Behind Closed Doors: A Thematic Analysis of Diabetes Community Health Worker Home Visit Content

Abstract

Background and Objectives.

The present work studies how CHWs perform the role of educator and how this relates to the implementation of other CHW roles, skills and qualities. Prior studies on this topic have relied on interviews or focus groups rather than analysis of CHW interactions.

Methods.

Authors conducted a thematic analysis of 24 transcripts of conversations occurring between CHWs and participants during home visits as part of the Mexican-American Trial of Community Health Workers, a randomized controlled trial that improved clinical outcomes among low-income Mexican-American adults with type 2 diabetes.

Results.

Three themes describing interactions related to diabetes self-management education accounted for about half of encounter content. The other half of encounter content was dedicated to interactions not explicitly related to diabetes described by four sub-themes.

Conclusions.

In a successful CHW intervention, focused educational content was balanced with other interactions. Interactions not explicitly related to diabetes may have provided space for the implementation of core CHW roles, skills and qualities other than educator, particularly those related to relationship building. It is important that interventions provide CHWs with sufficient time and flexibility to develop strong relationships with participants.

Community Health Workers (CHWs) have emerged as critical contributors to the struggle for health equity^1,2. This expansion of CHW services has been supported by a growing body of research that has evaluated CHW interventions and found many of them to improve health outcomes^3-6. CHWs play an especially important role in immigrant communities generally, and there is specific evidence for their impact in Latino communities^7-11. Established guidelines describe core CHW roles, skills and qualities which are thought to be the essential components of CHWs’ success¹². Roles are functions that CHWs serve in communities and the health system such as health educator or cultural mediator. Skills support roles and are the ability to do something well, such as communication, listening or relationship-building. Qualities are characteristics a CHW requires to be successful such as compassion or connection to the community.

Discussion about core CHW roles, skills and qualities inform intervention design and policy decisions¹², but how CHWs implement core roles, skills and qualities has only been partially described. Most research has relied on reports from CHWs and clients collected in either interviews or focus groups^13-20.CHWs’ or participants’ accounts of their interactions may be affected by outside influences including the expectations of program administrators. Descriptions of CHW activities in these studies tend to be broad and the inability of researchers to directly observe interactions makes it difficult to arrive at more nuanced interpretations. One recent study was able to partially overcome these limitations by using “unobtrusive observers” to document interactions during a lay health worker led family-based smoking cessation intervention²¹, but the authors of the present study were unable to find additional studies employing similar methods. In the present study, authors analyzed audio recordings of CHW interactions that took place in participants’ homes as part of the Mexican American Trial of Community Health workers (MATCH)²². The purpose of the study was to investigate how CHWs perform the role of educator and how this relates to the implementation of other CHW roles, skills and qualities.

Methods

Research Design.

The present work was a thematic analysis of transcripts of audio recordings of home visit encounters between CHWs and intervention participants. All data came from CHW home visit interactions recorded during the MATCH study. The analysis of directly observed clinical encounters is a well-established method for studying patient-provider interactions in medical contexts and allows for the analysis of actual clinical interactions, most commonly via video or audio recording^23,24. Direct observation of clinical encounters is particularly well suited to the study of the nuances of communication and relationship building and has been employed to investigate topics such as the use of humor in clinical interactions or how physicians learn about their patients’ social context^25,26. Thus far direct observation has not been widely employed in the study of CHW interactions, although at least one prior investigation employed a similar methodology ²¹. The MATCH study, including the current qualitative study, was approved by the Institutional Review Board of Rush University Medical Center.

Parent Study: The Mexican American Trial of Community Health Workers (MATCH).

MATCH was a randomized control study of community health workers undertaken in Chicago, IL²². MATCH tested the hypothesis that a CHW-delivered diabetes self-management intervention would result in significant improvement in A1C levels among community-dwelling Mexican American adults with type 2 diabetes relative to an attention control group at one- and two-years post-randomization. To ensure intervention fidelity, the MATCH study included the collection of audio recordings of CHW home visits occurring as part of the intervention. These recordings are the focus of the study reported here and were analyzed to describe the roles taken by CHWs in the intervention and how they built relationships with patients.

