Article Text
Abstract
Background The purpose of this study is to determine the prevalence of diabetes distress and its relationship with health behaviours and clinical outcomes in low-income patients.
Methods Secondary analyses were conducted using baseline data from a clinical trial evaluating a diabetes self-management intervention. Interviews were conducted with 666 participants receiving care at nine safety net clinics in Missouri. Distress was measured using the Diabetes Distress Scale, and outcomes included medication adherence, physical activity, nutrition and clinical biomarkers (haemoglobin A1C (HbA1C), blood pressure, low-density lipoprotein (LDL) cholesterol).
Results In a sample of 666 participants, 14.1% and 27.3% of patients were identified as highly and moderately distressed, respectively, with higher rates among younger, female and lower income patients. When compared with moderately and no distress groups, highly distressed patients were less adherent to medications (20.7% vs 29.9% vs 39.4%, p<0.001) and had higher HbA1C values (9.3% (SD=2.0) vs 8.2% (SD=1.8) vs 7.8% (SD=1.7), p<0.001), diastolic blood pressure (81.8 (SD=9.4) vs 80.2 (9.7) vs 78.9 (SD=8.8), p=0.02) and LDL cholesterol (104.6 (SD=42.4) vs 97.2 (34.3) vs 95.5 (37.9)) In multivariable analyses, high and moderate distress were associated with lower medication adherence (OR=0.44; 0.27 to 0.23, p=0.001) and (OR=0.58; 0.42 to 0.79; p=0.001), respectively, and higher HbA1C in only the highly distressed group (B=1.3; 0.81 to 1.85; p<0.001) compared with the no distress group.
Conclusions Diabetes distress is prevalent and linked to poorer adherence to health behaviours and glycemic control in a sample of patients receiving care from low-income clinics.
- DIABETES
- PRIMARY CARE
- SOCIO-ECONOMIC
Statistics from Altmetric.com
Introduction
Individuals living with diabetes are faced with daily challenges to appropriately manage their disease. In addition to regular visits with physicians, routine self-care typically involves adhering to complex prescription medication regimens, monitoring blood sugar and blood pressure, watching for signs and symptoms of diabetes-related complications, following dietary advice, and engaging in physical activity. Patients are also asked to avoid unhealthy behaviours such as smoking and excessive drinking. Numerous studies have documented the difficulties patients face in managing their diabetes, yet fewer studies have examined the psychological impact of living with diabetes.1–5 This is problematic, as these challenges are likely to influence patients’ ability to successfully engage in health behaviours and achieve optimal health outcomes.
The concept of ‘diabetes distress’ has emerged to describe patient experiences with diabetes-related depression, anxiety, social support and coping skills. The multifaceted concept of diabetes distress appears to be more clinically salient, and recent studies suggest that it is likely to have a greater impact on glycemic control than depression.6–8 Given the relative paucity of literature on this topic and the varying results, there is a clear need for further research to understand the emotional impact of diabetes on adherence to health behaviours and clinical outcomes particularly in vulnerable populations.
In this study, we sought to determine the prevalence of diabetes-related distress among out-of-glycemic control (haemoglobin A1c (HbA1C) >6.5%) adult patients receiving care from safety net clinics. Safety net clinics specifically deliver care to uninsured and underinsured patients. We used a previously developed and validated assessment tool, the Diabetes Distress Scale (DDS),9 and also evaluated the association between distress, health behaviours and relevant clinical outcomes, including HbA1C, blood pressure and low-density lipoprotein (LDL) cholesterol. We specifically targeted the safety net population as these patients are disproportionately affected by low health literacy, poor social support and inadequate financial resources. It was hypothesised that the prevalence of diabetes-related distress would be high and that those reporting distress would be less likely to engage in recommended health behaviours and achieve optimal health outcomes.
Methods
Data presented here are part of a larger clinical trial evaluating the effect of an educational intervention on diabetes outcomes. Analyses include only baseline data, which were collected prior to patient randomisation to the intervention activity. Participants were recruited from 10 safety net clinics that provide care to medically underserved patients in one of three cities in Missouri. The chosen cities represent urban, suburban and rural settings across the state; every clinic provided care regardless of insurance or ability to pay.
Sample
Patients with a medical chart-documented diagnosis of diabetes at each of the participating clinics were initially eligible for the study. Additional criteria included (1) 30 years of age or older, (2) HbA1C >6.5%, (3) English-speaking and (4) no hearing, vision or cognitive impairment. Physicians and staff identified eligible patients and provided an overview of the study, when possible, at clinic encounters. Signed consent was obtained prior to participation, and multiple Institutional Review Boards including Northwestern University and each study site approved the study procedures. Recruitment occurred between September 2008 and January 2010. In total, 671 patients were enrolled in the trial, with 666 patients having complete data for this study's analysis. Following American Association for Public Opinion Research (AAPOR)10 guidelines, the participation rate approximated 80% across sites.
