Associations Between Diabetes-Related Distress and Cardiovascular Complication Risks in Patients With Type 2 Diabetes and Lower Socioeconomic Status: A Pilot Study

This was a cross-sectional study ascertaining the associations between data from the PAID questionnaire and UKPDS Risk Engine results. This study was performed at LifeLong Medical Care (LLMC) in East Oakland, Calif. LLMC is a federally qualified health center serving patients with lower SES. The research protocol was approved by the Touro University California institutional review board.

Potential participants were identified via a database search. Inclusion criteria included one or more visits at LLMC East Oakland between July 2013 and October 2014, diagnosis of type 2 diabetes, age 40–80 years, and diabetes care regimen including one or more antidiabetic medications. The reasons for choosing these inclusion criteria were our desires to 1) sample middle- to old-aged individuals living with type 2 diabetes and 2) include those who had shown some initiative in their health care by presenting at the clinic before our data collection period. These factors were used as the input for the search function in the electronic medical record system at the outpatient clinic. About 900 potential participants met the inclusion criteria, and 63 of these potential subjects presented to LLMC for pharmacy appointments and agreed to be participants during the study period; this group became our study population. All 63 of the potential participants signed an informed consent form outlining the overview and purpose of the study, the risks and benefits associated with the study, and the research team’s approach in using the data collected. The participants were then interviewed by the research team using the survey instrument.

The survey instrument included five sections: 1) demographic characteristics including occupation, 2) history of stroke and heart attack, 3) the PAID questionnaire, 4) information required to predict 10-year cardiovascular complication risks (CHD and stroke) using the UKPDS Risk Engine 2.0, and 5) exercise patterns and depression status. Participants were excluded from the study if they met any of the following criteria:

• • History of stroke or heart attack

• • Inability to complete or understand the PAID questionnaire

• • Insufficient information to calculate the 10-year cardiovascular complication risks using the UKPDS Risk Engine 2.0

• • Total PAID score of 0 (indicating that the individual did not perceive having any DRD

A score of 0 on the PAID questionnaire would not be informative as a value for the independent variable when constructing the regression models against 10-year cardiovascular complications risks (the dependent variable). From a clinical standpoint, the PAID instrument is a questionnaire surveying respondent’s self-perceived DRD. As with all such surveys, there was the possibility that participants did not understand the wording of the statements, thus choosing 0 for all questions. To take that into account, we decided to exclude those who reported a total score of 0.

After comparing the collected data with the exclusion criteria, we enrolled 48 patients in the study, excluding 15 of the original 63 individuals who completed the informed consent form.

Participants’ SES—defined by income, occupation, and education—was revealed by the Nam-Powers-Boyd scale, which is a “pure socioeconomic scale” using only an individuals’ occupation as a proxy to indicate their SES. Scores range from 1 (e.g., dishwashers) to 100 (e.g., dentists, physicians, or surgeons) (12).

The PAID questionnaire was used to measure participants’ distress levels specifically relating to diabetes management. This tool contains 20 Likert-scale questions, with possible responses ranging from 0 (not a problem) to 4 (a serious problem). The following statements are examples from the PAID questionnaire: “Worrying about the future and the possibility of serious complications” and “Feeling alone with diabetes” (13). The total PAID score is determined by the sum of the 20 individual scores and then multiplied by 1.25, so the maximum PAID score is 100 (14). Higher scores indicate higher levels of emotional distress, specifically reflecting self-perceived emotional status related to living with type 2 diabetes (14).

The UKPDS Risk Engine 2.0 is a risk calculator specifically designed to predict the risk of cardiovascular complications for people with type 2 diabetes. To estimate a patient’s 10-year cardiovascular complication risks for nonfatal or fatal CHD and nonfatal or fatal stroke resulting from diabetes (all of which were defined outcomes of the risk engine), we collected the following information from the enrolled participants’ electronic medical records: current age, duration of diabetes, sex, diagnosis of atrial fibrillation, race, smoking status, A1C, systolic blood pressure, total cholesterol, and HDL cholesterol (15).

All statistical analyses were performed using Stata 13 (Stata Corp., College Station, Tex.). The primary outcomes of this study were the associations between the total PAID scores (the predictor variable) and the 10-year risks predicted by the UKPDS Risk Engine 2.0 for developing cardiovascular complications (the outcome variable). These associations were tested by multiple linear regression models with and without the occupational status scores (as the mediator variable) from the Nam-Powers-Boyd occupational scale. We included two sets of regression models to explore the impact of SES, measured by occupational index scores, on the primary outcomes. Each multiple linear regression model included patients’ sex, years since diagnosis, smoking status, depression status, and race as covariates.

