Mixed-Methods Randomized Evaluation of FAMS: A Mobile Phone-Delivered Intervention to Improve Family/Friend Involvement in Adults’ Type 2 Diabetes Self-Care

Abstract

Background

Family and friends have both helpful and harmful effects on adults’ diabetes self-management. Family-focused Add-on to Motivate Self-care (FAMS) is a mobile phone-delivered intervention designed to improve family/friend involvement, self-efficacy, and self-care via monthly phone coaching, texts tailored to goals, and the option to invite a support person to receive texts.

Purpose

We sought to evaluate how FAMS was received by a diverse group of adults with Type 2 diabetes and if FAMS improved diabetes-specific family/friend involvement (increased helpful and reduced harmful), diabetes self-efficacy, and self-care (diet and physical activity). We also assessed if improvements in family/friend involvement mediated improvements in self-efficacy and self-care.

Methods

Participants were prospectively assigned to enhanced treatment as usual (control), an individualized text messaging intervention alone, or the individualized text messaging intervention plus FAMS for 6 months. Participants completed surveys at baseline, 3 and 6 months, and postintervention interviews. Between-group and multiple mediator analyses followed intention-to-treat principles.

Results

Retention, engagement, and fidelity were high. FAMS was well received and helped participants realize the value of involving family/friends in their care. Relative to control, FAMS participants had improved family/friend involvement, self-efficacy, and diet (but not physical activity) at 3 and 6 months (all ps < .05). Improvements in family/friend involvement mediated effects on self-efficacy and diet for FAMS participants but not for the individualized intervention group.

Conclusions

The promise of effectively engaging patients’ family and friends lies in sustained long-term behavior change. This work represents a first step toward this goal by demonstrating how content targeting helpful and harmful family/friend involvement can drive short-term effects.

Trial Registration Number

NCT02481596.

A mobile phone-delivered intervention improved family support for diabetes self-care among diverse adults. Improvements in family support drove improvements in diabetes self-efficacy and diet behavior.

Introduction

Sustaining daily Type 2 diabetes self-care is challenging, and disparities based on race/ethnicity and socioeconomic status persist [1, 2]. Family members and close friends are often involved in or present for adults’ daily self-care efforts [3, 4]. According to family systems theory [5], interventions that improve the family and social context in which self-care occurs may be more effective in sustaining health behavior change than interventions directed only at individual change. Family systems theory posits that behavior change is initiated by the individual but sustained by the system through feedback loops [5]. When an individual introduces a behavior change, family/friends’ responses/actions either reinforce the change (e.g., making it easier or more rewarding to perform the new behavior) or undermine the change (e.g., making it more difficult or less rewarding). This affects the likelihood that the individual will maintain the change—either positively or negatively—creating a feedback loop. If family/friend reactions foster the change, it is more likely to be sustained. This theory is consistent with prior findings that instrumental received support from family and close friends (i.e., practical, observable support that makes self-care possible or easier) is more strongly associated with regular self-management than other types of support (e.g., emotional support) among adults managing chronic illnesses like diabetes [6, 7].

Despite robust effects of family/friend instrumental support on long-term self-management in Type 2 diabetes [8–10], family interventions have not been consistently effective [7, 11, 12]. One potential reason may be the lack of attention to both helpful and harmful aspects of received support for adults’ diabetes self-management [13, 14]. Research among adults with Type 2 diabetes has consistently shown that support for self-care behaviors (e.g., preparing healthy meals, collaborative problem solving in response to blood glucose readings, and assisting with medication reminders) [7, 10, 15, 16] often co-occurs with harmful involvement [7, 10, 13, 16–20]. This harmful involvement includes sabotaging/undermining (making self-care behaviors more difficult through, for instance, bringing around unhealthy foods or suggesting that diabetes medications are not needed) and nagging/arguing about self-care, which reduces patient self-efficacy and motivation [7, 16, 17, 20, 21]. Very few interventions engaging patients’ friends/family include content attending to harmful involvement [11]. This may be a critical oversight because increasing social support for self-care may inadvertently increase harmful involvement. For instance, friends/family receiving education about diabetes management may begin to nag or “police” the patient about self-care, having detrimental effects. Harmful family/friend involvement is more strongly associated with adherence and glycemic control than helpful involvement [7, 9, 10, 16], particularly among patients with vulnerabilities such as limited health literacy [22] and increased life stressors associated with financial strain [23].

Alternative approaches to interventions engaging family support for adults’ Type 2 diabetes management may be needed to enhance inclusivity, relevance to adults’ lived experiences, and efficacy. First, it may be important to broaden the definitions of “family” [12, 14] to include close friends, neighbors, and coworkers involved in daily self-care. Given modern demographic trends to not marry and have fewer children, the availability of family caregivers may be reduced [24]. Thus, interventions should recognize the diversity of social interactions and meaningful relationships for adults [14], with interventions that empower patients to manage multiple social relationships having more traction than those targeting a single relationship. Second, out-of-home and long-distance friends/family members are an underutilized resource for supporting chronic disease self-management [25], and interventions delivered via technology can engage long-distance supporters. Finally, family/social support intervention research that evaluates the targeted mechanisms will help us understand why interventions were or were not effective in improving patients’ behavioral or clinical outcomes [11, 26, 27].

