 6

Community Resource Specialist Patient-Adjusted Clinical Groups) (n = 418).

Figure 7

Chronic Conditions of Community Resource Specialist Patients (n = 418).

Figure 8

Ages of Community Resource Specialist Patients (n = 418).

Figure 9

Race/Ethnicity of CRS Patients and Overall Clinic Patients.

Patient Outcomes

Our project used several methods to assess patient satisfaction with CRS services, progress toward goals, patient-reported behavior change (with both qualitative and survey data), cost of services to patients, and health care utilization. Among other methods, we administered patient surveys at baseline and 3-month follow-up. Because of delays related to human subjects' approvals, some patients received a follow-up but not a baseline survey, while others received a baseline but not a follow-up because the field period ended before their 3-month follow-up was due. We mailed baseline surveys to 354 patients at the time of CRS registry entry. Of these, 30 were ineligible, 169 completed the survey (52% response rate, with ineligibles removed from the denominator), 115 refused, and 40 never responded. We mailed follow-up surveys to 380 patients at 3 months after registry entry. Of these, 16 were ineligible, 196 completed the survey (53% response rate, with ineligibles removed from the denominator), 71 refused, and 97 never responded. Reasons for ineligibility included health conditions that prevented participation (baseline n = 18, follow-up n = 4), did not speak or read English (baseline n = 4, follow-up n = 9), not a Kaiser Permanente Washington member (baseline n = 5, follow-up n = 2), and deceased (baseline n = 1, follow-up n = 1). Refusal reasons were passive refusal (indicated willingness but did not return survey and/or repeatedly missed phone appointments; baseline n = 72, follow-up n = 46), too busy/did not have time (baseline n = 17, follow-up n = 4), not interested or no reason given (baseline n = 23, follow-up n = 18), unhappy with health care system (baseline n = 2, follow-up n = 1), and privacy concerns (baseline n = 1, follow-up n = 2). After all exclusions were applied, 67 people responded to both baseline and follow-up surveys and indicated that they had interacted with a CRS. Table 4 summarizes survey response rates by clinic.

Services received

Of the 418 patients who had at least 1 visit with the CRS, services ranged from providing information about community resources to developing action plans. Figure 10 shows the percentage of patients who received particular services. In 25% of cases, CRSs provided resource information and developed an action plan with the patient. A few patients received only an action plan with no resource information. Pilot clinics varied in services provided (Figures 11 and 12). The CRS in Clinic B did more action planning than the CRS in Clinic C. Differences in the number of contacts by service type demonstrated variation in the intensity of services provided (Figures 11 and 12). Action planning involved more contacts per client; approximately 60% of clients with action plans had 3 or more visits. The opposite was true for the resource-only service, with more than 60% of clients having only 1 visit.

Figure 10

Services Received by Patients With a Significant Community Resource Specialist Encounter (Intake or Follow-up; n = 418).

Figure 11

Types of CRS Service in 2 Final Clinics (n = 359).

Figure 12

Types of CRS Services Received by Number of Contacts With CRS (n = 359).

Use of resources

To understand whether patients referred to a resource actually used it, we looked at EHR data for patients with resource referrals (n = 356). Table 8 shows that, for 70% of patients who received a resource referral, a CRS made at least 1 attempt to follow up with the patient, as recorded in EHR documentation. Of the 51% of patients with 1 or more successful follow-up visits with a CRS, 43% discussed the resource, and of those almost half reported that they used the resource. However, whether these data overestimate or underestimate the number of patients using resources is unclear. On the one hand, successful connection with the resource might have meant the patient did not feel a need for CRS follow-up (the case for several focus group participants). On the other hand, patients with a successful follow-up might have been more motivated overall and more likely to use resources than patients without follow-up. Patient survey results showed an increase in use of community resources by CRS patients, with more patients reporting that they had connected with any community programs (p = 0.006) and that those programs had met their needs (p = 0.040) after working with a CRS, compared with their experience before meeting with the CRS.

Table 8

Patients Given Resource Information (n = 356).

Goal and referral categories

Using EHR data, we coded patient goals and resource referrals into topical categories. Table 9 summarizes the frequency of different referral types to external resources. Patients could be referred to more than 1 resource. Table 10 shows goal categories, whether patients had action plans accompanying goals, and indications of progress toward goals. If a patient set a goal, likelihood of an accompanying action plan was higher. Overall, 84% of patients (n = 73) with an action plan and follow-up encounter progressed toward their goal.

Table 9

Types of Resource Referrals That CRS Patients Received (n = 356).

Table 10

Goal Setting Frequency and Progress (n = 93).

Patient satisfaction with CRS services

Three-month follow-up survey data for patients who responded that they had a CRS interaction (n = 106) showed high satisfaction with CRS services. Satisfaction levels were similar between the 2 final pilot clinics, even though patient follow-up varied notably by clinic. Survey results were consistent with focus group data showing high satisfaction with the CRSs (Figure 13).

