Development of a Typology of Diabetes-Specific Family Functioning Among Adults With Type 2

Abstract

Background

Family members’ responses to adults’ diabetes and efforts to manage it vary widely. Multiple aspects of diabetes-specific family functioning have been identified as important for self-management and psychosocial well-being in theoretical (i.e., theories of social support and collaborative coping) and observational literature.

Purpose

Develop a typological framework of diabetes-specific family functioning and examine cross-sectional associations between type and diabetes outcomes.

Methods

We used electronic health record (EHR) data to identify a cohort of 5,545 adults receiving outpatient care for type 2 diabetes and invited them to complete a survey assessing 10 dimensions of diabetes-specific family functioning. We used k-means cluster analysis to identify types. After type assignment, we used EHR data for the full cohort to generate sampling weights to correct for imbalance between participants and non-participants. We used weighted data to examine unadjusted associations between participant characteristics and type, and in regression models to examine associations between type and diabetes outcomes. Regression models were adjusted for sociodemographics, diabetes duration, and insulin status.

Results

We identified and named four types: Collaborative and Helpful (33.8%), Satisfied with Low Involvement (22.2%), Want More Involvement (29.6%), and Critically Involved (14.5%; reflecting the highest levels of criticism and harmful involvement). Across these types, hemoglobin A1c, diabetes distress, depressive symptoms, diabetes medication adherence, and diabetes self-efficacy worsened. After covariate adjustment, type remained independently associated with each diabetes outcome (all p’s < .05).

Conclusions

The typology extends theories of family support in diabetes and applications of the typology may lead to breakthroughs in intervention design, tailoring, and evaluation.

We identified four types of diabetes-specific family functioning among adults with type 2 diabetes. Types were associated with clinical, behavioral, and psychological outcomes.

Introduction

Adults’ management of type 2 diabetes mellitus (T2DM) requires daily behaviors that often occur with people they consider to be family [1–3]. As a result, the type and quantity of support for self-care provided by family members [4, 5], family communication styles when discussing diabetes management [6, 7], and the family’s approach to coping with issues or problems related to diabetes [8–10] are incredibly pertinent to the daily lived experience of managing diabetes. These different aspects of diabetes-specific family functioning have been linked empirically to patients’ motivation and ability to initiate and sustain self-care behaviors directly via actions that facilitate or impede self-care [3–6, 10–12] and indirectly via effects on psychological well-being and diabetes distress [13–17].

Broadly, our theoretical understanding of how a family might affect outcomes for adults with T2DM comes from the literature on social support [18, 19] and the literature on coping [8, 20]. These theories identify family constructs—including positive/helpful and negative/harmful aspects of family behaviors and communication styles, and coping styles—and patient constructs—including how the person with T2DM appraises their diabetes (e.g., as the patients’ issue to deal with alone versus a shared issue) and their family’s functioning with respect to diabetes (e.g., satisfactory, effective in addressing challenges). Most studies examine only one or two of these theoretically important aspects when studying the role of family in adults’ T2DM, and the selected aspects vary across studies (c.f. Refs. [7, 11–13, 15, 21–23]). As a result, it has been difficult to integrate findings across studies for a comprehensive understanding of different profiles of family functioning specific to diabetes. Therefore, we sought to develop a typology across multiple dimensions of diabetes-specific family functioning which have been identified as important in the extant literature.

Theoretical Background

The formative literature on social support makes a distinction between the structural (e.g., marital status, number of close relationships) and functional (e.g., needs served, support provided) aspects of support [19, 24, 25]. Structural and functional supports are related, but they have independent effects on physical and mental health outcomes and operate via different pathways [19]. Functional support is also multidimensional, consisting of both perceived and received support. Perceived support includes recipients’ perceptions about the availability of support and its adequacy to meet their needs (e.g., their global satisfaction) [18, 26], and measures seek to characterize respondents’ confidence that support would be given if needed and/or the family environment (e.g., communication styles, cohesion) [18, 19]. In contrast, received support measures ask respondents to recall specific actions or behaviors rather than general impressions [26]. Received family support for diabetes includes actions the family undertakes related to diabetes, whether they are wanted and have the intended supportive effect or are unwelcomed and undermine the recipient. Received support can increase stress, particularly if/when it is unwelcomed [26]. Perceived and received support are weakly to moderately correlated (r = .35 in a meta-analysis [26]). As an example, a person may have a family that is frequently trying to help with T2DM management (high-received support) but perceive her family as overly involved and critical and therefore have very low perceived support. Received family support appears more strongly linked with adults’ chronic disease self-management than structural (e.g., marital status, living alone) or perceived aspects [11]. Because definitions of both perceived and received support can include negative or harmful aspects, the literature on social “support” has long recognized that relationships can be mostly helpful, mostly harmful, or mixed [19].