Procedure and measurement
Structured interviews were conducted at baseline where sociodemographic information, patient health behaviours and a health literacy assessment (Test of Functional Health Literacy in Adults Short Form)11 were collected. Specifically, medication adherence was assessed by the Morisky index,12 a common measure that includes four questions about intentional and non-intentional non-adherence to drug regimens. Scores with three or more positive item responses are considered ‘high adherence’. Blood sugar monitoring was measured by the yes or no response to one item, “In the last 12 months, has your blood sugar gotten too low?” Physical activity and nutrition were measured using scales from the Behavioral Risk Factor Surveillance System13; each scale was scored dichotomously as either meeting US recommendations or not. Clinical outcomes included HbA1C, systolic and diastolic blood pressure, and LDL cholesterol; these were obtained from patients’ medical record on the date closest to the baseline interview.
Diabetes distress
Distress was measured with the DDS,9 which includes 17 items to determine an individual’s diabetes-related concerns using a six-point Likert scale. Example items include, “feeling that diabetes is taking up too much of my mental and physical energy everyday” and “feeling that I am often failing with my diabetes regimen”. The scale yields a total diabetes-related distress score as well as four subscale component scores (emotional burden, physician-related, regimen-related and interpersonal). Scores are calculated by totalling the patient's responses for each item in the scale and dividing by the number of items in each scale. According to Fisher and colleagues in 2012, a score of less than 2.0 indicates little to no distress, a score between 2.0 and 2.9 indicates moderate distress and 3.0 and greater is considered high distress. Those in the moderate and high distress groups have been found to have poorer behavioural and clinical outcomes.14 The diabetes distress scale score has been examined as a continuous and categorical variable in prior studies; in the analyses presented we chose to use the latest categorical cut-offs given the possible increasing clinical detriment at each of the two identified thresholds.
The variables of self-reported depression (patient-reported outcomes measurement information system (PROMIS)) and anxiety15 were also examined in order to understand their role in relation to the construct of distress, and the contribution of depression and anxiety within any analyses of behavioural or clinical outcomes.
Analysis plan
Descriptive statistics (percentages, mean and SD) were calculated for each demographic variable. Bivariate analyses (χ2) were used to evaluate the association between the demographic variables and diabetes distress (none, moderate, high). Bivariate associations (χ2 or one-way ANOVA) between overall diabetes distress and each of the four subscales with health behaviours and clinical outcomes were then examined. Finally, associations between overall health behaviours and clinical outcomes and diabetes distress were then examined using multivariable, logistic and general linear regression models, while controlling for relevant covariates found to be statistically significant in bivariate analyses and clustering by clinic (p<0.05).
Subanalyses
We performed mediational analyses to determine the extent to which self-reported depression and anxiety attenuated the relationship between diabetes distress and the health behaviour and clinical outcomes of interest.16 The independent relationship between diabetes distress and the health behaviour and clinical outcomes was revisited, adjusting for all covariates. Next, the relationship between depression and anxiety and the health behaviour and clinical outcomes was examined separately. Finally for those associations found to be significant in bivariate analyses, separate models for depression and anxiety were added to the baseline distress models as mediators, and changes in the coefficient (or OR) for diabetes distress were analysed. All statistical analyses were performed using STATA software V.10.0 (StataCorp, College Station, Texas, USA). Analyses were reported to be significant when p<0.05.
Results
Table 1 provides an overall description of the study population, stratified by diabetes distress. The average age of participants was 54.8 years (SD=11.1). Overall, two-thirds of the sample was female (62.7%), white (66.2%), unemployed or retired (62.2%), with adequate health literacy (67.1%). The majority of participants (67.1%) had been living with a diabetes diagnosis for 5 years or longer and were managing on average 3.3 comorbid conditions (including diabetes, SD=1.5).
Approximately 14.1% of patients met the clinical threshold for high diabetes distress and 27.3% for moderate distress. Those who were identified as highly distressed were more likely to be younger, female, report a lower household income and have more comorbid conditions compared with those moderately or not at all distressed. In terms of the diabetes distress subscales, more participants met criteria (moderate and high combined) for having regimen-related (56.8%) and emotional burden (52.3%) distress, while fewer participants identified problems with interpersonal (33.6%) and physician-related distress (11.0%).
In bivariate analyses, patients who were highly distressed were less likely to be highly adherent to their medication regimens (high: 20.7% vs moderate: 29.9% vs non-distressed: 39.4%, p<0.001, table 2). In addition, the presence of diabetes distress was significantly associated with higher HbA1C levels (high: mean (M)=9.3%, (SD=2.0) vs moderate: M=8.2 (SD=1.8) vs non-distressed: M=7.8 (SD=1.7), p<0.001) and higher diastolic blood pressure (high: M=81.8 (SD=9.4) vs moderate: M=80.2 (SD=9.7) vs non-distressed: 78.9 (SD=8.8), p=0.02).