Figure 13

Patient Satisfaction With CRS Overall and by Clinic.

I would say that I could see so many people being helped by this … and I hope to see it expanded. I hope it goes to other clinics. I think this is one of the best ideas, especially new ideas … that I have ever seen. I'm thrilled. (patient, Clinic C) And I was referred to [the CRS], and I met her and it kind of blew my mind … she was very sweet, considerate. And she was able to provide me with a lot of resources and, you know, pushed me in the right way. (patient, Clinic B)

Behavior change and wellness

A key goal for the CRS role was supporting patients in making changes to increase overall health and wellness. We found no significant differences over time in patient survey responses between baseline and follow-up for use of programs related to weight control, healthy eating, exercise, drug and alcohol use, depression, personal enrichment, social services, and chronic disease (data not shown, N = 67). Focus group data, however, indicated some patients felt that working with the CRS contributed to positive behavior change and/or increased wellness. Focus group participants reported many behavior changes, most related to healthy eating or active living.

I kind of liked it because the goals were little ones, like I promised I wasn't going to park in the handicapped parking … I got to park at the end of the lot and that would give me a little walk, and we kind of went from there … it helped me because I did push myself to at least walk a little more. (patient, Clinic B) It's been like 2 or 3 weeks that I have been going to the gym every day— and the last time I weighed myself it was 248, and today was 238, so I lost 10 pounds. (patient, Clinic C)

Almost all focus group participants reported improved health and well-being. Some improvements were general, such as feeling better or finding ways to be physically and mentally healthier.

When you're working hard to get healthy, it's not just physical, it's mental and spiritual. So what I got out of it was not to let anything toxic in my life, be it food, relationships, or situations. So now I look at my life in its totality. And there are things that I won't let in, like, not only Snickers bars but this person who's depressed and drains your battery. (patient, Clinic B)

Other improvements were specific, such as improving strength or losing a specific amount of weight.

I've been in a yoga class now for 11 weeks. I haven't missed at all. I can see and feel how the yoga is making my knee stronger, taking the pain away from my shoulders. (patient, Clinic C)

Several focus group participants described increased activation and self-efficacy, both as knowing how to access health care and community resources and as increased overall goal-setting abilities.

It's lifted my spirits to know that I have the ability to find what I need and that I can go to [the CRS] for that kind of help. You can't go to the doctor, it's not that. She is a whole other kind of help. (patient, Clinic C) I think what was really important to me was she taught me to organize my action plan for life. … Because, you know, I would talk about it but I wouldn't specify, actually execute what I wanted to do. And she really taught me how to motivate myself … to actually do it. That was really helpful and that still stays with me. (patient, Clinic B)

Clinical outcome measures

We intended to use administrative data to compare patients who interacted with the CRS with matched controls to evaluate clinical outcomes such as hemoglobin A1c, body mass index, blood pressure, and depression severity. However, we were unable to perform these analyses because our matching process focused on obtaining sample populations that could be compared for patient health care utilization. We found that the CRS cases and controls were not sufficiently matched for analysis of clinical outcome measures (see Appendix K).

Health care utilization

While the matching process did not result in a comparison group for clinical measures, it was effective for utilization measures; therefore, we were able to examine utilization outcomes. Table 6 shows changes over time in several utilization measures, comparing patients who had a CRS encounter versus the matched comparison group (rates are per member per month). With the caveat that groups were not matched on clinical measures, most utilization measures showed no significant differences between the groups in utilization patterns at 3- and 6-months post-CRS contact. However, we did observe increases in both face-to-face visits (adjusted difference in rates per member per month, 82.7 at 3 months) and secure message utilization (adjusted difference in rates per member per month, 580 at 3 months and 588 at 6 months; Table 6).

We asked focus group participants about additional expenses or costs from working with the CRS. In this implementation, patients were not charged for using the service. Several participants expressed appreciation that the service was free, mentioning that copays are often a burden or deterrent to using health care. Some mentioned the expense of joining a gym, but quickly noted that they would have paid for a gym regardless of working with the CRS. A few participants noted that working with the CRS saved them money by helping them access transportation vouchers, health care assistance, and home repair assistance.

Summary of Key Findings

For aim 1 (develop, implement, and evaluate new methods to involve patients intensively in care design), we found that the process we designed and implemented resulted in the recruitment of 12 patients, chosen or referred by clinic leaders, for design workshops that required significant time and asked patients to return more than a year later. We found no example in the literature (including gray literature) of any patient engagement events that included this many patients in the same meeting or workshop. Our relatively large number of patients allowed patient voices to outnumber staff voices and empowered patients to be actively engaged in the design of care. This engagement was exemplified by their robust participation in both large- and small-group discussions and reports that they felt they were working “side by side” with people on the primary care teams. The level of reported collaboration also demonstrated that patients needed only modest training and preparation to work effectively with clinical staff. While pinpointing specific changes that were made to the CRS role because of patient involvement was difficult, design workshop participants reported patients changed the overall dynamic and tone of the conversations.