The recipients’ appraisal of support matters when understanding the effects of social support [25, 27]. The matching hypothesis of social support [27] suggests alignment must be considered between the support given and the nature of the stressor and/or the individual’s needs related to that stressor. For instance, received support might be frustrating, disappointing, or ineffective if the recipient perceives it as misaligned with the demands of the stressor. The stress-buffering/exacerbating hypotheses [25] also emphasize the recipients’ perception of the adequacy of available support resources for the needs elicited by the stressor. These hypotheses all require the support recipient to appraise the adequacy of the support provided/available for the specific demands of the stressor, in this case T2DM.

Appraisal of the stressor itself is central in theories of dyadic or communal coping. Communal coping is defined as the combination of the appraisal of the stressor (T2DM) as a shared issue and the use of frequent collaboration to cope with and manage it [8, 9, 28]. This theory hypothesizes collaborative coping is maximally effective and related to less psychological distress and better health outcomes when T2DM is viewed as a shared issue. Collaborative coping may be less effective or problematic if the illness is viewed as a personal/individual issue. One of the central distinctions between “family” and other types of relationships is that family members work together to make daily decisions and solve everyday problems [29]. The frequency with which families do this with respect to T2DM varies, and the extent to which they enjoy working together with respect to T2DM varies as well. When considering older adults, collaboration to solve problems may be required to compensate for cognitive decline [30]. In other words, some adults with T2DM may need family to collaborate whereas others may not need collaboration but enjoy and benefit from it [31]. Therefore, collaborative coping is multidimensional, consisting of its frequency, the degree to which it is compensatory, and the degree to which it is enjoyable [31].

Evidence Among Adults With T2DM

Observational literature has identified multiple interrelated dimensions of diabetes-specific family functioning as important among adults with T2DM. Instrumental support and assistance (i.e., helpful received support) for diabetes self-care behaviors make self-care easier or more rewarding [3, 11]. Examples include providing/preparing healthful foods, exercising together, helping problem solve around blood glucose readings, or praising consistent self-care efforts [3, 32]. These family behaviors have also been described as markers of observable collaboration in the literature on communal coping [9]. Harmful received support is common, and includes experiencing criticizing and nagging intended to make the person with T2DM perform consistent self-care—frequently overlapping with “social control”—as well as behaviors that undermine or sabotage self-care efforts—overlapping with “family barriers” [3, 22, 32–35]. Observational research has shown many families provide both types of support (i.e., as helpful increases, often harmful does as well) [5, 12, 24, 32], consistent with the description of mixed relationships in the social support literature. Helpful and harmful received support are each independently associated cross-sectionally and over time with self-care [4, 5, 11, 21], diabetes self-efficacy [21], and glycemic control [4, 21].

Perceiving one’s family as providing more autonomy support (i.e., support for one’s personal agency) for T2DM self-care is also associated with long-term glycemic control [7, 36] and may buffer the detrimental effects of diabetes distress on long-term glycemic control [7]. In contrast, perceiving family members to be more critical or judgemental of one’s efforts to manage diabetes is independently associated with worsening diabetes distress over time [14]. These communication styles can be viewed as measures of the type of support one expects to get from their family if/when needed or discussing T2DM (i.e., perceived support).

A couple of studies in T2DM found evidence supporting the importance of the recipients’ appraisal of their family support. For example, Song et al. [37] found the gap between received support and the support needed/desired by the person with T2DM was independently associated with self-care, more than any support construct alone. Tang et al. [17] reported satisfaction with social support was associated with improved diabetes-specific quality of life, whereas received support (helpful or harmful) was not. Based on these few promising findings, more research on adults’ appraisal of family support or family functioning in T2DM is warranted.

Appraisal of T2DM has been shown to moderate the relationship between received support from a partner and diabetes distress, such that support was associated with decreases in distress only when T2DM was viewed as a shared issue [35]. In addition, evidence of collaboration in the way couples talk about diabetes (i.e., “we” talk) was associated with better self-care behaviors [20]. However, we are not aware of any literature on communal coping or shared illness appraisal in T2DM that does not involve only romantic partners.

Objective

We sought to develop a typology that could be applied to understand diabetes-specific family functioning regardless of family structure. Drawing upon the literature summarized above, we assessed constructs of functional family support and communal coping specific to T2DM. We examined who was included in patients’ consideration of “family,” assessed whether our typology was applicable across family structures and whether type was cross-sectionally associated with diabetes self-management and psychosocial well-being.

