Assessing Diabetes Distress in Emerging Adults with Type 1 Diabetes: Development and Validation of the Problem Areas in Diabetes—Emerging Adult Version

Abstract

Background:

Emerging adults (ages 18–30) with type 1 diabetes experience sub-optimal glycemic and psychological outcomes compared with other groups. The emotional burden of the unending self-care needs of diabetes management appears to be related to these poor health outcomes. However, there is no validated measure of this emotional burden in the developmental context of emerging adulthood.

Purpose:

The primary aim of this study was to examine the psychometric properties of a new measure of diabetes distress in emerging adults with type 1 diabetes in the United States.

Methods:

In this cross-sectional study, emerging adults with type 1 diabetes completed an online survey, including measures of diabetes distress, depressive symptomology and the newly developed measure, the Problem Areas in Diabetes—Emerging Adult version (PAID-EA). Participants also answered demographic and clinical outcomes questions. Internal consistency, reliability, construct validity and the underlying factor structure of the PAID-EA were assessed.

Results:

Participants (N = 287, 78% women) had a median age of 24 years; 43% were full-time students; 78% wore an insulin pump; 90% used a continuous glucose monitor; mean self-reported A1C was 7.1 ± 1.2%. The PAID-EA demonstrated good internal consistency and reliability (Cronbach’s α = .89), was comprised of one component accounting for 29% of the observed variance and demonstrated construct validity as it was significantly correlated with known measures of similar constructs, as well as with A1C levels (rho = 0.20, p = 0.001).

Conclusions:

The PAID-EA holds promise as a reliable and valid measure of diabetes distress in emerging adults.

Introduction

Emerging adults (ages 18–30) with type 1 diabetes navigate the many normative milestones of this developmental stage within the day-to-day context of diabetes management [1]. During this life stage, emerging adults face many transitions and challenges [2] that can exacerbate the diabetes-specific stressors of life with type 1 diabetes [3]. Recent evidence suggests that emerging adults with type 1 diabetes constitute a specific group that experiences worse glycemic control as well as more frequent episodes of severe hypoglycemia and diabetic ketoacidosis than any other age group [4,5].

Diabetes distress is the set of negative emotional or affective experiences that result from living with the unrelenting self-care demands of diabetes management [6,7]. Elevated diabetes distress is associated with higher hemoglobin A1C (A1C) levels, impaired quality of life and less frequent self-management behaviors [8,9]. It is believed that the effect of diabetes distress on glycemic control is mediated by self-care behaviors where the negative emotional burden of type 1 diabetes reduces self-efficacy and motivation to complete the repetitive, multiple daily management tasks of diabetes.

Diabetes distress is experienced within the developmental context of one’s life stage. During emerging adulthood, young people become geographically distinct from their parents for the first time, become increasingly financially independent and assume more personal control of their own health and healthcare [10]. Typically, friends and romantic partners become the most meaningful relationships in their lives instead of their family of origin [11]. These multiple transitions and stressors can exacerbate the diabetes-specific emotional burden of living with type 1 diabetes [12].

Yet, despite the acknowledgment of the effect of life stage on the experience of diabetes distress for children [13,14], adolescents [15], parents of young people with type 1 diabetes [14,16,17] and adults [18,19] there is no measure of diabetes distress specific to the developmental context of emerging adulthood. Thus, current measures inadequately capture the developmental hurdles, the many layers of transition and the high levels of uncertainty that characterize living with type 1 diabetes during this time period [20]. This underestimation may make it difficult to accurately determine significant associations between diabetes distress and other outcomes. To fill this critical gap, the Problem Areas in Diabetes—Emerging Adult version (PAID-EA) was developed, reduced and refined to measure diabetes distress in emerging adults living with type 1 diabetes in the United States [21,22].

Therefore, the purpose of this study was to further evaluate the psychometric properties of the PAID-EA. Specifically, internal consistency, reliability, construct validity and the underlying factor structure of the PAID-EA were assessed. Additionally, associations with demographic and clinical variables as well as between group differences in PAID-EA scores were explored.

