Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2016 Apr 1.
Published in final edited form as: J Behav Med. 2014 Nov 25;38(2):363–371. doi: 10.1007/s10865-014-9611-4

Stress, depression and medication nonadherence in diabetes: test of the exacerbating and buffering effects of family support

Lindsay Satterwhite Mayberry 1,2,3, Leonard E Egede 4,5, Julie A Wagner 6, Chandra Y Osborn 1,2,3,7
PMCID: PMC4355092  NIHMSID: NIHMS644702  PMID: 25420694

Abstract

Stressors and depressive symptoms have been associated with medication nonadherence among adults with type 2 diabetes (T2DM). We tested whether these associations were exacerbated by obstructive family behaviors or buffered by supportive family behaviors in a sample of 192 adults with T2DM and low socioeconomic status using unadjusted and adjusted regression models. We found support for the exacerbating hypothesis. Stressors and nonadherence were only associated at higher levels of obstructive family behaviors (adjusted interaction OR=1.12, p=.002). Similarly, depressive symptoms and nonadherence were only associated at higher levels of obstructive family behaviors (adjusted interaction OR=3.31, p=.002). When participants reported few obstructive family behaviors, neither stressors nor depressive symptoms were associated with nonadherence. We did not find support for the buffering hypothesis; stressors and depressive symptoms were associated with nonadherence regardless of supportive family behaviors. Nonadherent patients experiencing stressors and/or major depressive symptoms may benefit from interventions that reduce obstructive family behaviors.

Keywords: Stress/stressors, depression, family support, social support, medication adherence, diabetes

Introduction

Approximately one-third of adults with type 2 diabetes mellitus (T2DM) are nonadherent to their diabetes medications (Cramer et al., 2008; Ho et al., 2006). Although optimal performance of all self-care behaviors is important, medication nonadherence is both common (Cramer et al., 2008; Ho et al., 2006; Kim et al., 2010) and by itself a strong, independent predictor of poor glycemic control (Egede et al., 2011; Heisler et al., 2007; Kim et al., 2010; Ngo-Metzger et al., 2012), hospitalizations (Ho et al., 2006; Hong & Kang, 2011), premature death (Egede et al., 2013; Ho et al., 2006; Hong & Kang, 2011), and higher healthcare costs (Egede et al., 2012; Salas et al., 2009). Racial/ethnic minorities are less adherent to diabetes medications than non-Hispanic Whites (Egede et al., 2011; Heisler et al., 2007; Kim et al., 2010; Osborn et al., 2011), and disparities in diabetes medication adherence have helped to explain disparities in glycemic control (Egede et al., 2011; Heisler et al., 2007).

Stressors and depressive symptoms are related (Fisher et al., 2001; Naranjo et al., 2011) but independent (Osborn et al., 2014) predictors of medication nonadherence (Billimek & Sorkin, 2012a, 2012b; DiMatteo et al., 2000; Gonzalez et al., 2007; Ngo-Metzger et al., 2012; Osborn et al., 2014) which disproportionately affect minorities and socioeconomically disadvantaged patients (Fisher et al., 2001; Jeon et al., 2009; Jiang et al., 2008; Mendenhall et al., 2012; Spencer et al., 2006). However, not all patients who experience these overlapping vulnerabilities are suboptimally adherent. Whether experiencing more stressors or depressive symptoms leads to less diabetes medication adherence may depend on the quantity and quality of a patient’s social support. The exacerbating model of social support (Rook, 1998) posits the deleterious effects of stressors on outcomes will be stronger among persons experiencing more harmful behaviors from their social network than among persons experiencing little or none of these harmful behaviors. To our knowledge, no studies have examined the exacerbating hypothesis for any outcome in T2DM. The buffering model of social support (Cohen & Wills, 1985) posits that stressors have deleterious effects on outcomes among persons with little or no support, but no effect or a lesser effect among persons with high support. Several studies have examined whether support buffers adults with T2DM against the development of psychological distress or depression (Baek et al., 2014; Beekman et al., 1997; Littlefield et al., 1990; Thomas et al., 2007), yet few have examined buffering effects on diabetes management (Glasgow & Toobert, 1988; Griffith et al., 1990). Egede and Osborn (Egede & Osborn, 2010; Osborn & Egede, 2012) previously reported the absence of social support partially explains the association between depressive symptoms and diabetes medication nonadherence (i.e., mediation analysis). Moderation analysis, as proposed by the exacerbating and buffering models, suggests that the relationship between the predictor and the outcome vary based on the value of the moderator. To our knowledge, no studies have explored the buffering model with regards to diabetes medication adherence.

Although general social support is associated with more adherence to medications (DiMatteo, 2004), family members may be particularly well-suited to assist patients with diabetes medication adherence by reminding the patient to take medications as scheduled throughout their daily routine, ordering/picking up prescription refills, or helping the patient problem-solve when a side effect or hypoglycemic event occurs (Mayberry & Osborn, 2012). Such diabetes-specific supportive family behaviors are associated with more adherence to diabetes medications among adults with T2DM (Mayberry & Osborn, 2014). However, family interactions with the diabetic patient may not always be positive. Family members may also perform harmful behaviors (Carter-Edwards et al., 2004; Glasgow & Toobert, 1988; Henry et al., 2013; Mayberry & Osborn, 2012; Rosland et al., 2010), which are associated with patients being less adherent to diabetes medications (Mayberry & Osborn, 2012, 2014). These harmful or “obstructive” family behaviors include sabotaging behaviors (e.g., questioning the need for prescribed medications) (Carter-Edwards et al., 2004) and nagging or arguing about nonadherence (Carter-Edwards et al., 2004; Henry et al., 2013; Mayberry & Osborn, 2012; Rosland et al., 2010; Stephens et al., 2013).