To address these issues, we developed a mobile phone-delivered diabetes self-care support intervention called FAMS: Family-focused Add-on to Motivate Self-care [21]. Based on the concept of feedback loops in family systems theory [5], FAMS was designed to assist the person with diabetes in initiating diet and/or physical activity goals and equip him/her to manage feedback loops by learning skills to anticipate, shape, and manage responses—both helpful and harmful—received from family/friends as they meet new goals. Figure 1 shows the elements of the FAMS intervention mapped onto the family systems theory concept of feedback loops. FAMS participants received phone coaching and text message support and also had the option to enroll a support person. We sought to evaluate the acceptability of engaging adults’ family/friends with text messaging [28] and to explore if intervention effects depended on the inclusion of a participating support person.

Fig. 1.

Family systems theory feedback loop elements mapped onto FAMS intervention components.

Objective and Hypotheses

Ultimately, we are interested in whether improving family/friend involvement in patients’ self-care goals can sustain behavior change longer than interventions that do not improve social contexts, as suggested by family systems theory. However, our immediate goals for the present study are to determine whether FAMS was acceptable and well received in a diverse sample of adults with Type 2 diabetes, to ensure that the intervention improved the response/actions from family/friends as patients meet self-care goals (i.e., increased helpful received support without increased harmful received support), and to ascertain if improvements in family/friend involvement mediated improvements in dietary behavior, physical activity, and diabetes self-efficacy over 6 months. We used mixed methods to evaluate FAMS in a randomized controlled trial (RCT), in which participants were assigned to control, an individualized text messaging intervention alone, or to the individualized intervention plus FAMS. We hypothesized that FAMS would improve family/friend involvement, self-efficacy, diet and physical activity. We explored potential effect modifiers, including minority race/ethnicity, socioeconomic status (SES), and gender. We hypothesized that improvements in helpful and harmful family/friend involvement would mediate improvements in self-efficacy and self-care behaviors, whereas the individualized (i.e., not family/friend-focused) intervention would not improve family/friend involvement and would affect these outcomes directly. We further hypothesized that FAMS participants would experience improved family/friend involvement regardless of their choice to enroll a support person because FAMS targets patients’ skills to manage multiple family/friend interactions.

Research Design and Methods

This work was part of a larger parallel randomized trial examining effects of an individualized text message-delivered intervention called Rapid Encouragement/Education And Communications for Health (REACH) on patients’ medication adherence and glycemic control over 15 months. The intervention of interest, FAMS, was developed as a set of components to be delivered alongside REACH. Both interventions were developed with and for racially diverse adults with predominantly low SES via iterative usability testing [21, 29]. Content development included attention to health literacy and readability following a five-step methodology [30] as detailed previously [21]. The study protocol, including recruitment and randomization procedures, analytic plans, and secure data management, has been published [31]. The Vanderbilt Institutional Review Board approved all procedures prior to enrollment. The study is registered with ClinicalTrials.gov (NCT02481596).

Analyses to evaluate the effects of REACH + FAMS relative to REACH only and control from baseline through 6 months are presented herein. All analyses were planned a priori [31]; between-group analyses on family/friend involvement, self-efficacy, diet, and physical activity are registered with ClinicalTrials.gov.

REACH provided daily text message support, with content either tailored to address barriers to medication adherence or nontailored to address other self-care behaviors (i.e., diet, physical activity, and self-monitoring of blood glucose). REACH also sent daily text messages asking if the participant took all his/her prescribed diabetes medicine for the day, followed by a weekly feedback message on adherence progress. Details of REACH text message content and functionality have been published [29].

Participants assigned to REACH + FAMS received all the same text messages as those assigned to REACH only, except nontailored self-care texts were replaced with texts tailored to goals set during coaching. Therefore, REACH only and REACH + FAMS participants received the same frequency and pattern of texts during the 6 month intervention period [31]. Only participants assigned to receive REACH + FAMS set self-care goals, received monthly phone coaching with a trained coach, and were given the option to invite a support person to receive text messages (Fig. 1).

Recruitment, Enrollment, and Randomization

Recruitment for the RCT occurred from May 2016 through December 2017 from clinic sites in and around Nashville, TN, including 13 community clinic locations and three Vanderbilt University Medical Center (VUMC) adult primary care clinics. We oversampled patients with minority race/ethnicity or who had no insurance or public insurance only when recruiting from VUMC clinics to ensure similarity with patients recruited from the community clinics. Eligible participants were adults (age ≥18 years) with Type 2 diabetes receiving care at one of the clinics who were prescribed at least one daily diabetes medication, responsible for taking their own medications, owned a cell phone, and could speak and read English. Exclusion criteria included most recent electronic health record (EHR) hemoglobin A1c (%) value within 12 months <6.8%, auditory or communication limitations that precluded participation in phone coaching, failing a brief cognitive screener, and inability to receive and respond to a text after brief training from a research assistant (RA). Eligibility was assessed via EHR review, screening interview, and during the enrollment appointment. We screened 1,244 patients for eligibility (Fig. 2); 70% (512/733) of those eligible enrolled.

Fig. 2.

Participant recruitment and retention flowchart.

Enrolled participants were assigned to control, REACH only, or REACH + FAMS using a 2:1:1 design after completing all enrollment procedures, including informed consent. A computer randomization program used optimal multivariate matching [32] to ensure that the conditions remained balanced on baseline A1c, diabetes duration, insulin status, race, age, gender, income, and education. After randomization, RAs called participants to explain what to expect and collect condition-specific information (i.e., preferred times of day for text messages, complete first coaching session, and obtain support person information if assigned to REACH + FAMS). All participants—including those assigned to the control condition—received access to a study helpline for questions about the study and their diabetes medications (a clinical pharmacist returned calls), text messages advising how to access study A1c results, and quarterly newsletters on healthy living with diabetes.