Methods

Recruitment

Three online recruitment approaches were undertaken. First, the College Diabetes Network (CDN), a 501c3 non-profit company whose mission is to improve the lives of young adults living with type 1 diabetes, posted about study recruitment on their social media accounts (accounting for 23% of participants). Secondly, multiple emerging adult social media influencers in the diabetes online community were recruited to post on Instagram about life with type 1 diabetes, with approved recruitment language appearing in the comments[23] (53% of participants). These influencers received $20. Lastly, in order to ensure adequate representation of men, the Diabetes Education and Camping Association listserv shared recruitment language through emails with staff and alumni targeting young men (less than 24% of participants). Recruitment lasted for 3 months.

Procedures

By clicking on the survey link, participants were directed to the Qualtrics survey platform [24]. Inclusion criteria were: age 18–30 years old, have type 1 diabetes and able to read and write English. All participants had the opportunity to enter a raffle for one of five $100 Amazon gift cards. All phases of this project were approved by the Institutional Review Board at Boston College.

Measurement

The online survey included demographic and clinical questions. To approximate socioeconomic status, median household income (in US dollars) from US Census data was captured by the participant provided zip code that they identified as home [25]. Additionally, one question assessed the severity of the impact of the COVID-19 pandemic on the participant’s life and one question asked if the pandemic had affected their mood.

The first measure was the original Problem Areas in Diabetes (PAID), which includes 20-items about the emotional burden of living with diabetes [18]. Higher scores indicate more DD. For this study, high DD was defined as PAID score ≥ 40 [26] and moderate, possibly clinically significant, DD was defined as scores from 33 to 39 [27].

Next in the survey was the newly developed and refined PAID-EA. As previously described [21,22], the PAID-EA was developed first by considering the known and previously validated measures of diabetes distress, next items were expanded upon by diabetes experts of diverse specialties and then these items were piloted and discussed in an informal focus group of emerging adults to conceptually clarify the measure. During the revision and refinement of the PAID-EA using Rasch analysis in a previously recruited sample [21], model statistics were reviewed in parallel with cognitive interviews of emerging adults to ensure individual item comprehension and content validity. The refined 25-item PAID-EA, with 5-point Likert-type response options scored as 0–4, reflects the emotional burden of living with type 1 diabetes specific to the developmental stage of emerging adulthood. For this psychometric analysis (with missing data imputed), a total distress score was computed by summing responses. Higher scores indicate more diabetes distress, with scores that can range from 0–100. For future clinical use, calculating a mean score for all answered items and then multiplying by 25 may be more useful and avoid the need for imputation.

The final measure was the Center for Epidemiologic Studies Depression Scale (CES-D), a 20-item self-report assessment of depressive symptomatology [28]. Higher scores indicate greater frequency of depressive symptoms. For this study, clinically significant depressive symptoms were defined as a score ≥16 [29].

At the end of the survey, several open-ended questions were included. One question asked which feeling in the earlier items was the most important and another question specifically asked if there was anything that was missed in the prior items. The entire survey took 10–15 minutes to complete.

Analyses

Statistical analyses were performed using SPSS v26 [30]. Categorical data are presented as frequencies by n (%); data for continuous variables are summarized using measures of central tendency and dispersion (median or mean ± standard deviation). Few data points were missing after data cleaning (PAID: 2 items from 2 participants, PAID-EA: 18 items from 11 participants, CES-D: 7 items from 5 participants); missing data were replaced with the mode of the response option for that item as the data was categorical and to allow for comprehensive item analysis in this psychometric evaluation.

Item analysis was undertaken to assess item difficulty, discrimination, and corrected and uncorrected item-to-total correlations. Reliability analysis was conducted to assess internal consistency using Cronbach’s alpha. To determine if the PAID-EA was appropriate for factoring, Bartlett’s test of sphericity and Kaiser-Meyer-Olkin (KMO) test for sampling adequacy were assessed [31,32]. To examine the internal structure of the data, the PAID-EA items was subjected to exploratory factor analysis using principal components analysis (PCA) [33]. Inclusion of components were assessed by examining the scree plot [34] as well as comparison with a parallel analysis [35].

