Psychometric evaluation of the Problem Areas in Diabetes (PAID) survey in Southern, rural African American women with Type 2 diabetes

Type 2 diabetes is a leading cause of death for African American women [1]. Epidemiological evidence indicates that glycemic control in this patient group is suboptimal and that they suffer disproportionately from diabetes-related complications [2].

Though tight glycemic control is viewed as a primary indicator of favorable diabetes outcomes [3], a myriad of factors, including attention to diet, monitoring of blood glucose, medication, and physical activity [4], contribute to a patient's success in achieving desirable glycemic control. Additionally, psychological distress can negatively impact a patient's adherence to these necessary self-care regimens [5, 6].

The Problem Areas in Diabetes (PAID) survey was developed as a measure of diabetes-related stress that could be useful in measuring the association between psychological adjustment to diabetes and adherence to self-care behaviors [7]. This 20-item survey uses a Likert-scale format to assess the degree to which diabetes management and/or feelings about diabetes are problematic to patients. To date, research application of the PAID has been mostly in Caucasian, urban populations [8, 9] and, to a lesser extent, urban African Americans [1, 10]. Our goal was to assess the reliability and validity of the PAID in Southern, rural African American women with Type 2 diabetes.

Of the 160 patients that completed the survey, only 131 were included in the analysis. A total of 7 surveys were excluded because participants did not meet the age criterion. Twenty-two patients did not complete all 20 PAID survey items. Demographically, these patients were not different from those that completed all items. Table 1 provides a demographic and clinical profile of the 131 Southern, rural African American women that completed all PAID items. The profile depicts a middle age, obese, and low-income patient group with suboptimal glycemic control, the majority of which were non-smokers.

Internal consistency of the PAID was high (Cronbach's alpha = 0.94). Our factor analysis resulted in 2 factors with eigenvalues above 1.5 and with sound internal consistency (Factor 1 alpha = 0.85; Factor 2 alpha = 0.94). Orthogonal rotations best approximated simple structure for these 2 new factors. Factor loadings are presented in Table 2. Seven loaded on the first factor and 13 on the second factor. Items loading on the first factor were interpreted as a "Lack of Confidence in Self-Care Implementation" (Lack of Confidence Subscale) and those on the second factor as "Negative Emotional Consequences of Self-Care Implementation" (Emotional Consequences Subscale). We consequently elected to simultaneously investigate the construct validity of this 2-factor solution along with the single factor PAID solution.

Correlations of the Lack of Confidence Subscale, Emotional Consequences Subscale, and the overall PAID to the CES-D (Cronbach's alpha = 0.85) and SDSCA subscales (Cronbach's alpha range = 0.75 to 0.87-for 2-item scales only) were used to assess construct validity based on a priori hypotheses (Table 3). The Cronbach's alpha estimate of the "specific diet" subscale was too low to establish internal consistency so we did not use this subscale in construct validity studies. This lack of internal consistency of this subscale has also been reported in studies of largely Caucasian populations [12]. Correlations were high (r = 0.69–0.97) between the Lack of Confidence subscale, Emotional Consequences subscale, and the PAID. In keeping with our hypothesis that depression would be positively correlated with the PAID, positive correlations were observed between the Lack of Confidence Subscale, Emotional Consequences Subscale, and the PAID and the CESD (r = 0.55–0.58). All correlations between the Lack of Confidence and Emotional Consequences subscales and SDSCA subscales were negative. This confirmed our theoretical assumption that diabetes-related stress would be higher for patients that participated in self-care activities less frequently. For the Lack of Confidence Subscale, significant negative correlations were found with diet and foot inspection-related subscales only. For the Emotional Consequences Subscale, significant negative correlations were observed with diet and physical activity-related subscales. With the exception of the glucose monitoring subscale and the medication adherence question, significant negative correlations were found between the PAID single factor solution and other SDSCA subscales.

Correlations of the Lack of Confidence and Emotional Consequences subscales and the PAID with select demographic and clinical characteristics were used to examine predictive validity. Age was significantly negatively correlated with the each subscale and the PAID. Glycemic control, as measured by HbA1c, was significantly correlated (in a positive direction) with only the Lack of Confidence subscale. Body mass index was positively correlated with each of the subscales and the PAID. However, the association was only statistically significant with the Emotional Consequences subscale.

This is the first study to quantitatively evaluate the psychometric attributes of the highly utilized PAID survey in Southern, rural African American women with Type 2 diabetes. While the original 1-factor solution previously described performed well, a 2-factor solution revealed clearer correlations between psychological distress and glycemic control and BMI, two critical indicators for diabetes-related health outcomes.