The MATCH cohort consisted of 144 Mexican American adults with uncontrolled type 2 diabetes recruited from several contiguous Latinx neighborhoods in Chicago, Illinois between January 2006 and September 2008²⁷. Participants in the CHW arm showed reductions of A1C of 0.55% (6 mmol/mol) after one year, and 0.69% (8 mmol/mol) after two years (p = .021 & 005, respectively), relative to participants in the control group. All participants in the MATCH trial received education about seven core diabetes self-management skills recommended by the Association of Diabetes Care and Education Specialists²⁸.

In the intervention arm, the self-management skills were delivered by CHWs in 36 home visits over two years. Visits were conducted in either English or Spanish according to participant preference. The three MATCH CHWs were all women, and all resided in the community and had previous experience as CHWs. None of the three had diabetes themselves. The CHWs were trained to deliver basic diabetes education about healthy eating, increasing physical activity, self-monitoring glucose, responding to abnormal glucose readings, adhering to medication regimens, stress management, and prevention of complications through foot care and eye care²⁹. CHWs were not instructed regarding the amount of time they should dedicate to the defined topics. Instead, they were encouraged to allow participant priorities to guide each visit. After educating on a topic, CHWs coached participants in developing a behavioral action plan in which the participant would practice a new self-management skill. At follow-up encounters, CHWs elicited the participant’s successes and challenges and helped them problem-solve and set new self-management goals.

Data Sample and Collection.

The study protocol called for CHWs to record all scheduled intervention home visits with the participants’ consent. During supervisory meetings with CHWs, recordings were reviewed for quality assurance. After completion of the trial, these recordings were stored at the Rush Data Management Center which provided operational support for all aspects of the MATCH trial. All recordings were of home visits conducted between 2006 and 2008.

For the current analysis, a subset of 24 recordings consisting of 2 recordings of 12 different participants were selected by the Data Management Center for review. Recordings were selected with the following criteria to ensure that all selected participants had an established relationship with their CHW and that participants were roughly representative of the total MATCH intervention sample:

1. Selections were drawn from participants who had completed at least 18 of 36 scheduled intervention visits.

2. For each participant, one recording represented a “middle encounter” (10 visits before the last completed visit) and the other was a “late encounter” (3 visits before last completed visit)

3. Half of the selected recordings were from participants who achieved a reduction in A1C by at least 1.0 % between baseline measurement and the 2 years follow-up measurement at the end of the intervention period. The other half came from participants who did not have a significant decline in A1C during that period (<0.2%). All participants had a baseline A1C between 7.0% and 11.0%.

4. Recordings were selected to ensure sampling from each of the three CHWs (between 3 and 5 participants were selected for each of the study’s 3 CHWs)

Table 1 summarizes the characteristics of the 12 participants whose recordings were included in the analysis. The mean age was 59.8 (SD 11.5). Nine of the participants were female (75.0%) and 11 of the participants indicated their preferred language was Spanish (91.7%). The mean baseline A1C was 9.4 (SD 3.1). The participants had a low overall level of acculturation with a mean acculturation score on the Marin language subscale of 1.4 (SD 0.6). The Marin language subscale ranges from 1-5 with higher scores indicating greater acculturation³⁰.

Table 1.

Qualitative Study Participants, Demographic Characteristics

	Qualitative Sample No. (%) or Mean +/− SD
Age	59.8 +/− 11.5
Female	9 (75)
Preferred Language Spanish	11 (91.7)
Baseline A1C	9.4 (3.1)
Acculturation Score*	1.4 (0.6)

^*Marin language subscale; range =1-5; higher score indicates greater acculturation

As with over 90% of all intervention visits in the MATCH trial, all of the intervention visits in this qualitative study were conducted in Spanish. A professional Spanish-language transcription service transcribed these 24 recordings without translation. Published excerpts from the transcripts were translated into English with assistance from a professional translator.

Qualitative Data Analysis.