Spearman correlations were used to determine the strength and direction of relationships with clinical variables and to assess criterion-related validity [36]. Two-tailed independent sample t-tests and one-way between groups analysis of variance (ANOVA) with Scheffe test for post-hoc pairwise comparisons were conducted to explore demographic differences in diabetes distress. Responses to the open-ended questions were reviewed and organized by theme; these results were used to help inform the interpretation of the above statistics and for confirmation of content validity.

Results

Participant characteristics

The study sample consisted of 287 emerging adults with type 1 diabetes, with a median age of 24 years (range 18–30) (See Table 1). Two hundred and twenty-four participants identified as women (78%), 60 (21%) identified as men and 3 (1%) identified their gender in another way. Though 258 (90%) identified as white, 29 (10%) identified with another race, either exclusively or in addition to white, and 31 (11%) identified as being of Hispanic ethnicity. Many in the sample were students, with 124 (43%) as a full-time student and 21 (7.3%) as a part-time student. Additionally, reflective of the healthcare transitions that occur during this life stage, 134 (47%) had their own health insurance, whereas 149 (52%) remained on their parents’ health insurance and 4 (1.4%) were uninsured. A substantial proportion of the sample had access to advanced diabetes technologies: 223 (78%) wore an insulin pump recently, 257 (90%) used a continuous glucose monitor recently and 169 (59%) reported never checking a fingerstick blood glucose because they wear a continuous glucose monitor. Given the relatively high diabetes device use in this sample, it is not unexpected that the sample had a mean A1C of 7.1 ± 1.2% with a range of 4.9% - 11.4%. Finally, to begin to understand the impact of the COVID-19 pandemic, about two-thirds of participants reported that the pandemic had a moderate or severe impact on their life (188, 66%) and that their mood was worse than it was previously (199, 69%).

Table 1.

Participant demographics (N = 287)¹

Age (years)	24 (18–30)
Gender: Women	224 (78%)
Men	60 (21%)
Identify another way	3 (1%)
Age at diagnosis (years)	11(1–26)
Race (% White)	258 (90%)
Identify as another race (can include White)2	29 (10%)
Hispanic ethnicity	31 (11%)
Median household income for home zip-code	$77,020 ± $31,136 ($19,628 – $208, 212)
Currently a Full-time student	124 (43%)
Part-time student	21 (7.3%)
Have own health insurance	134 (47%)
On parents’ health insurance	149 (52%)
Uninsured	4 (1.4%)
Used insulin pump in last 30 days	223 (78%)
Used continuous glucose monitor in last 30 days	257 (90%)
Never check fingerstick blood glucose because of CGM	169 (59%)
Self-reported A1C (%)	7.1 ± 1.2 (4.9–11.4)
COVID-19 impact on life (% moderate or severe)	188 (66%)
COVID-19 impact on mood (% worse than before)	199 (69%)
PAID score3	38.4 ± 19.9 (0–96)
CES-D score4	20.7 ± 12.3 (0–55)

¹Data are median (range), n (%) or mean ± SD (range)

²Participants could select all that apply from the following options: Black or African American; American Indian or Alaska Native; Asian; Native Hawaiian or Pacific Islander; or an open-ended response.

³Problem Areas in Diabetes scale

⁴Center for Epidemiologic Studies Depression scale

Reliability analyses

Cronbach’s alpha for the PAID-EA measure was .89. Corrected item-total correlations (item discrimination) ranged from .280 to .656, all above the .2 minimum value for retention in the measure [37]. There were 2 items between .2 and .3 (Item 21: I worry about the cost of diabetes: .280 and Item 9: I worry that my blood sugar will go low or high during sex: .248). It is important to note that upon examination of the inter-item correlation matrix, Item 21 had a very low correlation with other items (range .096 to .264) as this item had the highest mean score of 3.52. Consideration was taken to explore how Item 9 (mean score 2.21, correlation with other items ranged from 0.14 to .294) functioned in the remaining analyses as it addressed intimacy while living with type 1 diabetes. Importantly, the deletion of any item would not have raised the Cronbach’s alpha above .89. Thus, all 25 items were retained.