Methods

We identified a cohort of adults (18–80 years old) receiving outpatient care for T2DM at Vanderbilt University Medical Center and invited them to participate in a survey. We used k-means cluster analysis to develop a typology of diabetes-specific family functioning. Criteria for evaluating typologies resulting from cluster analysis include: (a) clusters replicate using different methods or in different samples, (b) there are meaningful conceptual differences between clusters, and (c) cluster membership predicts differences in meaningful outcomes [38]. Therefore, we compared types identified using different analytic methods and examined the dimensions of interest among the identified types for meaningful conceptual differences. We then examined unadjusted associations between patient characteristics (including family structure) and types, and adjusted cross-sectional associations between types and diabetes outcomes including measures of diabetes self-management (glycemic control, diabetes medication adherence and self-efficacy) and psychosocial well-being (diabetes distress and depressive symptoms). We used EHR data to correct for response bias in analyses occurring after type identification. The Vanderbilt University Institutional Review Board approved this study.

Sample and Procedures

Cohort identification

We used an identifiable clinical data warehouse, the Research Derivative which houses restructured data from the EHR and other sources [39], to identify and recruit patients between October 2017 and May 2018. We refined an existing validated phenotype to identify adults with T2DM [40] which uses combinations of diagnostic codes, prescribed medications, and lab values to identify patients with T2DM [41]. We operationalized “receiving outpatient care for T2DM” as having ≥2 HbA1c results within the past 18 months and ≥1 outpatient visit within the past 12 months. We used diagnostic and billing codes to flag as ineligible patients with conditions or states which were selected because they meet any one of these criteria: likely alter family dynamics around the patients’ health, may present a threat to health and longevity that might reduce the patient’s and clinician’s focus on diabetes control, and/or may impede the patient’s ability to respond to self-report measures (Fig. 1).

Fig. 1.

Participant flowchart.

Survey recruitment

Identified patients received an email invitation to participate with a REDCap [42] survey link specific to their medical record number. Interested participants could click the link to learn more about the study, confirm eligibility, complete informed consent online, and complete the survey. Informed consent included a teach-back component in which participants answered multiple-choice questions to confirm comprehension of key elements of consent; correct information was shown if they choose the wrong response. If preferred, participants could participate by phone (n = 3 did so). Participation took approximately 30–45 min and participants were mailed a $20 gift card upon completion.

Measures

When asking survey participants to answer questions about their family, instructions advised “think about the people closest to you in your everyday life—it doesn’t matter if they live with you.” We used the networks module [43] from the National Social Life, Health, and Aging Project to help the respondent identify family ties. The networks module asks the respondent to answer a series of questions as each tie is identified. In brief, we first asked the respondent to identify “the person most involved in your diabetes on a regular basis” (primary tie), then identify “other people who are close to you” (close ties), and finally “other than the people you already listed, does anyone else live with you?” (non-close coresidents). For each tie, items queried relationship type, coresidence status, frequency of contact, and diabetes status.

Patient characteristics

Survey participants self-reported sociodemographic information (age, gender, race/ethnicity, income, years of education, marital status, number of people in their household) and diabetes characteristics (diabetes duration in years and months, and insulin status). Participants also completed a validated measure of health literacy, the Brief Health Literacy Screen, where higher scores indicate better health literacy and <9 suggests limited health literacy [44].

Dimensions of diabetes-specific family functioning

We assessed 10 dimensions of diabetes-specific family functioning. Cronbach’s α in our sample is provided when relevant.

We asked participants to characterize their appraisal of T2DM as an individual issue or shared issue using a single item that has been linked to greater use of collaborative coping and greater benefit of collaborative coping for diabetes outcomes [28]. This item specified the primary tie, asking the respondent to pick the response that best characterizes their thoughts about their diabetes: “it is my issue to deal with,” “it is my issue but I know it affects this person,” “it is a shared issue—we deal with it together,” or “it is this person’s issue to deal with.”

Three scales from the Perceptions of Collaboration Questionnaire (PCQ) [31] assessed different aspects of collaboration with the primary tie. Each scale consisted of three items on 5-point response scales that were averaged to generate scores ranging from 1 to 5. Cognitive Compensation assessed the degree to which collaboration is needed to overcome perceived deficits or decline (e.g., “I view working together as necessary because it is harder for me to do things by myself.” α = 0.80). Interpersonal Enjoyment assessed the degree to which collaboration provides encouragement and closeness (e.g., “I enjoy the support and encouragement I receive when we work together.” α = 0.64). Frequency of Collaboration assessed how often collaboration occurs (e.g., “Nearly every day we work together to make decisions or solve problems.” α = 0.79).

For the remaining dimensions, participants were asked to “think about the people closest to you in your everyday life.” The Family/Friend Involvement in Adults’ Diabetes [21] items assessed observable actions relevant to the respondent’s diabetes self-care activities in the prior 30 days. Response options range from 1 = “never in the past month” to 5 = “twice or more each week” and were averaged to generate two separate scores representing frequency of each type of received support. Helpful was assessed with nine items (e.g., “How often do your family members exercise with you or ask you to exercise with them?” α = 0.87) and harmful was assessed with seven items (α = 0.72) querying frequency of undermining/sabotaging behaviors (e.g., “How often do your family members bring foods around that you shouldn’t be eating?”) as well as nagging/arguing about self-care (e.g., “How often do your family members argue with you about your food choices or your health?”).