Objective

We tested the moderation hypotheses that the associations between stressors/depressive symptoms and nonadherence would be stronger in the context of more obstructive family behaviors (exacerbating hypothesis) and weaker in the context of more supportive family behaviors (buffering hypothesis).

Methods

Study Design

We used data collected as part of a larger cross-sectional study of modifiable determinants of medication adherence (Mayberry et al., 2013). Eligible participants where English- or Spanish-speaking adults (≥18 years old) diagnosed with T2DM and self-administering prescribed medications (oral agents and/or insulin) who did not have a sensory/cognitive impairment that prohibited participation (Mayberry et al., 2013). The larger study enrolled 314 participants consecutively as they arrived for a clinic appointment at a Federally Qualified Health Center in Nashville, TN over a 2.5 year period. Of the 588 patients with T2DM who arrived for a clinic appointment during the enrollment period, 377 were eligible and 314 enrolled (83.3% of those eligible) (Mayberry et al., 2013). The participants who enrolled in the final study year (n=192) completed measures assessing family behaviors and stressors as part of the regular study protocol and were included in these analyses. Eligible and interested participants were taken to a private clinic room before and/or after their appointment to complete informed consent and an interviewer-administered survey. The interview protocol was translated to Spanish using the forward-backward technique (Behling & Law, 2000) by licensed translators. Participants were compensated $20. The Vanderbilt University Institutional Review Board approved all study procedures prior to enrollment.

Measures

Predictors

Stressors were assessed with the Tool for Assessing Patients’ Stressors (TAPS) (Rothberg et al., 2011; Welch et al., 2011). Each of the 20 TAPS items assesses a stressor commonly reported by socioeconomically disadvantaged patient populations. Participants are asked “In the past year (12 months), have any of the following family issues been stressful for you?” with 1=yes and 0=no as response options. Example items include “lack of affordable local transport for my family or myself” and “living in an unsafe neighborhood.” The TAPS score is calculated by summing the “yes” responses across items for a possible range of 0–20 (Osborn et al., 2014). Although the checklist nature of the TAPS precludes examination of internal consistency reliability, the stressors most commonly reported on the TAPS have been similar across samples of patients from FQHCs (Rothberg et al., 2011; Welch et al., 2011). This instrument was developed based on extensive input from healthcare providers and diabetes educators who treat patients at FQHCs (Osborn et al., 2014), enhancing its content and face validity.

Depressive symptoms were assessed with the validated Patient Health Questionnaire-9 (PHQ-9) (Kroenke et al., 2001) which assesses the frequency of depressive symptoms according to DSM-IV criteria on a scale from 0=not at all to 3=nearly every day. The PHQ-9 yields a continuous severity score from 0 to 27. In our sample, the PHQ-9 had internal consistency reliability of α=0.87. PHQ-9 scores ≥10 coincide with a depression diagnosis from structured clinical interviews with 88% sensitivity and 88% specificity in general patient populations (Kroenke et al., 2001). Following the PHQ-9 developers’ recommendations, we categorized participants’ depressive symptoms as <5 none, 5–9 mild, ≥10 moderate to severe symptoms (Kroenke et al., 2001). Although 10 is the optimal threshold above which symptoms likely reflect depression in general populations (Kroenke et al., 2001), several studies (Anderson et al., 2001; Twist et al., 2013; van Steenbergen-Weijenburg et al., 2010) report the optimal cutoff point for correspondence with depression diagnosis from a clinical interview was higher among adults with diabetes. Twist et al. (2013) identified 12 as the optimal cutoff among adults with T2DM, reporting sensitivity of 86.9% and specificity of 80.3%. Therefore, we also categorized participants’ depressive symptoms as <5 none, 5–11 mild, ≥12 moderate to severe symptoms for sensitivity analyses.

Moderators

Supportive and obstructive family behaviors were assessed with the Diabetes Family Behavior Checklist-II (DFBC-II) (Glasgow & Toobert, 1988; Schafer et al., 1986). The 16-item DFBC-II asks respondents how often their family members have performed diabetes-specific behaviors in the last month on a scale from 1=never to 5=at least once a day. Supportive behaviors include “exercise with you” or “eat at the same time that you do,” whereas obstructive behaviors include “criticize you for not exercising regularly” or “eat foods that are not part of your diabetic diet.” We averaged the 9 supportive items and 7 nonsupportive items to create two subscales ranging from 1–5, with higher scores indicating more supportive or obstructive behaviors, respectively (Glasgow & Toobert, 1988). The DFBC-II also has a 2-item medication-specific subscale, which is calculated by subtracting the obstructive item score from the supportive item score for a possible range of −4 to 4 (Glasgow & Toobert, 1988; Schafer et al., 1986). In this sample, internal reliability (Cronbach’s α) was 0.85 for the supportive subscale and 0.78 for the obstructive subscale. Among adults with type 1 diabetes mellitus, Schafer et al. (1986) reported test-retest reliability ratings over a six-month period of 0.84 and 0.69 for the supportive and obstructive subscales, respectively, and correlations with family-member report ranged from 0.27 (p<.10) to 0.68 (p<.001), respectively (Schafer et al., 1986).

Outcome

Diabetes medication nonadherence was assessed with the Adherence to Refills and Medications for Diabetes (ARMS-D) (Mayberry et al., 2013). Items assess respondents’ adherence to taking and refilling diabetes medications (e.g., “How often do you…forget to take your diabetes medicine(s)?“ or “…run out of your diabetes medicine(s)?“). Response options range from 1=none of the time to 4=all of the time and are summed to produce a score ranging from 11 to 44 with higher scores indicating more nonadherence. The ARMS was originally designed to assess adherence to all medications and was validated against objective measures of refill adherence (r=0.32, p<.01 with retrospective six-month refill adherence) in a relatively similar sample to ours (i.e., predominantly African American, inner-city sample with a chronic disease) (Kripalani et al., 2009). The diabetes-specific ARMS-D correlates well with commonly used measures of diabetes medication adherence (all correlations >0.45, p<.001) and predicts glycemic control (0.25, p<.001) (Mayberry et al., 2013). In this sample, internal consistency reliability was α=0.85.