Key findings for aim 2 (design and pilot a new role for primary care teams to connect patients with community resources) were centered on the need to be adaptive and flexible when working in a real-world setting. We adjusted our pilot project to accommodate major shifts in the delivery system that caused higher-than-average turnover among staff and organizational leadership. We also found that significant time and resources were needed to establish and integrate a new nonclinical role into primary care teams.

The results from aim 3 (evaluate the design and efficacy of the new role) showed that CRSs saw 418 patients over 18 months, with an average of approximately 35 patient encounters a month per CRS between April and December of 2015. CRSs worked on a variety of topics with patients but mostly focused on healthy eating and active living resources and goals. CRSs varied somewhat in how they interacted with patients despite receiving the same training; one CRS emphasized action plans and longer-term follow-up, while another emphasized resource referral. Based on our observations, we hypothesize that this difference was the result of both clinic staff expectations/understandings of the CRS and the personality and preferences of the individual CRSs.

Patients from both clinics reported high levels of satisfaction with the CRS role, despite the variation in interaction styles. Most patients, regardless of clinic, received resource recommendations and many set personal goals. Of those who set goals and had CRS follow-up documented in the EHR, most made progress. Almost half who received a referral used the resource. Overall, patients were extremely satisfied with CRS services and appreciated coaching and motivational interviewing support from CRSs.

We were unable to evaluate patient clinical outcomes due to the methodologic challenges we described earlier. However, our matching method (described in Administrative Data under Aim 3: Methods) did result in a set of matched control patients who were comparable to CRS patients for utilization analysis. We found that primary care and secure messaging utilization increased for CRS patients compared with controls; however, all other utilization measures were similar between CRS patients and controls. Whether this increased utilization improved the quality of care these patients received is unclear. In contrast to the utilization data, staff perceived the CRSs as easing their workload and enhancing the quality of care that their team was able to provide.

Challenges and Limitations

This project overcame challenges inherent in implementing changes in a real-world health care setting. Staff turnover, complexity in developing EHR tools and metrics for clinic use, and major health care system changes provided an ever-shifting backdrop for this intervention. We adapted our original proposal by reducing the number of pilot clinics, which reduced the number of patients exposed to the intervention. Lower overall patient exposure made it difficult to compare the intervention to usual care. For both methodological and practical reasons, we revised our patient survey approach to focus on patients who received the intervention, rather than a more population-based survey at intervention and control clinics.

The reduction in the number of pilot clinics and delays in data collection because ramp-up for the role took longer than anticipated reduced our data for analyses. Furthermore, the heterogeneous nature of patients' health issues and resource needs meant that our clinical outcomes were diverse and subgroup populations were very small (ie, only 89 CRS patients had an indication of diabetes in the medical record and an even smaller number had associated prescriptions or laboratory results). These low numbers affected our ability to conduct subanalyses. Also, lack of randomization in CRS referral provided the potential for confounding. The matching of patients from CRS and control clinics based on criteria described in Aim 3: Methods, Administrative Data under Aim 3: Methods did not produce samples that could be used to compare patient clinical outcome measures. Therefore, we did not investigate whether these measures changed as a result of working with the CRSs. Furthermore, our challenges in matching controls to intervention patients for clinical measures meant that we could not determine if health care utilization by patients who interacted with CRSs was affected by increased poor health of these patients.

CRSs saw fewer than 700 patients during the study, raising concerns about efficient use of resources. We caution that the results from our proof-of-concept pilot project should not be extrapolated to CRS-type roles once they are established and implementation processes have been optimized. Furthermore, we were not in a position to fully document all the potential monetary savings the CRSs may have generated for patients or the delivery system, given the difficulties of collecting data on and monetizing patient satisfaction, disease prevention, and behavior change. One indicator of the value of this role was that the clinics with a CRS chose to continue paying for the role after external funding for the position ended. Another value indicator is that the delivery system has decided to spread the CRS role to all 25 clinics in the system.

Our qualitative methods relied heavily on self-report as this was often the only feasible way to gather data on patient and staff responses. Finally, selection bias was a concern for the data, which relied on volunteered information (eg, in focus groups and in surveys, which had low numbers of completed before-and-after surveys), referral to the CRS, and ability to obtain monthly measures for clinical outcomes.