Autonomy support was assessed using the 6-item Important Other Climate Questionnaire [45] which assessed respondents’ perception of the degree to which others support his or her personal agency. We adapted the items to be specific to diabetes management as the developers recommend [7, 45] and to reference more than one important other (e.g., “My important others try to understand how I see my diabetes before suggesting any changes” α = 0.89). Likert responses (1 = “strongly disagree” to 5 = “strongly agree”) were averaged to create a scale score with higher scores indicating more autonomy support.

Perceived criticism was assessed with four items from the Family Emotional Involvement and Criticism Scale [46] that could be adapted to be specific to diabetes (e.g., “My family finds fault with the things I do to manage my diabetes” α = 0.85). Response options range from 0 = “almost never” to 4 = “almost always” and were summed to generate a scale score with higher scores indicating more perceived criticism.

We assessed participants’ appraisal of their diabetes-specific family functioning with two separate items. One item assessed their appraisal of their family’s effectiveness: “How effective is your family at dealing with troubles or issues related to your diabetes management?” The other item asked about their satisfaction: “How satisfied are you with your family members’ involvement in dealing with troubles or issues related to your diabetes management?” Each item had response options ranging from 1 = “not at all” to 4 = “extremely.” These items had a moderately strong correlation (rho = 0.66, p < .001) and bivariate associations with diabetes outcomes of interest showed differing patterns (Supplementary Table S1) so we included them as separate dimensions in the cluster analysis.

Diabetes outcomes included measures of diabetes management (glycemic control, diabetes medication adherence, diabetes self-efficacy) and psychosocial well-being (diabetes distress and depressive symptoms).

To assess diabetes self-management, we used hemoglobin A1c (HbA1c, %), and measures of diabetes medication adherence and diabetes self-efficacy. Diabetes self-efficacy (also called perceived competence) measures one’s confidence in their own ability to carry out the multiple different behaviors needed to manage diabetes, and predicts multiple different diabetes self-care behaviors [47]. Using self-efficacy as a proxy for the other self-care behaviors is more efficient than assessing relationships with different behaviors. These three measures—including both objective and self-report—provide a picture of diabetes self-management.

We assessed glycemic control with the most recent HbA1c (%) value from the EHR. HbA1c values were dated median 9 days before the survey date (interquartile range, 77 days before to 34 days after). Diabetes medication adherence was assessed with the 11-item Adherence to Refills and Medications Scale for Diabetes [48] which assesses the degree to which respondents experience various barriers to refilling or taking their diabetes medications on a scale from 1 = “none of the time” to 4 = “all of the time.” We reverse coded items to generate a scale score ranging 11–44 with higher scores indicating better adherence (α = 0.81). Diabetes self-efficacy was assessed with a 4-item version of the Perceived Diabetes Self-Management Scale [49] which includes ratings of competence to manage diabetes (e.g., “I am generally able to accomplish my goals with respect to managing my diabetes” α = 0.82). The 8-item version of this scale has demonstrated associations with diabetes self-care behaviors and HbA1c [47].

To assess psychosocial well-being, we used measures of diabetes distress and depressive symptoms. Diabetes distress and depressive symptoms are related but distinct constructs, each of which leads to worse diabetes management, with diabetes distress the more proximal or mediating construct [50]. Diabetes distress was assessed with the 5-item version of the Problem Areas in Diabetes Scale, which has established sensitivity and specificity for recognition of diabetes-related emotional distress [51]. Items assess concern, worry, and fear about diabetes on a scale from 0 = “not a problem” to 4 = “serious problem.” Items are summed to generate a score ranging 0–20 (α = 0.90). Depressive symptoms were assessed with the Personal Health Questionnaire-8 [52] which assesses the frequency of depressive symptoms according to diagnostic criteria on a scale from 0 = “not at all” to 3 = “nearly every day,” yielding a continuous severity score from 0 to 24 (α = 0.87).

Analyses

Analyses are detailed in Supplementary Material. To develop the typology of diabetes-specific family functioning we used k-means cluster analysis with nine of the 10 dimensions assessed. We examined dimension scores across each cluster to ascertain if the clusters were conceptually distinct. We then used an alternative analytic approach (principal components analysis followed by k-means cluster analysis) to assess whether types identified were replicable and stable. Both methods identified four clusters; type membership stability was 90%–99%. Across the two methods, medians and interquartile ranges for each dimension were highly consistent, indicating both methods identified the same patterns across dimensions (Supplementary Table S2). Each participant was assigned to a cluster based on results from the first method. Henceforth, we refer to this cluster membership as “type.” The tenth dimension, shared illness appraisal, was a 4-level categorical variable rather than a continuous or ordinal variable like the other dimensions. Therefore, we examined the frequency of the response options across types with a Pearson chi-squared test.