Covariates

We collected self-reported age, gender, race, ethnicity, income, education, insurance status, and diabetes duration (time since diabetes diagnosis in years and months). The number and type of diabetes medication(s) were abstracted from participants’ medical records.

Analyses

We used Stata 12 for analyses. First, we used Spearman’s rank correlation coefficients (ρ) to examine bivariate associations between variables of interest. Next, we conducted a series of regression models with interaction terms using proportional odds/ordinal logistic regression. This approach treats the outcome as a series of ordinal variables (Harrell, 2001) and provides freedom from distributional assumptions for handling outcome data with floor or ceiling effects (e.g., 23% of our participants reported perfect adherence on the ARMS-D) (Harrell, 2001). We used an approximate likelihood-ratio test to confirm our data satisfied the proportional odds assumption that each probability curve has the same coefficient (Wolfe & Gould, 1998).

We conducted separate regression models with an interaction term between each predictor (i.e., stressors and depressive symptoms) and each type of family behavior (i.e., supportive, obstructive, and medication-specific support). We examined depressive symptoms continuously, to maximize our power to detect interactions, and categorically, to determine if interactions were consistent across different levels of depressive symptoms. We present results using the most common cutoff for PHQ-9 scores (≥10), but conducted sensitivity analyses with the higher cutoff suggested for use among adults with T2DM (≥12). Continuous variables were mean-centered to create interaction terms. For each interaction term with the supportive or obstructive subscales, we conducted a partially adjusted model and a fully adjusted model. To accommodate suppressor effects (Conger, 1974; Velicer, 1978) between supportive and obstructive family behaviors identified previously (Mayberry & Osborn, 2014), each partially adjusted model included the main effects, interaction term, and both types of family behaviors (e.g., obstructive family behaviors is a covariate in the model assessing the stress X supportive family behavior interaction). For each interaction term with the medication-specific subscale, we conducted an unadjusted model and a fully adjusted model. Each fully adjusted model included a priori covariates (i.e., age, gender, race, ethnicity, income, education, insurance status, diabetes duration, and insulin status). We then further adjusted for depressive symptoms in the models assessing interactions with stressors to ensure identified effects could be attributed to stressors and not to the association between stressors and depressive symptoms. Significant interaction effects are graphically depicted using post-hoc probing methods (Holmbeck, 2002). Results are presented as odds ratios (OR) or adjusted odds ratios (AOR) which are nonlinear, and graphed as log-odds, which are linear.

Results

The sample was racially/ethnically diverse (56% African American and 10% Hispanic) and socioeconomically disadvantaged (76% had incomes less than $20,000; 47% uninsured; 30% had less than a high school degree; see Table 1). Obstructive family behaviors were associated with more stressors (ρ=.20, p=.005) and more depressive symptoms (ρ=.18, p=.01). Supportive family behaviors, and, separately, the medication-specific subscale, were not associated with stressors or depressive symptoms. Both stressors (ρ=.29, p<.001) and depressive symptoms (ρ=.23, p=.001) were associated with nonadherence.

Table 1.

Participant characteristics.

N = 192 Mean ± SD or n (%)
COVARIATES
Age, years 51.6 ± 10.9
Gender
 Male 57 (29.7)
 Female 135 (70.3)
Race
 White 65 (33.9)
 African American/Black 107 (55.7)
 Other 20 (10.4)
Hispanic ethnicity 19 (10.0)
Education, years 12.0 ± 3.0
Income
 <$10,000 78 (43.6)
 $10,000 – $14,999 49 (27.4)
 $15,000 – $19,999 27 (15.0)
 ≥$20,000 25 (14.0)
Insurance status
 Uninsured 90 (46.9)
 Public Insurance 87 (45.3)
 Private Insurance 15 (7.8)
Diabetes duration, years 7.7 ± 7.2
Prescribed insulin 90 (46.9)
PREDICTORS
Stressors (TAPS) 4.8 ± 3.9
Depressive Symptoms (PHQ-9) 8.4 ± 6.6
 <5 (None) 67 (34.9)
 5–9 (Mild) 62 (32.3)
 ≥10 (Moderate to Severe)a 63 (32.8)
MODERATORS
Supportive Family Behaviors (DFBC-II) 2.4 ± 1.0
Medication-Specific Subscale (DFBC-II) 1.2 ± 1.6
Obstructive Family Behaviors (DFBC-II) 2.1 ± 0.9
OUTCOME
Medication Adherence (ARMS-D) 15.6 ± 4.9
Open in a new tab

ARMS-D = Adherence to Refills and Medications Scale for Diabetes; DFBC-II = Diabetes Family Behavior Checklist-II; PHQ-9 = Personal Health Questionnaire-9; SD = standard deviation; TAPS = Tool for Assessing Patients’ Stressors

a

≥12 (Moderate to Severe) n=46 (24.0%)

We found consistent support for the exacerbating hypothesis. The associations between stressors and nonadherence and depressive symptoms (continuous) and nonadherence were stronger as obstructive family behaviors increased (Table 2). The interaction with stressors remained significant after further adjusting for depressive symptoms (AOR=1.12, p=.002, not in Table). Post-hoc probing revealed a threshold of obstructive family behaviors over which stressors were associated with nonadherence. As shown in Figure 1, there was a significant association between experiencing more stressors and nonadherence among participants experiencing the sample mean or more of obstructive family behaviors (at +1 and +2 standard deviations), but not among participants experiencing less than the sample mean of obstructive family behaviors (at −1 and −2 standard deviations). Using the categorical measure of depressive symptoms, we found there was only an exacerbating effect among those with moderate to severe (but not among those with none or mild) depressive symptoms (Table 2). Therefore, we compared those with moderate to severe symptoms to all other participants for post-hoc probing. As shown in Figure 2, a similar threshold effect appears in this pattern of results; having moderate to severe depressive symptoms was only associated with nonadherence among participants experiencing the sample mean or more of obstructive family behaviors (at +1 and +2 standard deviations), but not among those experiencing less than the sample mean of obstructive family behaviors (at −1 and −2 standard deviations). Findings were consistent when using the higher cutoff of 12 to indicate moderate to severe depressive symptoms.