FAMS Intervention

FAMS included monthly family/friend-focused phone coaching sessions, text messages tailored to the diet or physical activity goal set during coaching (in addition to the medication adherence support provided by REACH messages), and the option to invite an adult support person to receive text messages (Fig. 1). Phone coaching was designed to improve the patient’s ability to identify family/friends’ actions that support or impede self-care goals and skills to ask for needed support and manage harmful actions to meet these goals [21]. Coaches followed a protocol to assist participants in identifying a diet or physical activity goal and to learn skills to elicit helpful involvement and redirect/cope with harmful involvement. Coaches were master’s level counselors, counselors-in-training, or health coaches with training in the FAMS protocols. Coaching occurred monthly for six full sessions (each lasting 20–30 min) and a brief wrap-up session and occurred with the person with diabetes alone. Each of the six core coaching sessions included the following components:

• (1) Check-in—assess participant’s current diet/activity and/or success with prior goal and assess experiences of family/friend involvement generally and related to goal pursuit.

• (2) Goal setting—set new or adjust prior goals. Goals were SMART (specific, measurable, attainable, realistic, and time bound), pertained to diet or physical activity, and could be performed daily (e.g., to walk 15 min 4 days per week). Goals were set and adjusted collaboratively with the coach during each session.

• (3) Discuss family/friend involvement—discuss potential or experienced helpful/harmful involvement relevant to the goal. Participants could and often did alternate between diet and physical activity goals across coaching sessions.

• (4) Skill building—administer a skill-building exercise (Fig. 1), including role-playing, brainstorming, or teach-back. For Session 1, the skill-building exercise included a brief didactic section on the role of others’ in diabetes self-management, followed by homework to observe and note family/friend responses as the participant worked to reach his/her goal. In Sessions 2–6, skill-building exercises (Fig. 1) were selected by the coach based on participants’ needs and could be repeated when relevant. All focused on skills to manage helpful/harmful family/friend behaviors.

• (5) Verbal contract—make a commitment to implement the skill with a friend/family member (who may be the same or different from the invited/enrolled support person).

Participants could choose to invite a support person to receive texts (Fig. 1) during any of the first four coaching sessions. Participants who chose to invite a support person provided a first name and mobile phone number, and an RA called to invite the support person to participate by completing verbal consent and survey via phone, receiving one-way text messages (three per week), and completing a follow-up interview after the intervention experience. Eligible support persons were adults (≥18 years old) who owned a mobile phone with text messaging capabilities and could read and speak in English. Importantly, there were no support person eligibility criteria regarding cohabitation or type of relationship with the patient. As published previously [28], participants who invited a support person did not differ from those who did not on race/ethnicity, gender, income, education, insurance status, diabetes duration, insulin status, health literacy, or self-care behaviors. However, they were slightly younger, more likely to be married/partnered, and reported more depressive symptoms, more helpful family/friend involvement, and more emergency department use in last 12 months (all ps < .05).

Data Collection

Participants completed baseline and 3 and 6 month follow-up assessments either in a private room at the clinic, phone, web, or mail per preference. Baseline surveys included sociodemographic information. Each assessment period included EHR review. All data were entered into Research Electronic Data Capture (REDCap) [33]. MEMOTEXT tailored and sent study text messages according to predefined intervention algorithms using data entered into REDCap via a secure application programming interface [31]. FAMS coaches were not involved in follow-up data collection.

Measures

Surveys assessed outcomes with self-report measures. The Family/friend Involvement in Adults Diabetes (FIAD) subscales [9] measure helpful (i.e., observable behaviors that make self-care possible/easier) and harmful (i.e., sabotaging/undermining and arguing) aspects of received support. The FIAD assessed helpful (nine items, e.g., “How often do your friends or family members exercise with you or ask you to exercise with them?”) and harmful (seven items, e.g., “How often do your friends or family members bring foods around that you shouldn’t be eating?”) aspects of received support for diabetes applicable to diverse living situations and relationships. Response options range from 1 = “never in the past month” to 5 = “twice or more each week” and are averaged to generate two separate scores. The scales have demonstrated reliability and associations with self-care behaviors and glycemic control [9] (Cronbach’s α = .87 and α = .63 in our sample).

Diabetes self-efficacy was assessed with the Perceived Diabetes Self-Management Scale, four-item version [34], which includes ratings of competence to manage diabetes (e.g., “I am generally able to accomplish my goals with respect to managing my diabetes”) with response options on a five-point scale ranging from “strongly disagree” to “strongly agree.” In our data, the four-item version had acceptable internal consistency reliability (α = .68). The eight-item version has demonstrated associations with diabetes self-care behaviors and glycemic control [35].

Dietary behavior was assessed with the diet subscales of the Personal Diabetes Questionnaire: use of dietary information for decision-making and eating behavior problems [34]. Items assessed the frequency of using different types of information to make dietary choices (e.g., “Over the past month, how often did you use information about the number of carbohydrates in foods to make decisions about what to eat?”) and frequency of problematic eating behavior (e.g., “Over the past month how often did you eat unplanned snacks?”) with response options on a six-point scale from “never” to “one or more times per day.” These subscales have excellent reliability (α = .85 and α = .88 in our sample) and associations with measures of diabetes self-care behavior and self-efficacy [34]. The two three-item subscales have a low correlation (rho = .10, p = .013 in our sample); therefore, we analyzed them separately as recommended by the developers.

Physical activity was assessed using the International Physical Activity Questionnaire-short form, which generates a total average metabolic equivalent (MET) minutes per week [36, 37] by querying frequency and duration of four different types of physical activity (e.g., average number of minutes per week doing moderate physical activity). This measure has been validated in many countries and patient populations [37].