Next, we used data extracted from the EHR (Table 1) to compare clinical and sociodemographic characteristics between participants and eligible non-participants. According to EHR data, survey participants were more likely to be White and were slightly younger with higher BMI. They were more likely to be prescribed a DPP4, sulfonylurea, and/or metformin and less likely to be diet controlled (i.e., not prescribed medication; Table 1). Because patient characteristics were available via EHR data for all patients meeting inclusion/exclusion criteria, we were able to account for response bias by weighting responding participants to match the characteristics of the eligible cohort [53] with all variables in Table 1. We confirmed the adequacy of our weighting (Supplementary Fig. S1).

Table 1.

Electronic Health Record Data for Eligible Participants and Non-Participants

Variable	N	Eligible participants (n = 379)	Eligible non-participants (n = 5,166)	Total cohort (N = 5,545)	p
Age, years	5,545	59.1 [50.4, 67.0]	60.9 [51.7, 69.0]	60.8 [51.4, 69.0]	.004
Female gender	5,545	50% (188)	49% (2,556)	49% (2,744)	.962
Race	5,504				.001
White		80% (303)	74% (3,797)	74% (4,100)
Black		16% (61)	22% (1,114)	21% (1,175)
Asian		3% (10)	3% (158)	3% (168)
Other		1% (3)	1% (28)	1% (29)
Unknown		1% (2)	1% (29)	1% (31)
Hispanic/Latino	5,545	2% (6)	3% (151)	3% (157)	.129
Body mass index	5,505	32.7 [28.8, 38.5]	31.7 [27.8, 36.8]	31.8 [27.9, 36.9]	.003
HbA1c value, %	5,545	6.9 [6.2, 7.9]	6.9 [6.2, 8.0]	6.9 [6.2, 8.0]	.868
Prescribed a statin	5,545	69% (263)	67% (3,465)	67% (3,728)	.353
Creatinine value	5,540	0.87 [0.76, 1.07]	0.90 [0.77, 1.08]	0.89 [0.77, 1.08]	.056
Systolic blood pressure	5,541	128 [118, 138]	128 [120, 140]	128 [119,139]	.114
Diastolic blood pressure	5,541	76 [68, 82]	75 [68, 82]	75 [68, 82]	.676
Prescribed diabetes medication (yes/no)
DPP4 inhibitors or gliptins	5,545	10% (37)	7% (349)	7% (386)	0.026
Insulin	5,545	21% (81)	20% (1,044)	20% (1,125)	0.587
Sulfonylurea	5,545	28% (108)	23% (1,169)	23% (1,277)	0.009
Thiazolidinediones	5,545	3% (12)	2% (91)	2% (103)	0.051
Metformin	5,545	70% (267)	53% (2,749)	54% (3,016)	<0.001
Other	5,545	3% (11)	2% (122)	2% (133)	0.507
None	5,545	20% (76)	34% (1,743)	33% (1,819)	<0.001

Data are presented as Median [Interquartile Range] or % (n). p values from Wilcoxon rank-sum tests (continuous variable) or Pearson χ ² tests (categorical variables).

We used weighted data to examine unadjusted associations (Kruskal–Wallis or Pearson chi-squared tests) between patient characteristics and types to understand how characteristics of patients and family structure were associated with the family functioning types. We used weighted data to examine cross-sectional associations between type and diabetes outcomes of interest with ordinary least squares regression, first unadjusted and then adjusted for a priori covariates. Covariates were age, gender, race/ethnicity, education, income, diabetes duration, and insulin status (yes/no). Missing data on covariates ranged from none to 4.2% for income. We generated m = 20 imputations for adjusted analyses via multiple imputation using chained equations. We used likelihood ratio tests to test the significance of associations between type and diabetes outcomes and marginal contrasts with 95% confidence intervals to compare types.

Results

A total of 10,077 adults with T2DM met our inclusion criteria; 4,532 (45.0%) were excluded resulting in an eligible cohort of 5,545 patients (Fig. 1). We invited all cohort members with an email address in their EHR to participate. Of the 4,647 (83.8%) invited to participate, 10.7% clicked the link, 8.0% consented, and 7.8% participated in the survey. N = 430 participated in the survey but n = 379 completed at least seven of the nine dimensions needed for cluster analyses. Therefore, we identified these n = 379 as eligible participants and all others in the cohort as eligible non-participants (n = 5,166).