Table 2.

Proportional odds/ordinal logistic regression analyses testing the exacerbating hypothesis: Odds ratios for moderation effects of obstructive family behaviors on the associations between stressors and diabetes medication nonadherence, and between depressive symptoms and diabetes medication nonadherence.

Outcome: Nonadherence Stressors Depressive Symptoms (PHQ-9)
TAPS X Continuous X Milda X Moderate-Severea X
Obstructive Family Behaviors Obstructive Family Behaviors Obstructive Family Behaviors Obstructive Family Behaviors
AOR p value AOR p value AOR p value AOR p value
Partially adjusted 1.10 .007 1.03 .08 1.42 .35 2.43 .02
Fully adjusted 1.12 .002 1.05 .02 1.56 .24 3.31 .002
Open in a new tab
a

Categorical PHQ-9: reference = none (PHQ-9 score<5)

Partially adjusted models include main effects, interaction terms, and supportive family behaviors. Fully adjusted models are further adjusted for age, gender, race, ethnicity, education, income, insurance status, diabetes duration, and insulin status (all as depicted in Table 1). AOR = adjusted odds ratio; PHQ-9 = Personal Health Questionnaire-9; TAPS = Tool for Assessing Patients’ Stressors. Note: Results were consistent when using the cutoff of PHQ-9≥12 to categorize participants with moderate-severe depressive symptoms: partially adjusted OR = 3.12, p=.004, fully adjusted OR = 4.35, p<.001.

Figure 1. Adjusted effects of having high stressors on nonadherence to diabetes medications at different levels of obstructive family behaviors (the moderator).

Figure 1

Open in a new tab

resents the simple slopes and significance values for the effect of having + 1 SD on the TAPS (versus -1 SD on the TAPS) on the log-odds of nonadherence at different values of obstructive family behaviors relative to the sample mean. The values on the y axis have no substantive meaning in the interpretation of log-odds; the simple slopes depict changes in the log-odds of more nonadherence and are interpreted relative to zero (i.e., no association) and to each another. Adjusted for supportive family behaviors and covariates (age, race, ethnicity, education, insurance status, diabetes duration, and insulin status). SD=standard deviation; TAPS=Tool Assessing Patients’ Stressors.

Figure 2. Adjusted effects of having a PHQ-9 score ≥10 on nonadherence to diabetes medications at different levels of obstructive family behaviors (the moderator).

Figure 2

Open in a new tab

Presents the simple slopes and significance values for the effect of having a PHQ-9≥10 (versus PHQ-9<10) on the log-odds of nonadherence at different values of obstructive family behaviors relative to the sample mean. The values on the y axis have no substantive meaning in the interpretation of log-odds; the simple slopes depict changes in the log-odds of more nonadherence and are interpreted relative to zero (i.e., no association) and to each another. Adjusted for supportive family behaviors and covariates (age, race, ethnicity, education, insurance status, diabetes duration, and insulin status). PHQ-9 = Personal Health Questionnaire-9; SD=standard deviation.

We did not find support for the buffering hypothesis. Results from fully adjusted analyses are presented. Neither supportive family behaviors (interaction AOR=1.06, p=.07) nor medication-specific support (interaction AOR=1.01, p=.76) moderated the association between stressors and nonadherence. Neither supportive family behaviors (interaction AOR=1.02, p=.22) nor medication-specific support (interaction AOR=1.02, p=.16) moderated the association between depressive symptoms (continuous) and nonadherence. However, among participants with major depressive symptoms there was a significant interaction in the opposite-than-hypothesized direction for both supportive family behaviors (interaction AOR=2.07, p=.03) and medication-specific support (interaction AOR=1.59, p=.04). These AORs were >1.0, indicating that having moderate to severe depressive symptoms was more strongly associated with nonadherence in the context of more support.

Discussion

In a racially diverse and socioeconomically disadvantaged sample of adults with T2DM, we found support for the hypothesis that obstructive family behaviors exacerbated the associations between both stressors and depressive symptoms and diabetes medication nonadherence. Neither stressors nor depressive symptoms were associated with nonadherence for patients with relatively low obstructive family behaviors, but both were increasingly associated with nonadherence at higher rates of obstructive family behaviors. Furthermore, the exacerbating effect for stressors was maintained after adjusting for depressive symptoms, suggesting a unique effect specific to stressors. Substantively, these results suggest patients may be able to maintain optimal medication adherence in the face of stressors and/or depressive symptoms as long as their family members do not sabotage, nag, or argue about self-care, although reverse causality is also plausible. For example, family members may be more inclined to sabotage, nag, or argue about self-care with family members who are nonadherent. To our knowledge, this study is the first to test the exacerbating model of social support (Rook, 1998) for any outcome in diabetes or medication nonadherence in any disease context.