After applying the exclusion criteria, the study enrolled 48 participants. The sex-specific averages for age, occupational status scores, and total PAID scores are presented in Table 1. The average total PAID score was 24.1, with the average score of men 2.9 points higher than that of women, indicating that, on average, male participants felt slightly more distress from having to manage diabetes. Table 2 shows average PAID score by race. Only three races are presented in Table 2 because the UKPDS Risk Engine includes only the racial categories “white,” “Afro-Caribbean,” and “Asian-Indian.” The average PAID score of Afro-Caribbeans was more than 10 points lower than those of white and Asian-Indian participants. Total PAID scores from 13 individuals (27.1%) were ≥33, indicating that they were experiencing elevated DRD/depressive symptoms (16).

Clinical outcomes data for participants’ diabetes management are presented in Table 3. The average A1C was 8.5% (SD 2.0).

A summary of the predicted 10-year cardiovascular risks by UKPDS Risk Engine and the contributing factors is presented in Table 4.

The primary outcomes were the associations between total PAID scores and the results of the UKPDS Risk Engine 2.0 predicting the 10-year risks for cardiovascular complications. Total PAID scores—the predictor variable—were analyzed with each of the predicted risks—the outcome variable—with covariates via multiple linear regression models. Regression coefficients from the regression models, illustrating the relationships between the predictor and the outcome variables, are presented along with 95% CIs in Table 5. The mediator variable—occupational status scores—was added to one set of the multiple linear regression models to highlight the impact of the variable on the primary outcomes.

The reason for DRD (captured by the total PAID scores) being the predictor variable was because various degrees of DRD could potentially contribute to different levels of the predicted cardiovascular risks; thus, we constructed regression models to understand the degrees of association or impact, demonstrated by regression coefficients. Study participants might have experienced DRD but did not realize the magnitude of it until this specific type of distress was quantified by the PAID scale, acting as the independent/predictor variable. On the other hand, the predicted 10-year risks for cardiovascular complications depended on many clinical and individual factors, which were ultimately translated into percentages via the UKPDS Risk Engine 2.0, to be studied with DRD distress.

Associations between total PAID scores and nonfatal and fatal CHD, with and without occupational status scores, were statistically significant (P = 0.049 and P = 0.048, respectively), whereas no associations with stroke were statistically significant (P >0.05). The regression coefficient between the risks of nonfatal CHD and the total PAID scores (as a means of statistically linking the cardiovascular risks with the corresponding levels of DRD) was 0.149 (P <0.05), indicating that, for each point increase in total PAID score, nonfatal CHD risk increased by 0.149% (Table 5). Based on these results, the mediator variable (occupational status scores) did not have a significant impact on the aforementioned associations; all associations were validated by their respective statistically significant 95% CIs, with similar coefficients with and without the inclusion of the occupational status scores. However, no statistically significant results were observed from the associations of total PAID scores with nonfatal and fatal stroke, with and without occupational status scores (Table 5).

For additional analyses, the study cohort was divided into two subgroups based on participants’ SES revealed by their occupational status scores. The average occupational status score for the study cohort was 44.9, which was used as the dividing point between the lower occupational status score subgroup and the higher occupational status score subgroup. In the lower occupational status score subgroup, the associations between total PAID scores and the two estimated 10-year stroke risks were shown to be statistically significant (P <0.05); in participants with lower occupational status score, higher DRD was associated significantly with higher predicted 10-year nonfatal and fatal stroke risks, whereas such an association was not observed between total PAID scores and the predicted 10-year CHD risks. Results are summarized in Table 6. No statistically significant associations were observed in the higher occupational status score subgroup (the P values from all four multiple regression models were larger than the set alpha level of 0.05).

Forty-eight participants were enrolled in this pilot study. Through multiple linear regression analyses performed on the total PAID scores with each of the four cardiovascular complications risks predicted by the UKPDS Risk Engine 2.0 calculator, the results showed that DRD was positively associated with the predicted nonfatal and fatal CHD risks with and without including occupational status scores. When translating these statistically significant associations into clinical application, the coefficients from the regression models without occupational status scores indicated that, for each point increase in total PAID score, nonfatal CHD risk increased by 0.149%, and fatal CHD risk increased by 0.129%. The same concept applied to the statistically significant results from the regression models that included occupational status scores (Table 5). These results agree with findings in a previous study that DRD increased patients’ risk of developing ischemic heart disease (6).