Intervention engagement and fidelity

Participants’ responses to weekly interactive text messages were tracked by MEMOTEXT. We tracked the number of FAMS coaching sessions completed by each participant. The outcomes of each coaching component were also tracked: the goal participants set during coaching, type of family/friend involvement discussed, the skill-building exercise employed, the verbal contract, participant’s confidence rating of his/her ability to complete the verbal contract, and, for subsequent sessions, the outcome of the verbal contract from the previous session. The principal investigator met with coaches twice monthly to review coaching sessions and discuss fidelity data.

Participant feedback

All participants who completed the REACH + FAMS intervention during a predetermined timeframe (July through November 2017) were invited to complete a feedback interview. Questions included: What was the most/least useful thing you did in coaching? Is there anything you wish had been discussed in coaching? In your opinion, what role do family and friends play in your diet/physical activity/diabetes? How did your thoughts change about this during your experience? Thematic analysis [38] was used to identify themes in responses. We used techniques from both inductive and deductive thematic analysis [38, 39]; codes were based on both preliminary open coding from the transcripts performed by E.B. and the refined based on the literature on family/friend involvement per feedback from L.S.M. This approach integrates data-driven codes with theory-driven codes or grouping of codes [39]. After this refinement process, the codebook was finalized, and all interviews were coded independently by E.B. One-third were also coded by a trained research assistant to determine interrater reliability (pooled kappa = 0.85).

Between-group and Subgroup Effects

Targeted mediators were helpful family/friend involvement and harmful family/friend involvement. The primary outcomes were dietary behavior and physical activity and the secondary outcome was diabetes self-efficacy. Analyses followed intention-to-treat principles [40], with all participants’ data analyzed according to randomization regardless of compliance with or completion of their assigned intervention.

Between-group analyses compared improvements in the REACH + FAMS group and, separately, the REACH-only group relative to the control group on mediators and outcomes at 3 and 6 months. Ordinary least squares regression models examined the change in the mediator/outcome of interest adjusted for the baseline value with restricted cubic splines to allow for a nonlinear effect of the baseline value. Adjusting for the baseline value of the outcome of interest models change in the outcome from baseline to follow-up and enhances the precision of estimates. Restricted cubic splines accommodate potential nonlinear relationships between the baseline value and the later value of the outcome (e.g., a low baseline value has more opportunity for improvement over time than a high baseline value). We ran separate models to examine the 3 and 6 month effects. Regression models used m = 20 datasets generated with multiple imputation using chained equations. Imputation models included all variables used in analyses.

To explore potential moderators of intervention effects, we then added interaction terms to outcome models examining REACH + FAMS versus control, each in a separate model. Potential effect modifiers included racial/ethnic minority status (vs. non-Hispanic White), socioeconomic disadvantaged status (as indicated by any one of the following at baseline: income <$25,000 USD, homelessness, uninsured, and years of education completed <12), and gender. Because we were underpowered to detect interaction effects (i.e., if the effect in one subgroup was significantly different than the effect in another), we also analyzed subgroup effects (i.e., to determine whether intervention effects in each subgroup were significantly different from zero).

Mediation Analyses

We tested if improvements in response from family/friends mediated improvements in behavior and self-efficacy as posited by the family systems theory concept of feedback loops (Fig. 1). We used change in diabetes-specific helpful and harmful family/friend involvement as our assessments of response from family/friends. We tested multiple mediator models with bootstrapped confidence intervals (CIs) for indirect effects [41, 42]. All mediation models included condition as the predictor, the outcome of interest at 6 months (i.e., self-efficacy, use of dietary information for decision-making, problem eating behaviors, and physical activity—each outcome examined in a separate model), and both mediators: change in helpful family/friend involvement from baseline to 6 months and change in harmful family/friend involvement from baseline to 6 months. Each model included the baseline value of each outcome of interest as a covariate to model change in the outcome from baseline to 6 months. We also examined mediation models with condition further specified as REACH + FAMS with a support person versus REACH + FAMS without a support person versus control.

Due to challenges with obtaining bootstrapped CIs when using multiply imputed data [43], we ran mediation models using complete cases (14%–18% missing depending on comparison and outcome) to obtain point estimates and bootstrapped asymmetric 95% bias-corrected CIs for indirect effects (3,000 bootstrapped samples) [41, 42]. We then ran a sensitivity analysis with the multiply imputed data to assess whether point estimates were substantively biased by missing data. Results from the mediation analyses on imputed data were consistent with findings from the original data, so we present results from the complete case data for ease of interpretation.

Results

We enrolled N = 512 participants, with a mean age of 56.0 (standard deviation: 9.5) years old and 54% females. Over half the sample (53%) were non-White, including 39% non-Hispanic Black and 6% Hispanic. Participants with lower SES were oversampled: 42% had a high school degree or less education, 56% had annual incomes less than $35,000, and 49% were uninsured or had public insurance only. Matched randomization led to strong balance across conditions at baseline (Table 1), with all standardized mean differences <0.3 and all p values >.05. Study measure completion among n = 506 randomized was 95% at 3 months and 92% at 6 months. Retention was not different across conditions (Fig. 1).

Table 1.