We did not find support for the hypothesis that supportive family behaviors or medication-specific family support buffered associations between stressors or depressive symptoms and diabetes medication nonadherence (Cohen & Wills, 1985). On the contrary, we found evidence that the association between having moderate to severe depressive symptoms and nonadherence was stronger in the context of more supportive behaviors. This unexpected finding cannot be attributed to the positive association between supportive and obstructive family behaviors (Mayberry & Osborn, 2012, 2014) because analyses adjusted for obstructive family behaviors. Similar unexpected results have been reported by studies exploring social support as a moderator of the associations between health literacy and health status (Lee et al., 2009) between health literacy and medication adherence (Johnson et al., 2010) in general patient populations. We hypothesize that our unexpected finding may be attributable to (a) our cross-sectional design and/or (b) our use of measures of received support, instead of perceived support. Family members may become more supportive of patients’ self-care once they become aware the patient is experiencing elevated depressive symptoms and/or is nonadherent. Such reverse-causality cannot be examined with our cross-sectional design. In general, consistent evidence for the buffering hypothesis has been documented by studies using measures of satisfaction with support (i.e., perceived support), rather than measures of actual support behaviors (i.e., received support) (Cohen & Wills, 1985; Thoits, 2011), such as the DFBC-II supportive subscale. Received support, but not perceived support, may be confounded with need (Thoits, 2011). In diabetes, Griffith et al. (Griffith et al., 1990) found evidence supporting the buffering hypothesis when examining the association between stress and glycemic control using a measure of perceived support, and studies (Egede & Osborn, 2010; Osborn & Egede, 2012) identifying social support as a mechanism explaining the association between depressive symptoms and nonadherence also measured perceived support.

In addition to using cross-sectional data and measuring received support, there are other study limitations to note. We sampled from a single clinic and results may not generalize to other patient populations. Although we had a relatively small sample size, our sample was larger than most studies exploring the buffering hypothesis in diabetes (Baek et al., 2014; Glasgow & Toobert, 1988; Griffith et al., 1990; Littlefield et al., 1990). Our lack of support for the buffering hypothesis cannot be attributed to reduced power to detect interactions because nonsignificant interactions were in the opposite-than-hypothesized direction, suggesting we would find further evidence against the buffering hypothesis with a larger sample. We also used a single self-reported measure of medication adherence. There is no perfect method to assess medication adherence (Farmer, 1999; Hansen et al., 2009). Self-report is the most practical, but is subject to recall and reporting bias, and may overestimate adherence (Liu et al., 2001), whereas prescription refills assess medication quantity, but not dosing (Choo et al., 1999; Steiner & Prochazka, 1997). Electronic devices attempt to capture dosing, and are often considered the gold standard for measuring adherence, but may not precisely capture when or how much medication was ingested, are often impractical and expensive (Choo et al., 1999; Farmer, 1999), and may underestimate adherence (Liu et al., 2001). Self-report measures of adherence in T2DM have been demonstrated to be sufficiently accurate and comparable to other types of measures (Gonzalez et al., 2013; Krapek et al., 2004). Nonetheless, our use of a single measure presents limitations, as does our use of self-report measures of depressive symptoms (instead of a clinical interview) and family behaviors. Because we are the first to test or find support for the exacerbating hypothesis, future studies should examine this hypothesis with multiple measures of medication adherence (Liu et al., 2001).

Conclusion

The extant evidence suggests support buffers the individual against developing depressive symptoms or experiencing distress in the context of chronic disease (Baek et al., 2014; Beekman et al., 1997; Littlefield et al., 1990; Thomas et al., 2007) but, per our findings, support might not buffer the detrimental effects of stressors or depressive symptoms on adherence. Additional work is needed to understand these relationships, longitudinally, and in other patient populations. It may be that adult patients who experience numerous stressors and/or increased depressive symptoms are able to maintain adherence in the absence of the additional strains presented by family members who are sabotaging or nagging/arguing with them. Alternatively, family members may resort to nagging and arguing out of frustration after attempts to support the patient’s adherence are ineffective, especially if the patient is managing depressive symptoms alongside their diabetes. Furthermore, family members of patients who experience numerous life stressors may also be coping with numerous stressors and may not have the energy or psychological resources to identify appropriate supportive behaviors and thus resort to nagging or arguing about adherence. Regardless, our findings suggest patients experiencing numerous life stressors and/or depressive symptoms may benefit from developing skills to talk with their family members about replacing nagging/arguing behaviors with welcomed supportive behaviors (e.g., ordering/picking up prescription refills, carrying extra medication doses for the patient), and how to tell family members when their support attempts are not perceived as helpful. Interventions seeking to improve adherence among patients who experience numerous life stressors and/or major depressive symptoms should seek to reduce family members’ diabetes-specific sabotaging and nagging/arguing behaviors.

Acknowledgments

Sources of Support: This research study was funded with support the National Center for Research Resources, Grant UL1 RR024975-01, which is now at the National Center for Advancing Translational Sciences, Grant 2 UL1 TR000445-06. This study was supported in part by grant P30DK092986 (PI: Elasy). C.Y.O. was supported by an NIH/NIDDK Career Development Award (K01DK087894). L.S.M. was supported by an NIH/NIDDK National Research Service Award (F32DK097880). J.A.W. was supported by the NIH/NIMHD (R01 MD005879), the NIH/NIDDK (DP3 DK097705), the American Diabetes Association (7-13-TS-31), and the Chicago Center for Diabetes Translation Research. L.E.E. was supported by NIH/NIDDK (K24DK093699) and NIH/NIDDK (R01DK098529).

The authors would like to acknowledge Cecilia C. Quintero, Sahbina Ebba, Karen Calderon, Leo Cortes, Anne Crook, Carmen Mekhail, the Vine Hill Community Clinic personnel, and the participants for their contributions to this research.

Footnotes

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Conflict of Interests: Lindsay Satterwhite Mayberry declares that she has no conflict of interest. Leonard E. Egede declares that he has no conflict of interest. Julie A. Wagner declares that she has no conflict of interest. Chandra Y. Osborn declares that she has no conflict of interest. All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2000(5). Informed consent was obtained from all patients for being included in the study.