As the PAID-EA item responses were continuous, the arithmetic mean for each item (item difficulty) suggests that participants endorsed much of the emotional burden described in the items (See Table 2) and the mean item score ranged from 1.32 to 3.52. These results support the clinical continuum of diabetes distress identified in the prior Rasch analysis [21].

Table 2.

Mean item scores and component loadings for the PAID-EA (N = 287)¹

Number	Item content	Mean ± SD	Component 1 loadings
1	I feel that diabetes is taking up too much of my mental energy every day.	2.5 ± 1.3	0.694
2	I feel annoyed when people say something ignorant about having diabetes.	3.3 ± 0.9	0.353
3	I am tired of having to explain diabetes to others.	2.4 ± 1.3	0.537
4	I have other things in my life that keep me from managing my diabetes.	2.0 ± 1.3	0.515
5	I feel judged by others because I have diabetes.	1.9 ± 1.4	0.520
6	I worry about being able to socialize because of how alcohol affects my blood sugar.	1.5 ± 1.5	0.539
7	I worry that a new romantic partner will see my diabetes devices or supplies.	1.3 ± 1.5	0.446
8	I worry about having kids in the future because of my diabetes.	3.1 ± 1.3	0.516
9	I worry that my blood sugar will go low or high during sex.	2.2 ± 1.5	0.272
10	I worry that diabetes will get in the way of what I want to do with my life,	2.3 ± 1.4	0.649
11	I feel that I must be perfect in my diabetes management.	2.6 ± 1.3	0.478
12	I feel alone with diabetes.	1.8 ± 1.5	0.594
13	I avoid doing diabetes management tasks when other people are around.	1.5 ± 1.4	0.530
14	I feel like a failure when I have a high A1C.	3.1 ± 1.1	0.488
15	I feel overwhelmed about having to do diabetes all by myself.	2.4 ± 1.4	0.674
16	I worry about living alone because I have diabetes.	2.2 ± 1.6	0.475
17	I don’t want to know my blood sugar when it is high.	1.9 ± 1.6	0.509
18	I feel frustrated about interruptions from diabetes (during sleep, work, school).	3.4 ± 0.9	0.508
19	I worry about diabetes complications.	3.2 ± 1.1	0.502
20	I don’t know how to make diabetes a priority when I have a lot of changes in my life.	3.2 ± 1.1	0.664
21	I worry about the cost of diabetes.	3.5 ± 0.9	0.320
22	I worry about having a low blood sugar.	2.9 ± 1.2	0.394
23	I am too tired of having diabetes to take care of it.	1.7 ± 1.5	0.636
24	I feel like I cannot take as many risks as my friends.	2.8 ± 1.4	0.627
25	I feel like I am trying my hardest to take care of diabetes, but it never works.	2.1 ± 1.4	0.713

¹The PAID-EA has a 5-point Likert-type response options scored as 0–4, where “Disagree” is 0, “Somewhat disagree” is 1, “Neutral” is 2, “Somewhat agree” is 3, and “Agree” is 4.

Factor analyses

An exploratory approach (PCA) was employed to examine the dimensionality of the PAID-EA and understand the number of components underlying the measure. The KMO test for sampling adequacy was .897 and the Bartlett’s Test of Sphericity was significant, χ²(300) = 2,150.71, p <.001, indicating an analyzable correlation matrix and appropriateness for factoring. A single component solution accounted for 28.9% of the variance (See Table 2). Inspection of the scree plot revealed a clear break after the first component (See Supplemental Figure 1) and this was further supported by the results of parallel analysis, demonstrating only one component with an eigenvalue exceeding the criterion value for a randomly generated data matrix of the same size (25 variables for 287 participants). All items loaded strongly on the single factor and all items were above the minimum factor loading for exploratory factor analysis of .2 (range .272 to .713).