According to participants’ self-report, the average age was 57.8 years (SD = 12.1), 49.5% were female, 77.0% were non-Hispanic White, 14.8% were non-Hispanic Black, 2.1% were Hispanic and the remaining 6.1% reported another race or multiracial. Most (60.2%) had a college degree (14.9% ≤high school degree, 24.9% some college), and 60.9% reported an annual household income ≥$55,000 USD (21.7% <$35,000 USD). Two-thirds (66.1%) were married or had a romantic partner. Most (58.1%) lived with one other person, 19.8% lived alone, and only one lived with ≥4 people. In the networks module, participants reported 2.3 (SD = 0.8) ties, ranging from 0 to 6 ties with interquartile range [2, 3]. Nearly all (98.7%) reported a primary tie, and 86.3% reported other close ties (interquartile range [1,1]). A quarter (27.2%) reported one or two other ties who live with them but were not included as a primary or close tie. Table 2 details identified ties. Average HbA1c from the EHR was 7.2% (SD = 1.5%).

Table 2.

Number and Characteristics of Different Types of Ties Respondents Identified

	Primary tie	Close ties	Non-close co-residents
Number reported	N = 374	N = 371	N = 120
Characteristics	% (n)	% (n)	% (n)
Relationship type
Spouse/partner/romantic	62.5% (235)	5.7% (21)	5.7% (7)
Son/daughter	10.4% (39)	31.8% (118)	58.3% (70)
Sibling	3.7% (14)	13.5% (50)	5.8% (7)
Parent	6.4% (24)	14.0% (52)	10.0% (12)
Other relative	1.9% (7)	2/7% (10)	8.3% (10)
Friend	7.5% (28)	24.5% (92)	9.2% (11)
Other (coworker, clergy, counselor)	7.2% (27)	7.5% (28)	2.5% (3)
Coresident status
Coresident	73.0% (273)	16.2% (60)	100% (120)
Same state	22.2% (83)	61.2% (227)	NA
Different states	4.8% (18)	22.4% (83)	NA
Frequency of contact
Daily	84.2% (315)	36.1% (134)	70.0% (84)
Weekly	9.9% (37)	45.6% (169)	20.0% (24)
Monthly	2.4% (9)	11.3% (42)	3.3% (4)
Less than monthly	3.5% (13)	7.0% (26)	5.8% (7)
Diagnosed with diabetes	19.4% (72)	16.2% (60)	12.5% (15)

Typology of Diabetes-Specific Family Functioning

Figure 2 shows standardized dimension scores for each of the four types of diabetes-specific family functioning identified. The means, standard deviations and bivariate correlations for each measure are shown in Supplementary Table S1. Shared illness perception was associated with type of diabetes-specific family functioning (p < .001; Supplementary Table S3) as described below.

Fig. 2.

Standardized scores for dimensions and summary descriptions for each type identified with cluster analyses.

Each type was named based on patterns across the 10 dimensions. Collaborative and Helpful (n = 128, 33.8%) was characterized by high collaborative coping, helpful involvement, autonomy support, and very high satisfaction and effectiveness, combined with low harmful involvement and perceived criticism. Participants with this type were much more likely to characterize T2DM as a shared issue relative to other types (54.2% vs. <20% for other types). Satisfied with Low Involvement (n = 84, 22.2%) was characterized by very low or low scores across all dimensions combined with high satisfaction. Participants with this type were much more likely to characterize T2DM as an individual issue to deal with relative to other types (80.9% vs. <60% for other types). Want More Involvement (n = 112, 29.6%) was characterized by moderate collaborative coping, combined with very low or low scores across other dimensions (like Satisfied with Low Involvement), but is distinct due to very low effectiveness and satisfaction. Participants with this type were likely to describe T2DM as either an individual issue (44.4%) or an individual issue that affects their primary tie (39.3%; i.e., the two more individualized rather than the two more shared responses). Critically Involved (n = 55, 14.5%) was characterized by moderate levels of collaborative coping and autonomy support with the highest levels of helpful and harmful involvement and very high perceived criticism. Participants in this type reported moderate effectiveness but low satisfaction. Like Want More Involvement, this type characterized T2DM with the more individualized response options (52.6% individual issue, 32.5% individual issue that affects their primary tie).

Participants’ age, gender, marital status, health literacy, and number of people living with the participant were each associated with type (all p’s < .05, Supplementary Table S3). Race/ethnicity, educational attainment, income, diabetes duration, and insulin status were not. In both unadjusted and adjusted analyses, type was associated with each diabetes outcome (Fig. 3). Figure 3A shows that unadjusted patterns in diabetes outcomes across types were consistent; HbA1c, distress, and depressive symptoms got higher (worse) and medication adherence and diabetes self-efficacy got lower (worse) in concert when looking across the types from Collaborative and Helpful to Critically Involved. Adjusted marginal contrasts are shown in Fig. 3B and described in summaries of each type below.

Fig. 3.