Among diabetes distress subscales, those who had high or moderate emotional-burden, regimen-related distress and interpersonal distress were significantly less likely to be highly adherent to their medications compared with those not identified as distressed (table 3). Regimen-related distress was associated with a lesser likelihood of meeting recommended physical activity guidelines (high: 27.4% vs moderate: 32.6% vs non-distressed: 39.2%, p=0.03). Additionally, all distress subscales, except physician-related distress, were associated with poorer glycemic control (table 4). For the remaining intermediary clinical outcomes, higher emotional burden and regimen-related distress were associated with higher diastolic blood pressure, while higher physician-related distress translated to higher LDL cholesterol. There were no significant associations for systolic blood pressure.
Multivariable models were conducted for those outcomes that were linked to overall diabetes distress scores (table 5). The presence of diabetes distress, both moderate and high, were significant, independent predictors of lesser likelihood of high medication adherence (adjusted OR (AOR) 0.58, 95% CI 0.42 to 0.79, p=0.001 and AOR=0.44, 95% CI 0.27 to 0.73, p=0.001), respectively. High diabetes distress was a significant predictor for higher HbA1C levels (β=1.33%, 95% CI 0.81 to 1.85, p<0.001). Diabetes distress was not independently associated with diastolic blood pressure.
Exploratory analyses with depression and anxiety
Analyses were performed with depression and anxiety to understand their role within the construct of diabetes distress in addition to the outcomes of interest. Both PROMIS symptom indices for depression and anxiety were moderately correlated with diabetes distress (depression r=0.44, p<0.001; anxiety r=0.39, p<0.001); there was a strong correlation between depression and anxiety (r=0.73, p<0.001). Bivariate associations were performed between depression and anxiety with every behavioural and clinical outcome (table 6). Patients who were more depressed and/or anxious were less likely to be highly adherent (depression: M=48.3 (SD=0.74) vs M=51.6 (SD=0.56), p=<0.001; anxiety: M=50.0 (SD=0.69) vs M=53.9 (SD=0.55), p<0.001). In addition, patients who were more depressed were less likely to meet recommended physical activity guidelines (M=49.0 (SD=0.75) vs M=51.2 (SD=0.55), p=0.02) and have higher diastolic blood pressure (r=0.12, p=0.002). There were no other significant bivariate associations with the outcomes.
Multivariable analyses were conducted to examine the explanatory effect of either depression or anxiety in previous models that included diabetes distress (medication adherence, glycemic control, diastolic blood pressure) (table 7). When depression and anxiety were included with diabetes distress, anxiety but not depression was a significant predictor for medication adherence (OR=0.96, 95% CI 0.93 to 0.99, p=0.01), but both levels of distress also remained significant (high: OR=0.55, 95% CI 0.31 to 0.97, p=0.04; moderate: OR=0.64, 95% CI 0.45=0.93, p=0.02) although slightly attenuated. When depression and anxiety measures were added to the models for HbA1C and diastolic blood pressure, neither variable was a significant predictor of the outcome, and the association between diabetes distress and HbA1C and diastolic blood pressure remained unchanged.
Exploratory analysis examining pathways
The significant relationship between distress and HbA1C led to an investigation of causal mechanisms. As medication adherence was the only health behaviour found to be related to diabetes distress and tested in multivariable analysis, we added this variable to the multivariable model for HbA1C to understand its contributory role in predicting glycemic control in relation to diabetes distress. Once added, the β coefficient for high diabetes distress remained unchanged (β=1.3) and statistically significant.
Discussion
This is the first study to our knowledge to document the prevalence of diabetes distress exclusively among patients receiving care in the safety net, who embody the most vulnerable populations. Our findings indicate a very high prevalence of diabetes-related distress: 41%, with one in seven patients reporting high levels of distress. In our sample, having diabetes distress had a detrimental impact on medication adherence and was strongly associated with worse glycemic control. Given the extent of our findings, there are several practice implications for consideration.
Findings from this investigation not only identified diabetes distress as a formidable risk factor for poorer glycemic control but also demonstrated that while the construct of distress may include aspects of depression and anxiety it remains a distinct set of symptoms. Our analyses found that while depression and anxiety may be more routinely assessed mental health symptoms, they were only moderately correlated with diabetes distress. Further, measures of depression and anxiety were also less predictive of clinical outcomes and did not explain or attenuate the relationship between distress and glycemic control. Our finding that associations between outcomes and moderate and high diabetes distress were strengthened rather than attenuated in mediational analyses suggests that anxiety and depression may have served as effect modifiers to the relationship. While diabetes distress, anxiety and depression were moderately correlated with one another, all were independently associated with outcomes. These variables are clearly connected to a degree, enough to enhance distress associations yet not explain them. This suggests that diabetes distress as a risk factor may be more clinically meaningful and should be considered for monitoring and intervention.