Methods for thematic analysis guided the interpretation of study transcripts³¹. The authors employed a coding technique that combined elements of both emergent approaches to coding, in which themes are allowed to emerge from the data during analysis, and a priori approaches, in which themes are defined before beginning data analysis. Although these two approaches are often presented in contrast to each other, combining elements of both may allow for a coding technique that fits specific data and research questions³². The authors chose the framework of thematic analysis, a hybrid approach to code generation, to allow for flexibility during analysis and engagement with established ideas as well as unexpected findings in the data.

Two of the authors were responsible for all coding of transcripts. The first coder (SM) was a medical student and fluent non-native Spanish speaker certified in medical interpretation, with an undergraduate degree in qualitative methods, prior experience with several qualitative projects and 10 years of experience living and working in Spanish-speaking communities in the United States and Latin America. The second coder (TB) was a non-native speaker with three years of experience working as a Spanish interpreter in a clinical context.

Before reviewing any transcripts, the authors generated an a priori list of four themes. Two themes were related the CHW role of educator and the educational model that was foundational to MATCH: prescriptive education and collaborative education²⁹. Two other themes, encouragement and resource referral, were related to other CHW roles and skills identified in the literature. While reviewing the first four transcripts, some observed interactions were not well described by any of the initial themes, and new themes were generated to capture them. These additional themes were questions and reporting (also thought to be related to education), and other topics not explicitly related to diabetes. For all initial themes except other topics, the authors applied a narrow interpretation of diabetes self-management and focused exclusively on interactions where the CHW and participant explicitly discussed diabetes.

All themes were formalized into a codebook and the first four transcripts from the study sample were independently coded by two researchers using QSR International’s NVivo 12 qualitative data analysis software. Coded transcripts were reviewed by both researchers together to compare how theme definitions had been applied and the codebook was refined. To test the reproducibility of the codebook, both researchers independently coded five new transcripts. The coding comparison function of Nvivo was used to check the interrater reliability and it was found that all themes had been coded with an interrater agreement of greater than 90%. One researcher coded the remaining transcripts. Finally, all coded transcripts were reviewed together to identify patterns across encounters and to theorize the role of different sorts of interactions between CHWs and intervention participants. Excerpts of interactions between CHWs and participants were selected to illustrate the nuances of each theme. By the end of analysis, new themes had ceased to emerge from the data and the authors concluded that the sample of 24 transcripts representing 2 visits from 12 different participants was sufficient to obtain data saturation.

To supplement this process of thematic analysis, the authors used the matrix query function of Nvivo to tabulate the percentage of words coded under each theme for each transcript and determine the proportion of the content of each visit dedicated to each theme. These results were then averaged across all participants to describe the proportion of visit content dedicated to each theme across visits.

After completion of this initial round of coding, authors noted that a large portion (51.9%) of interactions were coded under the theme of other topics not explicitly related to diabetes. To better characterize these, one author reviewed all interactions with this code and identified four important sub-themes using an emergent coding process: health topics not explicitly related to diabetes, feelings and emotions, family and shared national identity.

Results

The coding process generated four themes that were present in the analyzed interactions. (Table 2). Three of the four themes describe portions of the visit that were explicitly related to diabetes self-management education: prescriptive education, collaborative education, and questions and reporting. A fourth category, other topics, describes portions of the visit that were not explicitly related to diabetes management.

Table 2 –

Theme and sub-theme descriptions

Themes
Prescriptive education	Statements of a factual nature about diabetes in general or statements about what a participant should do to better manage their diabetes One directional transmission of knowledge
Collaborative education	Discussions about practical ways to implement self-management techniques Dialogue with participant as active participant
Questions and reporting	CHW questions or participant offers information about diabetes self-management and symptoms
Encouragement	Statements affirming the progress an intervention participant has made or affirming that they will improve in the future
Resource referral	Referral to community programs and resources
Not explicitly related to diabetes	Interactions not explicitly related to diabetes See sub-themes below
Sub-themes – interactions not explicitly related to diabetes
Health	Discussion of health topics where diabetes is not explicitly mentioned
Feelings and emotions	Discussions where participants feelings or emotions are explicitly discussed
Personal economics	Discussions relates to a participant’s individual economic situation: personal finances, prices, income, working conditions etc.
Family	Discussions of familial relationships and interactions
Shared national identity	Discussions of Mexico or Mexican culture: visiting Mexico, family in Mexico, Mexican food etc.