References

  1. Anderson RJ, Freedland KE, Clouse RE, Lustman PJ. The prevalence of comorbid depression in adults with diabetes: a meta-analysis. Diabetes Care. 2001;24(6):1069–1078. doi: 10.2337/diacare.24.6.1069. [DOI] [PubMed] [Google Scholar]
  2. Baek RN, Tanenbaum ML, Gonzalez JS. Diabetes burden and diabetes distress: the buffering effect of social support. Ann Behav Med. 2014 doi: 10.1007/s12160-013-9585-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Beekman AT, Penninx BW, Deeg DJ, Ormel J, Braam AW, van Tilburg W. Depression and physical health in later life: results from the Longitudinal Aging Study Amsterdam (LASA) J Affect Disord. 1997;46(3):219–231. doi: 10.1016/s0165-0327(97)00145-6. [DOI] [PubMed] [Google Scholar]
  4. Behling O, Law KS, editors. Translating questionnaires and other research instruments: Problems and solutions. Vol. 133. Thousand Oaks: Sage; 2000. [Google Scholar]
  5. Billimek J, Sorkin DH. Food insecurity, processes of care, and self-reported medication underuse in patients with type 2 diabetes: results from the California Health Interview Survey. Health Serv Res. 2012a;47(6):2159–2168. doi: 10.1111/j.1475-6773.2012.01463.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Billimek J, Sorkin DH. Self-reported neighborhood safety and nonadherence to treatment regimens among patients with type 2 diabetes. J Gen Intern Med. 2012b;27(3):292–296. doi: 10.1007/s11606-011-1882-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Carter-Edwards L, Skelly AH, Cagle CS, Appel SJ. They care but don’t understand”’: Family support of African American women with type 2 diabetes. Diabetes Educator. 2004;30(3):493–501. doi: 10.1177/014572170403000321. [DOI] [PubMed] [Google Scholar]
  8. Choo PW, Rand CS, Inui TS, Lee ML, Cain E, Cordeiro-Breault M, Platt R. Validation of patient reports, automated pharmacy records, and pill counts with electronic monitoring of adherence to antihypertensive therapy. Med Care. 1999;37(9):846–857. doi: 10.1097/00005650-199909000-00002. [DOI] [PubMed] [Google Scholar]
  9. Cohen S, Wills TA. Stress, social support, and the buffering hypothesis. Psychol Bull. 1985;98(2):310–357. [PubMed] [Google Scholar]
  10. Conger AJ. A revised definition for suppressor variables: a guide to their identification and interpretation. Educ Psychol Meas. 1974;34(1):35–46. [Google Scholar]
  11. Cramer JA, Benedict A, Muszbek N, Keskinaslan A, Khan ZM. The significance of compliance and persistence in the treatment of diabetes, hypertension and dyslipidaemia: a review. Int J Clin Pract. 2008;62(1):76–87. doi: 10.1111/j.1742-1241.2007.01630.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. DiMatteo MR. Social support and patient adherence to medical treatment: a meta-analysis. Health Psychol. 2004;23(2):207–218. doi: 10.1037/0278-6133.23.2.207. [DOI] [PubMed] [Google Scholar]
  13. DiMatteo MR, Lepper HS, Croghan TW. Depression is a risk factor for noncompliance with medical treatment: meta-analysis of the effects of anxiety and depression on patient adherence. Arch Intern Med. 2000;160(14):2101–2107. doi: 10.1001/archinte.160.14.2101. [DOI] [PubMed] [Google Scholar]
  14. Egede LE, Gebregziabher M, Dismuke CE, Lynch CP, Axon RN, Zhao Y, Mauldin PD. Medication nonadherence in diabetes: longitudinal effects on costs and potential cost savings from improvement. Diabetes Care. 2012;35(12):2533–2539. doi: 10.2337/dc12-0572. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Egede LE, Gebregziabher M, Hunt KJ, Axon RN, Echols C, Gilbert GE, Mauldin PD. Regional, geographic, and racial/ethnic variation in glycemic control in a national sample of veterans with diabetes. Diabetes Care. 2011;34(4):938–943. doi: 10.2337/dc10-1504. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Egede LE, Lynch CP, Gebregziabher M, Hunt KJ, Echols C, Gilbert GE, Mauldin PD. Differential impact of longitudinal medication non-adherence on mortality by race/ethnicity among veterans with diabetes. J Gen Intern Med. 2013;28(2):208–215. doi: 10.1007/s11606-012-2200-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Egede LE, Osborn CY. Role of motivation in the relationship between depression, self-care, and glycemic control in adults with type 2 diabetes. Diabetes Educator. 2010;36(2):276–283. doi: 10.1177/0145721710361389. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Farmer KC. Methods for measuring and monitoring medication regimen adherence in clinical trials and clinical practice. Clin Ther. 1999;21(6):1074–1090. doi: 10.1016/S0149-2918(99)80026-5. discussion 1073. [DOI] [PubMed] [Google Scholar]
  19. Fisher L, Chesla CA, Mullan JT, Skaff MM, Kanter RA. Contributors to depression in Latino and European-American patients with type 2 diabetes. Diabetes Care. 2001;24(10):1751–1757. doi: 10.2337/diacare.24.10.1751. [DOI] [PubMed] [Google Scholar]
  20. Glasgow RE, Toobert DJ. Social environment and regimen adherence among type II diabetic patients. Diabetes Care. 1988;11(5):377–386. doi: 10.2337/diacare.11.5.377. [DOI] [PubMed] [Google Scholar]