Sample characteristics at baseline

Median [IQR] or n (%)	Control N = 253	REACH only N = 127	REACH + FAMS N = 126
Participant characteristics
Age, years	57.0 [50.7, 63.3]	57.4 [50.4, 62.8]	57.2 [49.6, 64.3]
Gender, male	118 (47%)	57 (45%)	57 (45%)
Race/ethnicity
Non-Hispanic White only	121 (48%)	62 (49%)	59 (47%)
Non-Hispanic Black only	99 (39%)	49 (39%)	50 (40%)
Hispanic	15 (6%)	8 (6%)	8 (6%)
Other and multiracial	18 (7%)	8 (7%)	9 (7%)
Education, years	14.0 [12.0, 16.0]	13.0 [12.0, 16.0]	14.5 [12.0, 16.0]
Income (annual household, USD)
<$10,000	53 (21%)	22 (17%)	17 (13%)
$10,000–$34,999	93 (37%)	41 (32%)	55 (44%)
$35,000–$54,999	36 (14%)	20 (16%)	12 (10%)
≥$55,000	51 (20%)	32 (25%)	32 (25%)
Refused/don’t know	20 (8%)	12 (9%)	10 (8%)
Health insurance
Uninsured	66 (26%)	26 (20%)	25 (20%)
Public only	65 (26%)	30 (24%)	31 (25%)
Private	121 (48%)	69 (54%)	69 (55%)
Refused/don’t know	1 (0%)	2 (2%)	1 (1%)
Diabetes duration, years	10.0 [4.0, 16.0]	10.0 [5.5, 15.0]	10.0 [4.0, 15.0]
Diabetes medications
Oral only	132 (52%)	60 (47%)	68 (52%)
Insulin only	42 (17%)	23 (18%)	18 (17%)
Oral and insulin	79 (31%)	44 (35%)	40 (31%)
Clinic site: community clinic	109 (43%)	54 (43%)	53 (42%)
Hemoglobin A1c, %	8.2 [7.2, 9.5]	8.4 [7.3, 9.7]	8.0 [7.2, 9.5]
Outcomes of interest
Family/friend involvement
Helpful involvement	1.9 [1.2, 2.6]	1.6 [1.1, 2.4]	1.8 [1.2, 2.3]
Harmful involvement	1.6 [1.4, 2.0]	1.6 [1.1, 2.0]	1.6 [1.3, 2.1]
Diabetes self-efficacy	14.0 [11.0, 17.0]	14.0 [12.0, 17.0]	14.0 [12.0, 16.0]
Use of dietary information	2.7 [1.0, 4.3]	3.0 [1.0, 4.7]	2.7 [1.3, 4.0]
Problem eating behaviors	3.3 [2.7, 4.0]	3.3 [2.3, 4.0]	3.3 [2.7, 4.0]
MET minutes per week	1,299 [236, 3,492]	1,148 [520, 3,560]	1,158 [339, 2,655]

Strong balance across arms; all standardized mean differences <0.3 and all p values >.05. Family/friend involvement assessed with the helpful and harmful involvement subscales of the Family/friend Involvement in Adults Diabetes measure. Diabetes self-efficacy assessed with the Perceived Diabetes Self-Management Scale (four-item version). Use of Dietary Information and Problem Eating Behaviors are subscales of the Personal Diabetes Questionnaire. Metabolic equivalent (MET) minutes per week assessed with the International Physical Activity Questionnaire-Short form.

IQR interquartile range.

Of participants assigned to REACH + FAMS, 47% invited a support person and 42% had a support person enroll. Enrolled support persons were 77% female; 55% cohabitated with the participant and 23% lived out of town; 51% were a spouse/partner, 19% an adult child, 19% another relative, and 11% a friend.

Intervention Engagement and Fidelity

Median response rate to the weekly interactive text was 84% (interquartile range: 79%–97%). Ninety percent of participants randomized to REACH + FAMS completed four or more of the six core coaching sessions. Participants set a personalized goal in 96% of sessions (in the remaining 4%, goals were set from a list provided by the coach); 52% were diet goals and 48% physical activity goals. The discussion of potential or experienced family/friend involvement in the goal included helpful involvement only in 57% of the sessions, harmful involvement only in 10% of sessions, and both types of involvement in 30% of sessions; in only 3% of sessions, the participant did not want to discuss family/friend involvement despite prompting by the coach. Of the personalized skill-building exercises employed in Sessions 2–6, 28% were assertive communication, 63% were collaborative problem solving, and 9% were cognitive behavioral coping. In 99% of sessions, participants agreed to apply the skill learned during coaching with an identified friend/family member, and 98% rated confidence in applying the skill ≥7 on a scale from 1 = not at all confident to 10 = very confident. In 80% of sessions, participants reported fulfilling the prior sessions’ agreement and said doing so led to a positive interaction with a loved one (96%) and the desired change in helpful/harmful involvement (83%).

Participant Feedback on FAMS

Seventy-seven percent (34 out of 44) of participants invited to complete a postintervention interview did so. Themes identified in feedback about FAMS included (a) accountability and goal setting (n = 19), (b) a human connection with the coach (n = 11), and (c) information and advice (n = 11). Participants also reported changes resulting from their FAMS experience, including (d) changed perceptions of family/friend involvement and received support (n = 15) and (e) improved health behaviors (n = 8). These themes are described briefly below and illustrative quotes are shown in Table 2.

Table 2.