Two other themes occurred infrequently or not all. Encouragement, consisting of statements affirming the progress related to diabetes management an intervention participant has made or affirming that they will improve in the future, accounted for less than 1% of interactions. Resource referral, consisting of referral to community programs or resources explicitly related to diabetes, did not occur in any of the transcribed visits.

Prescriptive Education.

The first type of educational interaction observed was prescriptive education which accounted for 18.2% of visit content. This domain identified interactions in which CHWs made statements of a factual nature about diabetes in general or instructions about what a participant should do to improve diabetes self-management. A key characteristic of this category was that the interaction was mostly one-directional: the CHW transmits knowledge to the participant and the participant receives it. Examples include tips to help the participant eat healthy and exercise, reminders about correct targets for A1C or foods that should be avoided and information about possible complications of diabetes. In the quotation in Table 3 a CHW instructs a participant about how to choose a healthier variety of oatmeal. Here the interaction is oriented in one direction, with the CHW instructing the participant about which sort of oatmeal is healthier, and the participant passively listening.

Table 3 –

Representative quotations of themes

Theme	Quotation
Prescriptive Education	CHW: …if you look on the back of the oatmeal when it's flavored it has a lot of sugar…. When it's plain oatmeal that doesn't have anything in it, you can see it doesn't have sugar.
Collaborative Education	CHW: When you had a whole pot of food available, you didn’t rationalize to yourself, like someone else might have: “It’s so much, I have to eat a lot because there is a lot.” No. Participant: No honey…I know that I have [to set limits] and I can’t eat more than I should…If I don’t take care of myself who the hell is going to care of me? CHW: Yeah. That's what I tell you, it's building consciousness over time and then it's not as much work anymore.
Questions and Reporting	CHW: And what do you do, for example, to keep your sugar under control, what do you do? Do you exercise, are you careful about what you eat, do you relax, or not do anything? Participant: I just walk a little bit CHW: Everyday, or? Participant: Before I walked more than now, now I don't really. Well, right now, I'm about to start walking again, because before it was cold, before I used to walk, I'd walk the kids to school, but it was really cold, now I'm going to get to work walking.

Collaborative Education.

The second category of educational interaction was collaborative education which accounted for 8.6% of visit content. Collaborative education included discussions about practical ways to implement diabetes self-management techniques in which participants were involved in coming up with the information, solutions, or setting self-directed goals for improving their health maintenance. In contrast to prescriptive education, collaborative education included a horizontal, two-way dialogue between CHW and participant. In the quotation in Table 3, the participant describes the challenge of limiting portion size, and the CHW responds by reinforcing the participant’s healthy decision-making and building on that by encouraging the participant to engage in mindful eating. The interaction goes back and forth between CHW and the participant. Unlike the excerpt above illustrating prescriptive education interactions, here the participant takes an active role.

Questions and Reporting.

The final category of educational interaction is questions and reporting which accounted for 18.1% of visit content. Questions and reporting included interactions in which the CHW asked about diabetes self-care or symptoms of diabetes, or when the participant offered this information. In the quotation in Table 3, a CHW questions a participant about exercise.

Other topics not explicitly related to diabetes.

Just over half (51.9%) of the interactions involved topics not explicitly related to diabetes at all. Within this larger theme, four recurrent sub-themes stood out, frequently overlapping with each (Table 2). One important subtheme was health topics not explicitly related to diabetes. Some of these topics such as cholesterol or kidney failure are closely related to diabetes, but participants and CHWs never made an explicit connection in the transcribed interactions. Others topics such as a child’s dental issues or a patient’s knee pain were less closely related to the intervention’s focus on diabetes self-management.