Importantly, we were able to observe a significant positive correlation between the 1st subscale in our 2-factor solution, "Lack of Confidence in Self-Care Implementation" and Hba1c. The fact that Hba1c was not significantly correlated with the PAID, as a single scale, is inconsistent with other findings ([7, 9]) in largely Caucasian populations but consistent with findings in urban African Americans [10]. While the impact of race on PAID responses is not clear, it is possible that the PAID as a single scale, masks some of the more specific factors that impact emotions and behaviors that ultimately influence HbA1c in African American women. However, additional work is required to explore this notion. Figure 1 provides a conceptual model and shows that Lack of Confidence in implementing the diabetes self-care plan could potentially influence self-care, which might ultimately impact glycemic control. The clinical relevance of this finding is apparent, especially for this high-risk population, and is of critical importance in delineating relevant emotional factors that may influence self-care behaviors and, ultimately, impact glycemic control. This relationship between Lack of Confidence and glycemic control is consistent with the findings of others that confidence is a key determinant of self-care behaviors among African American women with Type 2 diabetes [13] and persons with Type 2 diabetes, in general [14]. Since there was not a significant correlation between the PAID and HbA1c, this finding also provides support for the additional utility of the PAID as 2-factor model.

Moreover, our 2-factor solution revealed a significant positive association between our 2^nd subscale, "Negative Emotional Consequences of Self-care Implementation" and BMI. This association is clinically relevant given the emerging obesity epidemic among African American women [15] and the disproportionate diabetes burden among African American women [2]. For example, items in the Emotional Consequences subscale represent the emotional consequences of implementing a diabetes self-care plan. Figure 2 illustrates this relationship. It shows that implementing a diabetes self-care plan could possibly result in negative emotions that might, in turn, impact self-care. Further, the degree to which specific components of the self-care plan are implemented can negatively influence BMI.

Therefore, our 2-factor solution allowed us to identify important correlates of HbA1c and BMI in Southern, rural African American women with Type 2 diabetes, ones that would not have emerged using the 1-factor PAID solution alone. Our work represents first steps in identifying relevant emotional factors for diabetes self-management and, ultimately, glycemic control among rural African American women with Type 2 diabetes. To the extent that our 2 subscales reliably and accurately correlate with Hba1C and BMI among this patient population, future work should focus on defining best methods for improving patient confidence in self-care activities that pose the most significant challenges and helping patients manage the emotional consequences of implementing self-care plans.

Our study has some limitations. Cognitive response interviews were not conducted. The addition of this method would have enhanced our ability to understand how patients perceived the meaning of individual PAID items. Our demographic questionnaire did not discriminate between patients receiving insulin via shots or pump. Therefore, our study results do not take into account differences attributable to distinct methods of insulin delivery and, possibly, adjustment issues. Another possible limitation was the potential for patients to respond to PAID questions in a manner that they felt was socially desirable (i.e. reporting that "feeling scared when you think about living with diabetes" was not a problem when indeed it was a significant one). Though impractical relative to minimizing patient fatigue during questionnaire administration, the additional administration of a social desirability questionnaire along with the PAID and other questionnaires would have been helpful in determining whether this was indeed true for this study [16]. Also, our most significant study results were found in a sample of obese patients with poorly controlled diabetes. Since our study only included rural, female patients, it is not clear the extent to which gender and locale alone correlated with these specific clinical characteristics. A more heterogeneous study sample would have been helpful in this regard. From a psychometric perspective, it was apparent from our factor analysis results that 3 of the PAID items loaded well (≥ 0.40) on both the Lack of Confidence and Emotional Consequences subscales. Since it was not our goal to modify the PAID items and, hence, possibly enhance factor loading clarity, we assigned these 3 items to the Emotional Consequences subscale based on three critical factors: 1) slightly superior loadings (with the exception of item 9 where loadings were equal); 2) face validity, and 3) better theoretical cohesiveness relative to our construct validity conceptualization. Finally, while it is appropriate and common practice to report correlations between 1- and 2-item SDSCA subscales and other scales, subscales with more items might lend to more robust and discriminating results.

The greatest strengths of our study were findings that lack of confidence in diabetes self-care implementation and the negative emotional consequences of self-care implementation were associated with Hba1c and BMI, respectively. Though the correlations that support these findings were modest, the clinical importance of identifying emotions that might impact glycemic control and body weight among Southern, rural African American women with Type 2 diabetes adds tremendous value to these novel findings. Other major strengths of this study were related to our recruitment efforts. We were able to recruit African American patients, a group that is typically perceived to be "hard to reach" relative to study recruitment ([17]). Additionally, we were able to recruit patients from rural locales, a success that is invaluable for involving geographically underserved populations in research ([18]).

A psychometrically sound two-factor solution to the PAID survey is identified in Southern, rural African American women with Type 2 diabetes. Lack of confidence in and negative emotional consequences of diabetes self-care implementation provide a better understanding of determinants of glycemic control and weight than an aggregate of the two scales. Further work is needed to replicate these findings and, subsequently, design interventions that to improve the lives of Southern, rural African American women with Type 2 diabetes.