Themes and illustrative quotes about participants’ FAMS experience

Accountability and goal setting
“[I] never thought about being accountable for following up during exercise or making sure that I communicate[d] with my family about what I wanted and…be accountable for what I promised to do…it helps me to realize how it important it was to complete things…I think the coaching handled what was normally my downfalls in keeping track of stuff. The coaching made me accountable.” Non-Hispanic White Female, 58 years old
“Forming the plans. I like that. I like having a plan to accomplish something. So, you know, that was good. I even made some charts.” Non-Hispanic White male, 67 years old
“When I set goals, and I was able to stick with it, and I gained from sticking with it. I gained knowledge, and I gained the support and respect of what I’m doing. I gained all that, you know, with my family, my husband.” Non-Hispanic Black female, 51 years old
A human connection with the coach
“I thought it was very helpful to have [coaching] every month….to have a person at the end of the phone to talk to who understood…And it was like I wasn’t in this alone.” Non-Hispanic White Female, 57 years old
“I guess the most useful thing is just verbally communicating what I’ve pretty much considered to be my weak areas and you know, acknowledging it with another person.” Non-Hispanic White Male, 49 years old
“I really appreciated [coaching], and it made me feel like you guys really care about me...I feel like you guys personally are concerned about my health and how things are going with me.” Non-Hispanic Black Female, 51 years old
Information and advice
“[Coaching] gave me ideas on things that I hadn’t been doing as far as, for example, how important it was with me and my daughter communicating what we were doing and trying to keep each other on track.” Non-Hispanic Black Female, 64 years old
“The emphasis on communication with, in my case, my husband. Communicating to him, the SMART goals, and what I needed. From, you know, using ‘I’ statements and just letting him know what I’m going through with this diabetes.” Non-Hispanic White Female, 57 years old
Changed perceptions of family/friend involvement and received support
“[Coaching] made me see how important it is to have that support system. Because before it was always kind of like I’m just basically on my own…I didn’t want to feel like I was whining and complaining about having medical issues. And so I opted not to mention it to them…But I realized that the diabetes is important enough that if I need to ask them help, just talk about it.” Non-Hispanic White Female, 58 years old
“I didn’t realize how large a role [family] was really. I didn’t realize. I thought it was all on me to manage, to deal with, to cope with. I didn’t realize how much people could help me until the coaching and until it was brought up, how to get family and friends, and how to get my husband to help.” Non-Hispanic White Female, 57 years old
“I didn’t realize I could use my wife so much as a tool, that I could lean on her even more. She’s very willing to do it.” Non-Hispanic White Male, 48 years old
“I expected to find the whole experience [of coaching] kind of annoying. And I didn’t. It did help me. I think because of it my opinion of having friends and family involved in what I’m doing changed.” Non-Hispanic White Male, 59 years old
Improved health behaviors
“[Coaching] made me pay more attention to what I eat. So, I started out eating fruits a couple of times a day, and then I expanded it to vegetables at least once a day. Then I started doing it twice a day. And now I eat fruits and vegetables twice a day… it’s become a habit, and I like that because it’s something I wouldn’t [otherwise] keep doing.” Non-Hispanic Black Male, 66 years old
“[Family] play a big role in my physical activity…my niece, she loves to dance, and she will ask me to dance with her. She kind of encouraged me. So now I’m doing it for a lifestyle change.” Non-Hispanic Black Female, 51 years old
“I talked to them [friends] about [my diet], and they had suggestions on things to help me eat…I started walking. That’s where I lost the weight. We walk 15 min every morning and then 15 min in the evening.” Non-Hispanic White Female, 63 years old
“They’ve [friends] have been cheering me on to lose weight. Some of us compare our glucose every morning now. We check our sugar. We talk about our sugars and stuff.” Non-Hispanic White Female, 63 years old

Participants described a sense of accountability from setting goals, checking in monthly on goal progress, and resetting the goals with the coach. Reflecting on the benefits of coaching, participants described how the coach was someone they could communicate and connect with and the support and encouragement that coaching provided made them feel that someone cared for them and their health. Participants noting the benefit of information and advice provided specifically emphasized increased awareness about diabetes self-care, learning about food options and meal planning, learning about the role of support from family and friends, and learning about additional available resources (e.g., websites). Participants also found coaching useful for the tools and advice it offered about how to communicate with family and friends, recounting specific skill-building exercises they found most useful. Several participants who were reluctant to discuss family/friends in their self-care when they began the FAMS experience described the evolution in their thinking and their realization that family and friends were willing to provide support. Finally, participants described how they had incorporated new self-care habits into their routine—some with and some without the involvement of family/friends.

Between-group and Subgroup Effects

Only the REACH + FAMS group demonstrated improvements in family/friend involvement relative to the control group at both 3 and 6 months (Table 3). Relative to the control group, both intervention groups (REACH only and REACH + FAMS) demonstrated improvements in diabetes self-efficacy and use of dietary information for decision-making (Table 3). Neither intervention group reduced problem eating behaviors or increased physical activity relative to the control group.

Table 3.

Between-group differences

	REACH + FAMS vs. control		REACH only vs. control
	β	p	β	p
Helpful involvementa
3 months change	.088	.024	−.023	.560
6 months change	.128	.003	−.005	.897
Harmful involvementa
3 months change	−.133	.001	−.042	.279
6 months change	−.115	.012	−.048	.268
Diabetes self-efficacy
3 months change	.114	.015	.147	.002
6 months change	.101	.034	.096	.049
Use of dietary info
3 months change	.083	.045	.095	.025
6 months change	.126	.006	.107	.012
Problem eating
3 months change	−.027	.537	−.071	.113
6 months change	−.037	.410	−.028	.474
Physical activity
3 months change	.032	.518	.010	.828
6 months change	.091	.092	.014	.790

Models used imputed data (m = 20). All models adjusted for baseline values of the outcome, splined with three knots to allow for nonlinear associations. Effects meeting a priori significance threshold of p < .05 are in bold.

^aWith suppression effect (i.e., models evaluating change in helpful involvement adjusted for baseline value of harmful involvement; models evaluating change in harmful involvement adjusted for baseline value of helpful involvement).