Construct validity

Evidence supporting criterion-related validity was observed for the PAID-EA as scores were significantly correlated with scores on the PAID (rho = 0.76, p < 0.001) demonstrating concurrent validity. Additionally, evidence of convergent validity was observed for the PAID-EA as scores were significantly correlated with scores on the CES-D (rho = 0.61, p < 0.001). Scores on the PAID-EA were also significantly correlated with A1C (rho = 0.20, p = 0.001).

Emotional burden of type 1 diabetes in emerging adults

This sample of emerging adults carried a substantial emotional burden of life with type 1 diabetes. Using the clinically significant cut-offs for the PAID, 135 (47%) of participants met the criteria for high diabetes distress, with an another 34 (12%) meeting the criteria for moderate diabetes distress. Additionally, this sample endorsed a considerable amount of depressive symptomology, where 171 (61%) met the criteria for clinically significant depressive symptoms. Though determining clinically significant cut-offs for the PAID-EA is beyond the scope of this paper, emerging adults reported mean PAID-EA score of 59.5 ± 17.6 (range 9–94) and a median score of 61, both of which are substantially above a score of 50 (average “Neutral” response). Importantly, no participant indicated “Disagree” for all items on the PAID-EA, demonstrating all participants identified in some way with items on the PAID-EA, with many items very strongly endorsed by participants (see Table 2). The most common concern for participants was Item 21: “I worry about the cost of diabetes” with 204 (71%) of participants endorsing “Agree.”

Demographic differences in diabetes distress

This sample included emerging adults who described their gender in a variety of ways; there was a significant difference in PAID-EA scores when comparing those who identified as women and men (p = 0.001), where those who identified as women (61.4 ± 16.1) were significantly higher than men (51.7 ± 20.7). Participants who endorsed another gender identity were not included in this comparison as that portion of the sample was too small (n = 3) but these participants had a rather high mean PAID-EA score (78.3 ± 6.8). Also, PAID-EA scores were significantly correlated with age (rho = −0.12, p = 0.04) but not with age at diagnosis (rho = 0.11, p = 0.07). Furthermore, there was no significant difference in PAID-EA scores between those participants who identified as white or identified as another race (which included participants who identified as both white and another race, 10% of the sample) (p = 0.58), as well as between emerging adults who identified as not being of Hispanic origin and those who did (p = 0.81).

Additionally, there was no significant difference in PAID-EA scores between those participants who had their own insurance, were on their parents’ insurance or were uninsured (p = 0.06). Also, there was no significant difference in PAID-EA scores between median household income categories of <$50,000, $50,001–75,000, $75,001–100,000 and >100,000 (p = 0.78).

The participants who reported their lives were moderately or severely impacted by the COVID-19 pandemic (66% of the sample) had significantly higher mean PAID-EA scores (62.4 ± 16.6) compared with those who were less impacted (54.1 ± 18.2; p < 0.001). Those participants who reported their mood was worse during the pandemic (69% of the sample) had significantly higher mean PAID-EA scores (63.5 ± 16.3) compared with those participants reporting no change or an improvement in mood (50.4 ± 17.1; p < 0.001).

Content validity

Open-ended questions provided additional insight into the content validity and functioning of the PAID-EA. Almost all participants (254, 89%) responded to at least one of the open-ended questions; 212 participants responded to the question about which emotions related to living with diabetes were not included in the survey and over half of those responses (51%) indicated that nothing was missed. The remaining responses identified many feelings that were in fact included in the measure. Some participants identified feelings that were intentionally not included because there are other known, validated measures of the construct, such as disordered eating behaviors, fear of hypoglycemia and anxiety. Additionally, the open-ended question about the most important feeling related to living with type 1 diabetes supported the content validity of several items that appeared particularly meaningful. These items included worry about cost of diabetes, worry about the future (complications, having children) and how diabetes affects relationships (romantic partners, friends). Taken together, the open-ended responses provided evidence of content validity of the PAID-EA from the stakeholders who live with these emotions every day.