Associations between the type of diabetes-specific family functioning and standardized outcomes of interest. (A) Weighted mean scores on outcomes of interest by family functioning type. Unadjusted F-tests for associations between type and each outcome, all p < .001. (B) Marginal contrasts: Weighted adjusted effect sizes and 95% confidence intervals. Models adjusted for age, race/ethnicity (non-Hispanic White/non-Hispanic Black/Other), gender (female/male), education (high school degree or less, some college, college degree, or more), income, diabetes duration, and insulin status (yes/no). Likelihood ratio chi-squared tests (3 degrees of freedom) used to evaluate the adjusted association between type and each outcome.

Collaborative and helpful

Participants with this type were most likely to be male (59.9%) and married/partnered (78.8%), and least likely to live alone (9.5%). They also had the lowest health literacy scores (13.7 vs. ≥14.1 for other types). This type had the highest unadjusted mean values on self-efficacy and adherence and lowest mean HbA1c, as well as relatively low mean values on diabetes distress and depressive symptoms. In adjusted contrasts between types, Collaborative and Helpful was the only type to have significantly (and substantively) better glycemic control than other types –0.55 [–0.93, –0.16] relative to Satisfied with Low Involvement, –0.43 [–0.80, –0.07] relative to Want More Involvement, and –0.84 [–1.32, –0.36] relative to Critically Involved.

Satisfied with low involvement

Participants with this type of family functioning were least likely to be married/partnered (45.1%) and most likely to live alone (39.7%). This type had the lowest depressive symptoms and the lowest diabetes distress in unadjusted and adjusted analyses (significantly less diabetes distress relative to Want more Involvement and Critically Involved; Fig. 3B). Medication adherence and self-efficacy were better than Want More Involvement and Critically Involved, although HbA1c was significantly worse than those characterized as Collaborative and Helpful in adjusted analyses (0.55 [0.16, 0.93]) representing a real difference in HbA1c of 0.8% [0.2%, 1.4%].

Want more involvement

Participants with this type were likely to be married/partnered (67.2%). This type had more depressive symptoms, more diabetes distress, less self-efficacy, and worse medication adherence than the prior two types. Despite doing worse on each of these outcomes, HbA1c was similar to Satisfied with Low Involvement.

Critically involved

Participants with this type were the youngest (mean age 52.1 years vs. >57 years for other types), most likely to be female (59.0%), less likely to be married/partnered (54.7%) and less likely to live alone (17.2%) compared with other types. Combined, these structural characteristics suggest this type has slightly more cohabitation with people other than a partner. This type had significantly worse diabetes distress and depressive symptoms compared to all other types, and worse self-efficacy and medication adherence than Collaborative and Helpful and Satisfied with Low Involvement (Fig. 3B). HbA1c was the highest for this group, with a substantial adjusted contrast to Collaborative and Helpful (0.84 [0.36, 1.32]) representing a real difference in HbA1c of 1.3% [0.5%, 2.0%].

Discussion

We developed an empirical typology of adults’ diabetes-specific family functioning that was replicable across different analytic techniques, applicable across diverse family structures, and cross-sectionally associated with distinct clinical profiles including measures of diabetes self-management and psychosocial well-being. The multidimensional assessment provided by the typology was more illustrative than any single measure. We identified four types or profiles of diabetes-specific family functioning. Collaborative and Helpful had the best overall diabetes outcomes, Satisfied with Low Involvement had relatively good diabetes self-management and the lowest diabetes distress, Want More Involvement had worse self-management and higher diabetes distress, and Critically Involved had the worst overall diabetes outcomes.

These four types and their distinguishing characteristics have important theoretical implications for family in T2DM. Our results extend a theoretical literature on communal coping that has focused nearly exclusively on romantic partners (e.g., nearly 40% of our sample had a primary tie that was not a romantic partner). This extension is necessary; an increasing number of individuals are not partnered [54] and adults with T2DM receive regular support for diabetes self-management from adult children and loved ones who live separately and long-distance [55]. Attending only to in-home family members or romantic partners limits our understanding of the role of family. The Collaborative and Helpful type is consistent with dyadic [10] and communal [8, 9] coping theories in that this group appraised T2DM as shared and reported high levels of working together, with overall positive self-management and psychosocial well-being. Frequently, these dyadic and communal coping theories make a distinction between supportive and collaborative behaviors; however, similar to our findings, recent work indicates that individuals do not make distinctions between working together collaboratively and instrumental supportive behaviors [28]. Similarly, most theories of family support [33] also include harmful aspects, although frameworks differ as to the specific nature of that harmful involvement (e.g., whether it be criticism, control, etc.). Our results support assertions that helpful and harmful aspects both happen in the same families and add to this literature by noting the relative amounts of helpful and harmful (high/high, low/low, and low/high) appear critically important. Another novel theoretical contribution of our study was the importance of satisfaction with low levels of involvement. Those who were Satisfied with Low Involvement may be independent and successful in their management, either because self-care activities are going well and/or due to more independent temperamental styles. Whereas those who Want More Involvement showed evidence of struggling with T2DM in similar low involvement families. This has been described as person-environment fit [33] and our study suggests it may be integral for understanding diabetes distress.