Interaction terms examining effect modifiers were nonsignificant. Subgroup analyses did indicate some patterns (Supplementary Tables); however, because interactions terms were nonsignificant, we conservatively describe patterns in effect sizes rather than significance cutoffs for these results. Treatment effects were more pronounced among racial/ethnic minority participants as compared with non-Hispanic Whites for helpful/harmful family/friend involvement, self-efficacy, and use of dietary information for decision-making; there was some indication of treatment effects on problem eating behaviors among non-Hispanic White participants but not participants who were racial/ethnic minorities. There was no indication of differential treatment effects based on socioeconomic disadvantaged status. There were no consistent differences in treatment effects by gender; patterns suggested stronger effects on harmful family/friend involvement and use of dietary information among women but stronger effects on self-efficacy among men.

Mediation

As hypothesized, REACH + FAMS improvements in diabetes self-efficacy, use of dietary information for decision-making, and problem eating behaviors were mediated by improvements in family/friend involvement (Table 4), and there was no evidence of mediation in the REACH-only group (Table 4). Mediation analyses indicated that overall indirect (mediated) effects were consistent for REACH + FAMS participants, regardless of support person enrollment (not in Table); however, mechanisms differed based on support person enrollment status. For REACH + FAMS participants with a support person, there was significant improvement in helpful family/friend involvement (a₁ effects ranged from 0.383 to 0.408, all p < .01) but no significant reductions in harmful family/friend involvement (a₂ effects not significant) in all mediation models. For REACH + FAMS participants who did not enroll a support person, there was significant reduction in harmful family/friend involvement (a₂ effects ranged from −0.185 to −0.200, all p < .05) but no significant improvement in helpful family/friend involvement (a₁ effects not significant) in all mediation models. There was also evidence of a total effect of REACH + FAMS on physical activity (952.69, p = .044) among participants who enrolled a support person but not among those without an enrolled support person (257.56, p = .493).

Table 4.

Mediation analyses examining change in helpful and harmful family/friend involvement as mediators of change in outcomes from baseline to 6 months

REACH + FAMS vs. control	Total effect, p value	Direct effect, p value	Bootstrapped indirect effect			Proportion of total effect that is mediated
			Effect	Bias-corrected CI
Diabetes self-efficacy	.652, p = .079	.411, p =.247	0.240	0.056	0.507	30.93%
via helpful involvement			0.133	0.014	0.349	17.14%
via harmful involvement			0.107	0.001	0.300	13.79%
Use of dietary info	.456, p = .003	.357, p = .018	0.098	0.010	0.231	22.12%
via helpful involvement			0.092	0.021	0.225	20.77%
via harmful involvement			0.006	−0.025	0.061	NA
Problem eating behaviors	−.040, p = .674	.015, p = .881	−0.055	−0.122	−0.010	73.33%
via helpful involvement			−0.025	−0.082	0.004	NA
via harmful involvement			−0.030	−0.077	−0.003	40.00%
Physical activity	422.73, p = .170	410.07, p = .192	12.665	−114.500	147.184	NA
via helpful involvement			42.464	−27.067	179.099	NA
via harmful involvement			−29.800	−151.508	27.295	NA
REACH only vs. control
Diabetes self-efficacy	.816, p = .039	.744, p = .056	0.072	−0.060	0.274	NA
via helpful involvement			0.027	−0.078	0.215	NA
via harmful involvement			0.045	−0.064	0.218	NA
Use of dietary info	.331, p = .023	.315 p = .030	0.016	−0.017	0.077	NA
via helpful involvement			0.008	−0.018	0.065	NA
via harmful involvement			0.008	−0.007	0.069	NA
Problem eating behaviors	−.049, p = .640	−.029, p = .777	−0.020	−0.067	0.016	NA
via helpful involvement			−0.005	−0.044	0.010	NA
via harmful involvement			−0.015	−0.057	0.021	NA
Physical activity	−215.86, p = .429	−226.11, p = .406	10.250	−59.155	110.100	NA
via helpful involvement			10.290	−31.540	113.542	NA
via harmful involvement			−0.038	−45.926	47.217	NA

Models adjusted for baseline values of the outcome. Indirect effects bootstrapped (3,000 samples) to obtain bias-corrected confidence intervals (CI). Effects meeting a priori significance thresholds are in bold. Each outcome was examined in separate models. The total effect (c path) is the effect of the assigned condition on change in the outcome without including any mediators. The direct effect (c’ path) is the effect of the assigned condition on change in the outcome after adjustment for the mediators. Indirect effects (ab paths) represent the effect of the assigned condition on change in the outcome via its effects on changes in the mediators. Multiple mediator models calculate a total indirect effect (combined across mediators) and individual indirect effects (one for each mediator). Indirect effects consist of a paths (effect of the assigned condition on change in the mediator) and b paths (effect of change in the mediator on change in the outcome, adjusting for condition).

NA not applicable.

Discussion

Participants who received the FAMS mobile phone-delivered intervention had greater improvements in family/friend involvement—increased helpful and reduced harmful involvement—as compared to a control group, whereas the individualized intervention group (REACH only) did not. There were significant overall treatment effects on the use of dietary information for decision-making and diabetes self-efficacy for both REACH + FAMS and REACH only. However, for only the group receiving FAMS, improvements in family/friend involvement mediated improvements in diabetes self-efficacy, use of dietary information in decision-making and problem eating behaviors. This finding indicates that FAMS did in fact improve the response from family/friends as patients met new self-care goals. FAMS was well received as evidenced by (a) high fidelity to the coaching protocols, (b) high completion rates of the monthly coaching, (c) high response rates to interactive text messages, and (d) positive feedback from participants.