Another important sub-theme was participant emotions and feelings not explicitly related to diabetes. Participants shared deeply with CHWs, explicitly discussing emotions such as stress, anger, sadness and happiness. In Table 4, there is a quotation from a participant discussing his insecurities when starting a new romantic relationship, and his happiness when his new partner accepted him for what he was. The family sub-theme also frequently captured participants opening up to CHWs about their private lives. Some discussions focused on the routine, such as birthday parties or what grandchildren ate after school. Family conflict was also a frequent focus of conversation. In the quotation in Table 4, a participant shares the struggles she experienced living with her alcoholic son while the CHW listens and provides advice.

Table 4 –

Representative quotations of subthemes

Sub-theme	Quotation
Health	Participant: Because my knee, when I do this, it seems messed up. CHW: Uh-huh, may be there isn’t any liquid any more Participant: That’s what I want to see, because maybe they gave me an injection and if its going to be anything, its that the liquid or whatever is running out, maybe they’ll do it again or something…because here he did give me an injection.
Feelings and emotions	Participant: This person, I had been in a relationship with her many years ago. She told me she had been waiting all this time and I don’t know what else, then I thought, that how I look, not just my age, but the cancer, my eye, was going to be … a disappointment for her. You see? But no, when she saw me she said “I keep thinking about you”…and that made me feel good.
Shared national identity	Participant: I remember when we were young we weren’t rich, beans from the pot with fresh cilantro from the yard, and you added a little bit of chile, and look, you ate so well CHW: With some fresh tortillas Participant: It’s the most tasty thing there is
Family	CHW: Your son needs help, he needs to work on his problem with alcohol Participant: Yeah, I don’t understand. Bottles and bottles and bottles. It’s a shame. He just drinks and drinks and drinks and complains, complains, complains and yells and yells… terrible, and hitting the bottle, that is wrong.

A final group of subthemes often related to shared points of reference between participants and CHWS: personal economics and shared national identity. Economics included personal economic struggles of the participant such as having hours cut at work or difficulty paying utilities, but also topics such as rising prices or the cost of travel that were relevant to both the participant and CHW. All CHWs and participants were Mexican-American, and most visits included some interactions related to this shared national identity, including topics such as visiting Mexico, relatives in Mexico and Mexican food.

Discussion

These results use qualitative data from recordings to provide one of the most detailed descriptions to date of what happens during CHW visits. Most prior studies describing CHW encounters have relied on CHWs’ and participants’ self-reporting of their interactions through surveys or interviews^13-20. These studies provided valuable insight into the role of CHWs, but reliance on self-reported accounts of CHW interactions has limitations. In some instances, CHWs and participants may be consciously or unconsciously transmitting preconceptions of what they were told should occur during a CHW visit. CHWs and participants may also experience challenges in recalling or characterizing their discussions.

Transcription and coding of recordings of CHW visits allowed for the direct analysis of interactions of CHWs with participants. This approach permitted a more nuanced description of CHW roles. For instance, previous studies have described the role of CHWs using broad descriptors such as educator¹³. In the present study, the authors were able to examine educational interactions more closely and subdivide them into more precise categories. Another major advantage of this approach was that it allowed the authors to calculate the proportion of each encounter devoted to different sorts of interaction. The MATCH model assumed that collaborative education was a central component of successful CHW interactions, and it was heavily emphasized during CHW training. The results of the present study provide evidence that CHWs employed it regularly after their training, although not as much as intended. Collaborative education occurred in all but two of the visits analyzed but on average accounted for only about one fifth of all time spent on education (8.6% of visit time was spent on collaborative education out of 44.9% of visit time was spent on all types of education).

These results show that CHWs made significant use of prescriptive education, directly advising participants about self-management behavior. The MATCH protocol de-emphasized this type of teaching, in favor of motivational interviewing and behavioral self-management. Prescriptive education nonetheless accounted for 18.2% of observed visit content. This is an important reminder not to neglect the continued reliance on traditional educational modalities CHWs frequently default to. The present analysis also helps to highlight the role of a third component of CHW educational interactions: questions and reporting. This modality of education was not explicitly emphasized in CHW training for MATCH and has not been well described as a component of CHW educational interactions elsewhere. Although in a few instances asking questions may have been of a form of behavior correction, more often it seemed to be a tool employed by CHWs to initiate reflection.