Examining the subscales, consisting with other studies the nature of patients’ distress in our sample appeared to stem mostly from the emotional toll of the disease and from the difficulties in managing their medication regimen.5 ,17 Themes included within the emotional burden subscale deal with a lack of physical and mental energy, feeling angry, scared, depressed, out of control, overwhelmed or hopeless because of their diabetes. Regimen-related distress highlights fears of failure in compliance with one's diabetes management and a lack of confidence or motivation to comply with their medication schedule, diet and blood sugar monitoring. While nearly one in three patients exhibited distress in one of these areas, it was the cumulative impact of distress related to diabetes, including interpersonal and physician-related distress, which translated to self-care concerns and poorer outcomes.
In considering possible pathways linking distress to clinical outcomes, our data were not entirely clear. It seemed that medication adherence would have the greatest impact, although non-significant trends for lesser adoption of other health behaviours were clearly evident. Because many patients had reported not maintaining several recommended health behaviours, it could be that the challenge of medication adherence was disproportionately difficult for those who were distressed. This is plausible, as Wolf and colleagues found that patients, especially those who are more cognitively distracted, may overcomplicate drug regimens by not consolidating complex regimens.18 However, these results should not be seen as conclusive as we did not assess in detail how patients were taking medicine. To that regard, many of our health behaviour measures were self-assessments of meeting a threshold level of an activity rather than direct measurement.
Our study has several other limitations that should be acknowledged. First, the cross-sectional design of the study limits our ability to understand or imply a causal mechanism linking distress to poor glycemic control. While it is intuitive to suggest that being distressed could distract a patient and negatively impact their health behaviours, leading to poorer glycemic control, the opposite could also be true. The sample was also over-represented by women, those who are less educated and with a lower household income; though this is typical of safety net clinic settings, where our findings are most generalisable. However, patients who were limited English proficient were excluded in this sample. As prior studies have repeatedly shown, language barriers have an obvious and direct impact on chronic disease self-management19 ,20; we are limited in our ability to understand the prevalence and nature of these patients’ distress.
Given the prevalence and extent of the association between diabetes distress with health behaviours and clinical outcomes, clinical screening may become necessary to detect those at risk and in need of intervention. The DDS was devised as a research and clinical tool, and its administration by clinical staff should be examined in greater detail. Before any screening efforts were to be underway, viable interventions for those identified as distressed must also be in place. According to the U.S. Preventive Services Task Force,21 a screening test can only be viewed as receiving a high ‘grade’ if there are good assessments in place that demonstrate appropriate sensitivity and specificity, and without unnecessary stigma, also a clear plan of action. In fact, the value of a distress measure could be not only applicable as an early detection tool, but for continued monitoring over time as part of intervention. However, it is unclear whether more extensive time with a primary care doctor or nurse, referrals to psychological services or case management would be the best method for reducing symptoms of distress.
Given the sizable impact on diabetes outcomes, a longitudinal study is warranted to fully detail the signal created by this distress scale and how various attempts to reduce certain aspects of distress allow patients with diabetes to be more involved in health behaviours and achieve their clinical goals.
What is already known on this subject
-
Diabetes distress is an emerging construct that encompasses a patient's emotional ability to cope with the self-management requirements of diabetes.
What this study adds
-
This concept has not yet been studied exclusively in a low-income, out-of-glycemic control population where distress is likely to be even more prevalent. As a result of this study, we know that diabetes distress is highly prevalent in low-income diabetes patients (42.1%) and that high distress is related to poorer adherence to health behaviours (medication adherence) and clinical outcomes (haemoglobin A1c, blood pressure and low density lipoprotein cholesterol).
References
Footnotes
-
Contributors All authors are justifiably credited with authorship, according to the authorship criteria. AUP was responsible for conception, analysis and interpretation of data, drafting of manuscript and final approval given. SCB, RMP, DS, DDeW and DF were responsible for critical revision of manuscript and final approval given. LMC was responsible for analysis and interpretation of data, critical revision of manuscript and final approval given. HKS and TCD were responsible for conception, critical revision of manuscript and final approval given. DCM was responsible for interpretation of data, critical revision of manuscript and final approval given. MSW was responsible for conception, design, analysis and interpretation of data, drafting of the manuscript and final approval given.
-
Funding This paper was supported by the Missouri Foundation for Health.
-
Competing interests None.
-
Ethics approval Northwestern University Institutional Review Board, Copernicus Group IRB.
-
Provenance and peer review Not commissioned; externally peer reviewed.