Discussion

Emerging adults with type 1 diabetes need to negotiate the many developmental challenges within the context of life with type 1 diabetes, therefore it is critical to accurately measure the emotional burden of living with type 1 diabetes within this context. Findings from this study provide evidence supporting internal consistency and demonstrate the PAID-EA is composed of one main factor. Construct validity of the PAID-EA is supported by strong correlations with both the PAID and the CES-D, as well as significant correlations with A1C and female gender. Additionally, the conceptual continuum of low to high diabetes distress identified in the prior Rasch analysis of the PAID-EA [21,22] is also supported by the current findings.

The PAID-EA includes many developmental stage-specific experiences, and, to our knowledge, it is also the first measure to include an item addressing the financial burden of diabetes in a measure of diabetes distress. Our findings demonstrate that this financial worry may be substantial for emerging adults living in the United States and may not only be related to insurance coverage or financial resources, consistent with recent evidence [38,39]. Further study is needed to clarify this finding, as insurance and the cost of diabetes are particularly complex in the United States and this experience may be different in other countries with universal healthcare. Also, future research must consider the often transitory addresses, fluctuating income and varying degrees of family support when evaluating healthcare access and financial resources during this developmental stage.

Though exploring intimacy and seeking a partner are major developmental milestones in emerging adulthood [2], these findings suggest that this experience may be highly individualized. These results may reflect the wide age range of this sample, where some participants are in college, some have started their careers, some are still seeking a partner, some are settled in a long-term relationship and some have decided to not pursue intimate relationships. Recent research highlights how the type of relationship, such as spontaneous intimacy with a new partner versus a long-term committed partnership, impacts the worry about diabetes during intimacy, including disclosing diabetes, how to deal with diabetes technology, and troubleshooting hyper or hypoglycemia [40,41]. In future studies, it will be important to tease out these differences to explain how type 1 diabetes influences intimacy during emerging adulthood.

It is becoming clear that the COVID-19 pandemic has had a substantial impact on many young people. The results of this study suggest that emerging adults who describe a more consequential impact of the pandemic or experienced a worsening of their mood during this time also reported more diabetes distress, consistent with recent reports [42,43]. Though further exploration is needed, these results begin to describe the consequences of this historical moment on emerging adults’ emotional experience.

Despite attention to rigor, this study does have a few limitations. This study used a convenience sample of emerging adults living in the United States recruited through social media and electronic communication, which may limit generalizability. In fact, this sampling technique captured a sample of mostly white women with higher socioeconomic status, with a majority using diabetes technologies and with close to target glycemic control; these sampling issues may impact DD [44–46] and further research is needed. Though the authors took an intentional approach to outreach, the sample ultimately had a limited amount of diversity. While it is unclear why this is the case, it may be that individuals engaging with the particular influencers used in this study tended to be white and female identifying social media users. As the users of social media continue to become more diverse[47], future researchers must explore ways to broaden engagement. Future validation work on the PAID-EA would benefit from in-person recruitment approaches at the conclusion of the COVID-19 pandemic. Additionally, in recognition of the transitory demographics of emerging adults, socioeconomic status was captured by using the zip code participants identified as home, however this proxy measure still has limitations. Lastly, as this was an online survey, it cannot be assessed if respondents were different than those who did not respond and all demographic and clinical variables were self-report with its inherent limitations.

Conclusion

These findings support the PAID-EA as a promising, reliable, valid and developmentally-embedded measure of diabetes distress in emerging adults. Though the PAID-EA must be subjected to further psychometric testing, including clarifying the ideal scoring strategy for clinical use and determining useful cut-off point scores, the PAID-EA has the potential to advance the science of measurement of the emotional burden of type 1 diabetes for both clinical and research use.

Supplementary Material

1

Supplemental Figure 1. Scree plot of eigenvalues for the PAID-EA items, demonstrating the point of inflection at the second component supporting a single-component solution.

Key messages:

• Emerging adults with type 1 diabetes experience worse glycemic control and psychological outcomes compared with other age groups.

• The Problem Areas in Diabetes—Emerging Adult version measure of diabetes distress holds promise as a reliable and valid measure.

• Measurement of developmentally embedded diabetes distress in emerging adults may explain these suboptimal outcomes and provide opportunity for intervention.