Our findings also have implications for the design and evaluation of interventions to improve diabetes outcomes. Interventions focused on improving the quantity or quality of diabetes-specific family involvement should focus on meeting the needs of patients who were characterized as Want More Involvement and Critically Involved. These two types had the worst outcomes and made up 44% of the sample. Critically Involved appears to be most in need of effective interventions, as this group was youngest with the worst outcomes. Our findings indicate one potential reason for very low satisfaction among Want More Involvement may be a higher need for collaboration to compensate for perceived cognitive deficits. Improved communication with family about patient needs and preferences may help alleviate diabetes distress. Interventions that coach individuals as to how to identify and ask for the kind of support they want/need, like the FAMS intervention [56], may enhance communication skills to use with family and friends about diabetes self-care goals and be especially beneficial for these two types. Interventions requiring family member participation may be fraught. Critically Involved may have someone readily available to participate but interventions may inadvertently increase criticism and harmful involvement unless content explicitly addresses this carefully. Want More Involvement may have family members who are unwilling to participate and/or provide the level of involvement desired, potentially exacerbating already elevated diabetes distress. The Collaborative and Helpful type may enjoy participating in (and readily sign up for) interventions with their family members but have less need for these types of interventions.

Concurrent with this study, Bouldin et al. [57] identified four relationship profiles among patients with heart failure and an out-of-home friend or family member with similarities to those identified here, suggesting these profiles may generalize across chronic health conditions. Both studies identified a more collaborative type and a more conflictual argumentative type. Boudin et al. also identified Avoidant and Distant types, which shared similarities with Want More Involvement and Satisfied with Low Involvement, respectively, but their profile did not include measures of patients’ appraisal, so it is difficult to conclude these types were analogous.

Strengths, Limitations, and Next Steps

Respondents considered several people (most often 2 or 3) when reporting on their family functioning which is both a strength and a weakness. We cannot examine concordance with family member report, which may have been illuminating. However, our study provides a view of family functioning that is not restricted to family members who would participate in research about the patients’ T2DM, thereby enhancing the generalizability of our findings. Our study also had the major strength of leveraging EHR data to address response bias. As a result, findings are representative of the cohort rather than only of survey participants. Nonetheless, the cohort includes patients receiving regular care from an academic medical center and was largely non-Hispanic white, with predominantly middle-to-high socioeconomic status, who used e-mail and had relatively good glycemic control. Replication in more diverse samples is needed to confirm the generalizability and utility of the typology. Other important next steps to enhance the utility of the typology in future research and clinical care include (a) examining if type predicts longitudinal outcomes to enhance clinical validity, (b) examining associations between type and different aspects of diabetes self-care (e.g., dietary behavior, physical activity), and (c) reducing the number of items needed to identify type (our assessment used 37 items, not including the networks module).

Conclusion

Typologies are pragmatic tools that provide organizational frameworks for complex phenomena [38]. Acknowledging and measuring multidimensionality can lead to breakthroughs that are unattainable with single measures. Applications of this typology can lend cross-study consistency in tailoring and support efforts to enhance the fit between the patient-family needs and the intervention type (individualized vs. family member enrolled) and intervention content [58]. Heterogeneity in treatment effects of family interventions for adults with T2DM [59] may be explained by type of family functioning, and explorations of effect modification by type may accelerate learning about what works for whom.

Acknowledgments This research was supported by the National Institute for Diabetes and Digestive and Kidney Diseases via R03-DK113329 (PI Mayberry) and the Vanderbilt Center for Diabetes Translation Research (P30-DK092986, PI Elasy), and used resources supported by the Vanderbilt Institute for Clinical and Translational Research grant (ULTR000445 from the National Center for Advancing Translational Sciences). Dr. Mayberry was supported by a career development award from the National Institute for Diabetes and Digestive and Kidney Diseases (K01-DK106306).

Compliance with Ethical Standards

Authors’ Statement of Conflict of Interest and Adherence to Ethical Standards Authors declare that they have no conflict of interest.

Authors’ Contributions L.S.M. was the principal investigator and designed the research study, oversaw data collection, and wrote the manuscript. L.S.M. is the guarantor of the work. R.A.G. planned the analyses and oversaw all stages of analysis and interpretation. L.-C.H. and S.Z. conducted analyses and wrote the online-only supplement. S.Z. also provided expertise in cluster analysis. C.A.B. provided content expertise throughout the project, aided in selection of the dimensions and measures, and participated in interpretation and naming of the types. All authors reviewed and edited the manuscript.

Ethical Approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed Consent Consent for electronic health data to be used for identification to participate in research is obtained in the course of clinical care. Documented informed consent for this research study was obtained from all participants in the survey.