  21. Gonzalez JS, Safren SA, Cagliero E, Wexler DJ, Delahanty L, Wittenberg E, Grant RW. Depression, Self-Care, and Medication Adherence in Type 2 Diabetes Relationships across the full range of symptom severity. Diabetes Care. 2007;30(9):2222–2227. doi: 10.2337/dc07-0158. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Gonzalez JS, Schneider HE, Wexler DJ, Psaros C, Delahanty LM, Cagliero E, Safren SA. Validity of medication adherence self-reports in adults with type 2 diabetes. Diabetes Care. 2013;36(4):831–837. doi: 10.2337/dc12-0410. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Griffith LS, Field BJ, Lustman PJ. Life stress and social support in diabetes: association with glycemic control. Int J Psychiatry Med. 1990;20(4):365–372. doi: 10.2190/APH4-YMBG-NVRL-VLWD. [DOI] [PubMed] [Google Scholar]
  24. Hansen RA, Kim MM, Song L, Tu W, Wu J, Murray MD. Comparison of methods to assess medication adherence and classify nonadherence. Ann Pharmacother. 2009;43(3):413–422. doi: 10.1345/aph.1L496. [DOI] [PubMed] [Google Scholar]
  25. Harrell FE. Regression modeling strategies: with applications to linear models, logistic regression, and survival analysis. Springer; 2001. [Google Scholar]
  26. Heisler M, Faul JD, Hayward RA, Langa KM, Blaum C, Weir D. Mechanisms for racial and ethnic disparities in glycemic control in middle-aged and older Americans in the health and retirement study. Arch Intern Med. 2007;167(17):1853–1860. doi: 10.1001/archinte.167.17.1853. [DOI] [PubMed] [Google Scholar]
  27. Henry SL, Rook KS, Stephens MA, Franks MM. Spousal undermining of older diabetic patients’ disease management. J Health Psychol. 2013;18(12):1550–61. doi: 10.1177/1359105312465913. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Ho PM, Rumsfeld JS, Masoudi FA, McClure DL, Plomondon ME, Steiner JF, Magid DJ. Effect of medication nonadherence on hospitalization and mortality among patients with diabetes mellitus. Arch Intern Med. 2006;166(17):1836–1841. doi: 10.1001/archinte.166.17.1836. [DOI] [PubMed] [Google Scholar]
  29. Holmbeck GN. Post-hoc probing of significant moderational and mediational effects in studies of pediatric populations. J Ped Psychol. 2002;27(1):87–96. doi: 10.1093/jpepsy/27.1.87. [DOI] [PubMed] [Google Scholar]
  30. Hong JS, Kang HC. Relationship between oral antihyperglycemic medication adherence and hospitalization, mortality, and healthcare costs in adult ambulatory care patients with type 2 diabetes in South Korea. Med Care. 2011;49(4):378–384. doi: 10.1097/MLR.0b013e31820292d1. [DOI] [PubMed] [Google Scholar]
  31. Jeon YH, Essue B, Jan S, Wells R, Whitworth JA. Economic hardship associated with managing chronic illness: a qualitative inquiry. BMC Health Serv Res. 2009;9:182. doi: 10.1186/1472-6963-9-182. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Jiang L, Beals J, Whitesell NR, Roubideaux Y, Manson SM. Stress burden and diabetes in two American Indian reservation communities. Diabetes Care. 2008;31(3):427–429. doi: 10.2337/dc07-2044. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Johnson VR, Jacobson KL, Gazmararian JA, Blake SC. Does social support help limited-literacy patients with medication adherence? A mixed methods study of patients in the Pharmacy Intervention for Limited Literacy (PILL) Study. Patient Educ Couns. 2010;79(1):14–24. doi: 10.1016/j.pec.2009.07.002. [DOI] [PubMed] [Google Scholar]
  34. Kim N, Agostini JV, Justice AC. Refill adherence to oral hypoglycemic agents and glycemic control in veterans. Ann Pharmacother. 2010;44(5):800–808. doi: 10.1345/aph.1M570. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Krapek K, King K, Warren SS, George KG, Caputo DA, Mihelich K, Lubowski TJ. Medication adherence and associated hemoglobin A1c in type 2 diabetes. Ann Pharmacother. 2004;38(9):1357–1362. doi: 10.1345/aph.1D612. [DOI] [PubMed] [Google Scholar]
  36. Kripalani S, Risser J, Gatti ME, Jacobson TA. Development and evaluation of the Adherence to Refills and Medications Scale (ARMS) among low-literacy patients with chronic disease. Value Health. 2009;12(1):118–123. doi: 10.1111/j.1524-4733.2008.00400.x. [DOI] [PubMed] [Google Scholar]
  37. Kroenke K, Spitzer R, Williams JW. The PHQ-9. J Gen Intern Med. 2001;16(9):606–613. doi: 10.1046/j.1525-1497.2001.016009606.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Lee SY, Arozullah AM, Cho YI, Crittenden K, Vicencio D. Health literacy, social support, and health status among older adults. Educ Gerontol. 2009;35(3):191–201. [Google Scholar]
  39. Littlefield CH, Rodin GM, Murray MA, Craven JL. Influence of functional impairment and social support on depressive symptoms in persons with diabetes. Health Psychol. 1990;9(6):737–749. doi: 10.1037//0278-6133.9.6.737. [DOI] [PubMed] [Google Scholar]
  40. Liu H, Golin CE, Miller LG, Hays RD, Beck CK, Sanandaji S, Wenger NS. A comparison study of multiple measures of adherence to HIV protease inhibitors. Ann Intern Med. 2001;134(10):968–977. doi: 10.7326/0003-4819-134-10-200105150-00011. [DOI] [PubMed] [Google Scholar]
  41. Mayberry LS, Gonzalez JS, Wallston KA, Kripalani S, Osborn CY. The ARMS-D out performs the SDSCA, but both are reliable, valid, and predict glycemic control. Diabetes Res Clin Pract. 2013;102(2):96–104. doi: 10.1016/j.diabres.2013.09.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Mayberry LS, Osborn CY. Family support, medication adherence, and glycemic control among adults with type 2 diabetes. Diabetes Care. 2012;35(6):1239–1245. doi: 10.2337/dc11-2103. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Mayberry LS, Osborn CY. Family involvement is helpful and harmful to patients’ self-care and glycemic control. Patient Educ Couns. 2014 doi: 10.1016/j.pec.2014.09.011. http://dx.doi.org/10.1016/j.pec.2014.09.011. [DOI] [PMC free article] [PubMed]