No statistically significant associations were observed between total PAID scores and participants’ 10-year stroke risks (Table 5). A trial with >22,000 U.S. adults studied the association between perceived stress and risk of CVD; within the follow-up period, patients with diabetes who reported elevated perceived stress had a statistically significant increase in the incidence of stroke (17). The statistically insignificant associations demonstrated by the regression models between total PAID scores and the predicted 10-year stroke risks in this study warrant further investigations with a more representative study population. One factor worth noting is that 72.9% of the study participants were African Americans, who were at a higher risk for multiple chronic conditions, including diabetes and CVD. Also, other factors besides DRD, such as experiencing microaggressions in everyday life, can also contribute to the risk of heart conditions in this population.

This pilot study showed that lower SES, but not higher SES, significantly affected the associations between total PAID scores and 10-year predicted stroke risks. Our findings revealed a trend similar to results in a previous study, in which lower SES, leading to possible health care inequalities, was found to have an impact on disease management outcomes (e.g., risks of cardiovascular complications) (11). These findings are applicable to patients who seek medical attention at safety-net clinics, many of whom live with limited resources to manage their type 2 diabetes.

Our findings should also encourage primary care providers to incorporate the PAID questionnaire into their clinical practices as a way to identify DRD in patients with type 2 diabetes and to address related problems accordingly. Our results suggest that, for patients who experience elevated DRD (total PAID score ≥33), clinicians should consider addressing psychosocial needs because DRD is associated with 10-year predicted cardiovascular complications risks.

Diabetes education that is better oriented to address the psychosocial aspect of diabetes is warranted, since 27% of the participants in our study expressed experiencing high emotional distress (PAID score ≥33, identified by Hermanns et al. [16]) related to having to manage their diabetes. Patients with this level of DRD might benefit from seeing a certified diabetes educator (CDE) to initiate a more focused diabetes management plan with the aim of increasing patients’ self-efficacy in diabetes management. At the time the study was being conducted, there was no CDE serving at the clinic to provide a comprehensive approach to diabetes management.

Limitations to our study include the fact that the recruitment process was not random. Because recruitment largely depended on whether identified potential participants kept their appointments with the pharmacy team, there was a possible bias from the differences in attitudes between participants and those who did not participate (i.e., did not come to the appointments) with regard to managing chronic medical conditions. Successful treatment and management of diabetes and related conditions heavily depend on self-management; participants who kept their appointments with the medical team might have been able to better manage their psychological distress as a result of receiving more information and having more contact with their HCPs.

Accurate capturing of emotional states relating to diabetes management by the PAID questionnaire relied heavily on participants’ abilities to acknowledge and recall their feelings retrospectively and thus may have been influenced by forgetfulness, embellishment, and bias. Inaccurate recollection might have led to inaccuracies, for example, in participants’ self-perceived social support.

Another limitation has to do with the cross-sectional design of this study. We only collected psychological, social, and clinical data from participants at a specific point in time, without taking into account possible changes in these variables as time progressed. As a result, a more robust study design (e.g., a prospective cohort study) is needed to further investigate the associations between total PAID scores and predicted 10-year cardiovascular risks.

A more thorough chart review would have provided the following useful data: participants’ BMI, insulin status, number of daily insulin injections, number of active medications, number of diabetes medications, number of active medical problems, and residential locations. With a larger sample size and more parameters included for each participant, a more nuanced association between 10-year predicted cardiovascular complications risks and total PAID scores could be explored and revealed.

One way to reduce bias in future studies may be to reach out via telephone and recruit patients who did not keep their medical appointments. In addition, asking participants to elaborate on the reasons behind each of their Likert-scale answers on the PAID questionnaire could help to increase recall accuracy. For a future study with larger sample size, regression analyses on subpopulations such as by sex, by race, and by two or more such demographic characteristics could be performed to reveal more in-depth and specific associations between total PAID scores and predicted 10-year cardiovascular risks.

Through the use of the PAID questionnaire and the UKPDS Risk Engine 2.0, DRD was shown to be associated with participants’ predicted 10-year risks for fatal and nonfatal CHD, regardless of their SES. Participants’ SES did play a role in the association between the total DRD and 10-year predicted stroke risks. The results of this pilot study suggest that incorporating the PAID questionnaire to screen for DRD, especially in safety-net clinic settings in which patients have a lower SES, could help clinicians better address such DRD and reduce patients’ risks for future cardiovascular complications.

The authors thank Paul Perry, PhD, for providing research advice regarding the design of this study.

No potential conflicts of interest relevant to this article were reported.

Touro University California College of Pharmacy provided funding for this pilot study.

C.F.Y. collected data, wrote the manuscript, contributed to discussion, and reviewed/edited the manuscript. J.C. wrote the manuscript and contributed to discussion. G.M. wrote the manuscript, contributed to discussion, and reviewed/edited the manuscript. C.F.Y. is the guarantor of this work and, as such, had full access to all the data and takes responsibility for the integrity of the data and the accuracy of the data analysis.