Also significant was the limited presence of CHW roles besides educator in interactions explicitly related to diabetes. The CHWs used their skills in cultural mediation and emotional support, but the direct connection to diabetes was not verbally stated. For instance, the participant who discussed the complex emotions he felt when starting a new romantic relationship (Table 4) later explicitly indicated the CHW was a source of social support. Provider of social support is recognized as a key CHW role¹². Even though the focus of the CHW’s empathy was directed to the patient’s experiences beginning a new relationship, this social support may still have facilitated diabetes self-management, for instance by strengthening the trusting relationship between participant and CHW.

Perhaps the most notable finding was that on average, slightly more than half of all interactions analyzed were not explicitly related to diabetes. The sub-themes used to describe these interactions focused on their explicit content. Since the researchers could not ask CHWs or participants to explain their conversations, it was difficult to make strong inferences about their deeper significance. Since all sub-themes were generated by emergent coding, this task was even more challenging than for the diabetes self-management education related themes which were structured by an established theory that facilitated the linkage of observed interactions with theoretical constructs.

However, at the level allowed for by these limitations, interactions not explicitly related to diabetes did seem to provide evidence for the implementation of a wider variety of CHW roles, skills and qualities. The CHW skill that seemed most present in the interactions not explicitly related to diabetes was relationship building, recognized as a core CHW skill¹². Qualities such as caring and compassion were also present. Sub-themes of economics, health and family may have represented CHWs simply engaging with participants on the topics that were most important to them. Discussions of Mexico or Mexican culture may have used shared national and cultural reference points to build a relationship. This also relates to the role of cultural mediator, another core CHW skill¹². This relationship may have allowed the CHW to serve as mediator between the participant and the health care system. Interactions described by the sub-theme emotions would have provided CHWs with opportunities to engage in empathetic listening, also relating to the CHW role of provider of social support and the quality of compassion.

Thus, CHWs devoted just as much time to building and maintaining relationships as they did to delivering the formal content of the intervention. This contrasts strongly with physicians, who spend 82% of their time with patients discussing biomedical, mental health or health behavior topics and only 4% discussing “small-talk” or other topics³³. These differences may point to CHWs strengths in relation to physicians. Other authors have proposed that CHWs are successful in delivering health education because of the personal relationships they build with their participants^12,17,18. The data presented here adds support to that conceptual model.

All interactions analyzed occurred from 2006 – 2008 when the MATCH intervention was implemented. Significant time has passed since then, however, chronic disease self-management home visit interventions similar to MATCH continue to be implemented and researchers remain interested in CHW roles such as those observed here^34-36. The results of the present study are directly relevant to recent CHW programs, but as historical data they may also serve as a reference point for understanding changes in CHW interactions over time.

Study Limitations.

Of more than 1,000 hours of visits that were recorded during the MATCH trial, authors only had the resources to transcribe and code 24 encounters each lasting about an hour on average. A future study with a larger sample size might allow for the selection of more visits for each participant to study how relationships between CHW and participants develop over time. All participants in the original MATCH study were Mexican-American adults from Chicago. The intervention was focused solely on diabetes, and all interactions occurred as part of home visits. The present study’s results may not be transferable to interventions working with other populations, focused on other conditions or taking place in in other settings. CHWs and participants knew they were being recorded and it is possible that this influenced their interactions.

Connections between observed interactions and established CHW roles could only be inferred by researchers as the study design did not allow for CHWs or participants to comment on their interactions. Researchers were further limited by their reliance on written transcripts which could not convey body language or tone of voice. As a consequence, when coding the researchers focused on what CHWs and participants said explicitly, for example limiting the application of diabetes self-management education themes to instances where diabetes was explicitly mentioned. This likely led to many interactions with an implicit connection to diabetes self-management being coded as other topics not explicitly related to diabetes.

Conclusion and Implications.

These findings help characterize how CHWs perform their role as educators and balance this with the implementation of other rolls, skills and qualities, particularly relationship building. The results of this study also imply that intervention protocols should not force CHWs to rigidly adhere to predefined curricular content. CHWs should be given time and flexibility to develop strong relationships with their participants through interaction that need not be explicitly related to the intervention focus area.