We mitigated potential biases by surveying all individuals referred to a CRS, rather than a sample, and we did not restrict focus group invitations in any way. Analyses that evaluated the care the CRSs provided reviewed all completed interactions. For analyses involving health care utilization, we attempted to match on a large number of factors, and investigated whether the matching process yielded similar preindex behavior between intervention patients and controls. We had robust survey methods with paper and phone follow-up and 3 modalities for completing surveys, including via phone (see Appendix E for details about the 4 quantitative data sources). While patient focus groups reported health benefits from CRS services, we found no pre versus post or control versus intervention group differences in clinical or health status at the population level using administrative data or patient surveys. Potential reasons for this finding include a relatively “light-touch” CRS intervention (half of CRS patients were seen only once), small sample sizes for several measures focusing on subgroups (eg, with diabetes or hypertension), and short follow-up time frame.

Implications for the Field

As recognition increases about the influence of SDOH on chronic disease,⁵ health care systems are experimenting with lay health workers on primary care teams.¹³ The function of these workers may include connecting patients with community resources that might mitigate negative SDOH effects. However, few studies have addressed how best to add lay health workers to primary care teams or their influence on health outcomes or health care utilization.

This project developed and evaluated methods that other health care systems might use for designing a variety of patient-facing roles. Health care systems may also benefit from our findings in how to engage patients in care design, which is another area in which practice is changing ahead of evidence on the benefits and challenges. Our study helps fill this evidence gap.

At the clinics, CRSs worked with patients to resolve a wide variety of social and behavioral needs, for example referral to grief support services and transportation assistance. Our findings illustrate the types of support and services that can be offered to patients through a lay role focused on connecting patients with community resources, especially patients with or at risk for chronic disease.

Our evaluation also examined the capacity and productivity of the CRS role as implemented in our pilot clinics. Our findings on how CRSs used their time demonstrate the possible variations in the role. We saw heterogeneity in service provision by CRSs based on their skills, interests, and expectations of other primary care team members. We saw relatively modest overall numbers of patients served. More research is needed to understand how to make the CRS more efficient and/or the ideal CRS patient caseload, including expectations for number of follow-up visits. Recent developments in the area of social needs screening may also prove helpful, as implementing a screening may allow for more systematic and consistent referral criteria and expectations.

The lack of charge for the service was important to patients, raising questions about patient interest in using CRSs if copayment was required. Since the position was PCORI funded, questions about funding for the position remain. The 2 final pilot clinics allocated 6 months of clinical operating funds to sustain the CRS role in 2016 and delivery system leadership decided to spread the CRS to all clinics in June 2017, so short-term sustainability is positive. Nonetheless, more evidence is needed to demonstrate the overall value of the CRS position, including a clear return-on-investment analysis and assessment of clinical outcomes. We plan to use our lessons learned about measuring utilization, patient satisfaction, and other outcomes to inform future evaluations of this role. The goal will be to develop methods that provide valid and reliable results to help health care leaders better assess key process and outcome measures and determine the overall impact of CRS-type roles in primary care.

For health care systems considering a CRS-type role in primary care, comparing the effectiveness of the CRS position with usual care would be valuable. However, due to several factors, including the need to reduce the number of clinics involved in the pilot, we were limited in our ability make these comparisons. We are interested in continuing to test these types of positions in order to better define appropriate scope. Additional future areas of study include investigating if having primary care-based lay health workers who link patients to community resources compares favorably to usual care or other options for referral to community services (such as 2-1-1 resource lines). For this type of study, we need to consider the most appropriate outcomes to measure. Based on our findings, emphasizing patient experience and/or psychological impacts such as sense of well-being and resilience might be the best way to measure CRS effects. Another option for studying the impact of CRS-type roles is to have CRSs work with populations that are fully aligned with the services offered, for example focusing on people with diabetes who would benefit from diet and exercise action planning and support. A narrow scope would increase the clarity of the CRS scope of work; allow CRSs to focus on deeper, more defined knowledge of clinical and community resources; and assist researchers in evaluating utilization and clinical outcomes. However, a narrower scope would mean CRSs would most likely provide services to fewer patients and would view patients by diagnosis rather than focusing on the whole person.

Appendix A.

Community Advisory Panel Membership (PDF, 126K)

Appendix B.

Co-Design Workshop Activities (PDF, 141K)

Appendix C.

CRS Job Description (PDF, 81K)

Appendix D.

CRS Epic Documentation: Tools/Template Example (Intake, Action Plan) (PDF, 475K)

Appendix E.

Overlap Between Key Data Sources (PDF, 65K)

Appendix F.

Compliance with PCORI Methods Standards (PDF, 150K)

Appendix G.

EPIC Code Book (PDF, 213K)

Appendix H.

Site Visit/Observation Protocol (PDF, 99K)

Appendix I.

Staff Interview Guide (PDF, 102K)

Appendix J.

Patient Focus Group Guide (PDF, 147K)

Appendix K.

Unadjusted Estimated Rates of Healthcare Utilization (PDF, 1.4M)