  44. Mendenhall E, Shivashankar R, Tandon N, Ali MK, Narayan KM, Prabhakaran D. Stress and diabetes in socioeconomic context: a qualitative study of urban Indians. Soc Sci Med. 2012;75(12):2522–2529. doi: 10.1016/j.socscimed.2012.09.040. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Naranjo DM, Fisher L, Arean PA, Hessler D, Mullan J. Patients with type 2 diabetes at risk for major depressive disorder over time. Ann Fam Med. 2011;9(2):115–120. doi: 10.1370/afm.1212. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Ngo-Metzger Q, Sorkin DH, Billimek J, Greenfield S, Kaplan SH. The effects of financial pressures on adherence and glucose control among racial/ethnically diverse patients with diabetes. J Gen Intern Med. 2012;27(4):432–437. doi: 10.1007/s11606-011-1910-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Osborn CY, Cavanaugh K, Wallston KA, Kripalani S, Elasy TA, Rothman RL, White RO. Health literacy explains racial disparities in diabetes medication adherence. J Health Commun. 2011;16(Suppl 3):268–278. doi: 10.1080/10810730.2011.604388. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Osborn CY, Egede LE. The relationship between depressive symptoms and medication nonadherence in type 2 diabetes: the role of social support. General Hosp Psychiat. 2012;34(3):249–253. doi: 10.1016/j.genhosppsych.2012.01.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Osborn CY, Mayberry LS, Wagner JA, Welch GW. Stressors May Compromise Medication Adherence Among Adults With Diabetes and Low Socioeconomic Status. Western J Nurs Res. 2014;36(9):1091–110. doi: 10.1177/0193945914524639. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Rook KS. Investigating the positive and negative sides of personal relationships: Through a lens darkly. The dark side of close relationships. 1998:369–393. [Google Scholar]
  51. Rosland AM, Heisler M, Choi HJ, Silveira MJ, Piette JD. Family influences on self-management among functionally independent adults with diabetes or heart failure: do family members hinder as much as they help? Chronic Illn. 2010;6(1):22–33. doi: 10.1177/1742395309354608. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Rothberg M, DuVal T, Luciano G, Frederici J, Welch GW. A scale for measuring social stress among diabetic patients. Paper presented at the 34th Society of General Internal Medicine Scientific Meeting; Phoenix, AZ.. 2011. [Google Scholar]
  53. Salas M, Hughes D, Zuluaga A, Vardeva K, Lebmeier M. Costs of medication nonadherence in patients with diabetes mellitus: a systematic review and critical analysis of the literature. Value Health. 2009;12(6):915–922. doi: 10.1111/j.1524-4733.2009.00539.x. [DOI] [PubMed] [Google Scholar]
  54. Schafer LC, Mccaul KD, Glasgow RE. Supportive and nonsupportive family behaviors - relationships to adherence and metabolic control in persons with type-I diabetes. Diabetes Care. 1986;9(2):179–185. doi: 10.2337/diacare.9.2.179. [DOI] [PubMed] [Google Scholar]
  55. Spencer MS, Kieffer EC, Sinco BR, Palmisano G, Guzman JR, James SA, Heisler M. Diabetes-specific emotional distress among African Americans and Hispanics with type 2 diabetes. J Health Care Poor Underserved. 2006;17(2 Suppl):88–105. doi: 10.1353/hpu.2006.0095. [DOI] [PubMed] [Google Scholar]
  56. Steiner JF, Prochazka AV. The assessment of refill compliance using pharmacy records: methods, validity, and applications. J Clin Epidemiol. 1997;50(1):105–116. doi: 10.1016/s0895-4356(96)00268-5. [DOI] [PubMed] [Google Scholar]
  57. Stephens MAP, Franks MM, Rook KS, Iida M, Hemphill RC, Salem JK. Spouses’ attempts to regulate day-to-day dietary adherence among patients with type 2 diabetes. Health Psychol. 2013;32(10):1029–1037. doi: 10.1037/a0030018. [DOI] [PubMed] [Google Scholar]
  58. Thoits PA. Mechanisms linking social ties and support to physical and mental health. J Health Soc Behav. 2011;52(2):145–161. doi: 10.1177/0022146510395592. [DOI] [PubMed] [Google Scholar]
  59. Thomas JL, Jones GN, Scarinci IC, Brantley PJ. Social support and the association of type 2 diabetes and depressive and anxiety disorders among low-income adults seen in primary care clinics. J Clin Pyschol Med S. 2007;14(4):351–359. [Google Scholar]
  60. Twist K, Stahl D, Amiel SA, Thomas S, Winkley K, Ismail K. Comparison of depressive symptoms in type 2 diabetes using a two-stage survey design. Psychosom Med. 2013;75(8):791–797. doi: 10.1097/PSY.0b013e3182a2b108. [DOI] [PubMed] [Google Scholar]
  61. van Steenbergen-Weijenburg KM, de Vroege L, Ploeger RR, Brals JW, Vloedbeld MG, Veneman TF, van der Feltz-Cornelis CM. Validation of the PHQ-9 as a screening instrument for depression in diabetes patients in specialized outpatient clinics. BMC Health Serv Res. 2010;10:235. doi: 10.1186/1472-6963-10-235. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Velicer WF. Suppressor variables and the semipartial correlation coefficient. Educ Psychol Meas. 1978;38(4):953–958. [Google Scholar]
  63. Welch GW, Osborn CY, Rotherber M, Zagarins SE, White RO, Friderici J. Level of social distress among Mexican and Puerto Rican American diabetes patients and relationship to glycemic control. Paper presented at the 71st Annual Scientific Session of the American Diabetes Association; San Diego, CA.. 2011. [Google Scholar]
  64. Wolfe R, Gould W. sg76: An approximate likelihood-ratio test for ordinal response models. Stata Technical Bulletin. 1998;42:24–27. [Google Scholar]

RESOURCES