We did not find overall REACH + FAMS effects on physical activity nor problem eating behaviors as hypothesized, although subgroup analyses shed some light on this. Physical activity improvements were significant only among the subset of participants who enrolled with a support person. Although REACH + FAMS did show improved physical activity relative to control and REACH only, very large standard errors made it difficult to detect significant treatment effects. Physical activity is notoriously difficult to measure, and our selected measure made detecting effects challenging. Non-Hispanic Whites may have experienced improvements in problem eating behaviors, whereas racial/ethnic minorities did not. Cultural differences in the perception of eating behaviors as problematic [44] may be responsible for this differential effect but more research is needed. Except for problem eating behaviors, there was some evidence suggesting that FAMS treatment effects were more pronounced among racial/ethnic minorities. We engaged racial/ethnic minorities and patients with low SES in the development of FAMS [21] and are gratified to learn that it was efficacious in these subgroups.

Participants who chose to enroll a support person experienced a significant increase in helpful involvement (without increasing harmful involvement). In a prior mixed-methods analysis [28], we found indications that participants who invited a support person had greater need for support at baseline (i.e., reported greater depressive symptoms and more emergency department use in the prior year and cited needing more help as reason for inviting a support person). Participants who did not enroll a support person experienced a significant reduction in harmful involvement. Prior observational research suggests that either increasing helpful involvement or reducing harmful involvement can be beneficial to patients’ self-care and glycemic control [16, 22], and we found that FAMS had significant indirect effects on outcomes regardless of the pathway. Our findings suggest that participants may have exercised some wisdom about their needs and, potentially, the capabilities of their friends/family when deciding whether to enroll a support person in the study. In interviews, several participants described how they initially felt resistant to engaging family and friends but ultimately adjusted their perspectives as a result of the intervention. Collectively, these findings support assertions that requiring friend/family enrollment in studies seeking to improve support for adults may not be wise [14, 45]. Not assigning participants to enroll with a support person means that our findings may have enhanced ecological validity as estimates are affected by self-selection, which would exist in a real-world evaluation of FAMS. Estimating the effect of enrolling a support person among people who would not invite a support person is less useful than the effect among those who would.

Notably, the REACH-only treatment effects on diet and self-efficacy were similar in magnitude to those of REACH + FAMS with fewer required resources. The REACH-only group did not participate in goal setting but saw improvements consistent in magnitude to the REACH + FAMS group, suggesting that well-designed text messages alone may be enough to improve behavior. On the other hand, there are numerous examples of text messaging interventions like REACH improving short-term (~6 months) self-care behavior; the tougher problem has been sustaining those effects [46, 47]. The concept of feedback loops in family systems theory is about sustaining individual-initiated behaviors through system change [5]. This study represents an important first step toward examining the impact of FAMS on patients’ long-term outcomes by evaluating the targeted mechanisms of change. Next steps include evaluating if improving family/friend involvement in concert with behavior improvement might lead to sustained effects beyond the 6 month period observed in the current study. Additional future research on the FAMS intervention might attend to effects on patients’ perception of their friends/family as available to provide support if needed (i.e., perceived support), as we focused on received instrumental support here. Future research should also examine FAMS effects on objective measures of diet and physical activity, as well as other diabetes self-care behaviors and outcomes, such as diabetes distress and glycemic control.

The context of the larger randomized trial contributed both limitations and strengths of this work. Because FAMS was an “add-on” to REACH, we are unable to evaluate the effects of FAMS as a standalone intervention. But participants did not self-select into a “family” intervention study but rather into a study evaluating mobile phone support for diabetes self-care, thereby enhancing generalizability. Also, we were able to compare FAMS effects with those of a similar intervention that did not focus on family/friend involvement, and the context of a larger trial provided sufficient sample size to test mediation and explore effect modification. Finally, our attention to oversampling racial/ethnic minorities and patients with lower SES also enhances generalizability to patients at elevated risk for poor outcomes.

Conclusion

This study represents an advance in the science of family and social support interventions by measuring the helpful and harmful aspects of family/friend involvement and examining whether these targeted mechanisms mediated intervention effects. Previous interventions targeting family/friend involvement in adults’ diabetes self-management have either not assessed the change in targeted support-related constructs or not assessed harmful aspects of family/friend involvement [11]. We are aware of no other intervention study, across chronic disease contexts [11, 27], evaluating whether improvements in family/friend involvement or social support mediated improvements in adult patients’ outcomes. Thus, our findings lend new support for the hypothesis that adults’ diabetes self-management can be enhanced by improving family/social support. Future evaluations of FAMS hold promise for understanding if improvements in family/friend involvement in concert with patients’ self-care changes can sustain improvements in self-management.

Funding

This research was supported by the National Institute of Diabetes and Digestive and Kidney Disease via a Career Development Award (K01 DK106306 to L.S.M.) and R01 DK100694 (to L.S.M.) and the Vanderbilt Center for Diabetes Translation Research (P30 DK092986 to T.A.E.).

Compliance With Ethical Standards

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

Authors’ Contributions L.S.M. designed the FAMS intervention, planned the study design, oversaw study execution, conducted analyses and wrote the manuscript. C.A.B. consulted on the FAMS intervention design, study design, and analytic plan. R.A.G. oversaw analyses and interpretation of findings. E.M.B. assisted with interview design and conducted qualitative analyses. K.A.W. advised on measure selection and study design. K.J.H. assisted with intervention content development, conducted coaching, oversaw coaches, and managed data collection. T.A.E. advised on recruitment and retention and manuscript planning. All authors reviewed/edited the manuscript and assisted with revisions.

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 Informed consent was obtained from all individual participants included in the study.