A Short-Form Measure of Diabetes Distress Among Adults With Type 1 Diabetes for Use in Clinical Practice: Development and Validation of the T1-DDS-7

Living with diabetes entails continuous engagement and self-management activities such as monitoring glucose levels and adjusting daily choices about insulin, food, and activities. Diabetes distress is a multifaceted construct that refers to the perceived emotional burden of living with and managing diabetes (1–4). International data suggest that 20–40% of adults with type 1 diabetes experience moderate to high levels of diabetes distress (3,5). Furthermore, prevalence may be higher among vulnerable groups, such as ethnic minorities, people with lower socioeconomic status, and young people (5–9). Higher levels of diabetes distress are associated with difficulty managing diabetes, elevated HbA_1c, and increased risk of complications that are associated with greater morbidity and mortality (2,3,10). This highlights the importance of developing valid and reliable tools to assess levels of diabetes distress and assist in guiding targeted support.

A recommended approach to detecting diabetes distress is routine screening, which has been increasingly recognized as part of diabetes care (2,11). Commonly used screening tools include the 20-item Problem Areas in Diabetes (PAID-20) instrument (4) and the 17-item Diabetes Distress Scale (DDS) (12). However, neither was developed specifically for adults with type 1 diabetes, who present different disease-related experiences and unique sources of diabetes distress when compared with adults with type 2 diabetes (5). The Type 1 Diabetes Distress Scale with 28 items (T1-DDS-28) is a variation of the DDS developed specifically for adults with type 1 diabetes that measures total distress with seven derived subscales representing distinct sources of type 1 diabetes distress. It has previously been found to be a valid and reliable measure (5,13–15). However, its length may inhibit its use in time-limited settings, such as routine clinical visits in which health care professionals must include many health-related assessments.

The development and choice of a diabetes distress screening tool entails a dilemma between tool precision and completeness and feasibility in busy clinical settings. Short-form diabetes distress measurements can be a practical solution to assess diabetes distress in settings where time is limited. Existing short-form measures, such as the two-item Diabetes Distress Scale (DDS-2) (16) and the five-item Problem Areas in Diabetes instrument (PAID-5) (17), are commonly used and valued in clinical care. However, no validated short-form measure of diabetes distress exists that is specific to the distinctive features of distress among people with type 1 diabetes.

Therefore, we aimed to identify a short-form measure of the T1-DDS-28 that maintained the conceptual model of the original scale by including one subscale item from each identified source of diabetes distress among adults with type 1 diabetes: powerlessness, negative social perceptions, physician distress, friend/family distress, hypoglycemia distress, management distress, and eating distress (5). Moreover, we aimed to explore the psychometric properties of the seven-item short-form measure among 2,016 adults living with type 1 diabetes.

Systematic and comprehensive validation relying on existing evidence is important when reviewing new instruments assessing multidimensional constructs that must be evaluated subjectively, such as diabetes distress (18). The validation process and method for shortening the scale were inspired by existing guidelines (18–20).

Validation consisted of two phases. In the initial phase, T1-DDS-28 items were prioritized for the short-form measure, using content validation. Experts in type 1 diabetes care and research subjectively assessed each item for influence and importance based on their experiences in diabetes-specific settings. To retain the underlying conceptual model of diabetes distress as a multidimensional construct, we identified a candidate item within each of the seven T1-DDS-28 diabetes distress sources. Thus, we exercised pragmatism and also respected the multidimensionality of the construct of diabetes distress. The second phase consisted of validating the proposed short-form measure with psychometric analyses, using data from a large diverse sample of Danish adults living with type 1 diabetes.

To identify and validate content for a short-form measure of the T1-DDS-28, we consulted selected health care professionals and researchers working in the field of diabetes who were most likely to be its primary users. We contacted 16 diabetes nurse specialists, psychologists, and researchers by email from August to September 2022. Of these, 14 experts participated: seven psychologists, four nurse specialists, and three diabetes researchers. The experts had worked in diabetes care and research for an average of 12 years, were all women and Danish, and represented a range of national institutions and locations.

We asked the experts to complete a survey evaluating the influence and importance of all individual T1-DDS-28 items. They were first asked to rate each item based on its potential negative influence on persons with type 1 diabetes, using a scale from 1 (“minor degree”) to 10 (“very high degree”). They then ranked subscale items by how well they represented the specific source of diabetes distress (e.g., physician distress), assigning a rank of 1 to the most important item, a rank of 2 to the second most important item, and so on. A free-text option allowed for additional comments about individual items or the study objective. The experts were informed in detail about the purpose and methods of the study, data processing procedures, anonymity, and withdrawal rights.

To ensure full representation of the seven sources of diabetes distress in the T1-DDS-28 and maintain the underlying conceptual model, we predefined the criterion for retaining items in the short-form scale as best performance within each subscale. Retained items had the highest influence rating and were ranked as most important by the largest proportion of experts. If influence and importance ratings conflicted, retention was based on the influence rating. We also prespecified a minimum criterion for item performance to ensure consistency: retained items must be rated an average of 5 or greater on influence (from 1 to 10) with at least 35% of experts ranking it as most important. The criteria were specified based on discussions among the authors. In case of suboptimal performance, the candidate item with the second-best performance would replace the item, and the validation process would be repeated. However, this did not prove necessary.

To validate the proposed T1-DDS-7, we used survey data from a T1-DDS-28 validation study (15). A total of 2,201 adults living with type 1 diabetes from Denmark (response rate, 38.2%) participated in the survey. We excluded 185 individuals with invalid responses on one or more T1-DDS-28 items meeting criteria for inclusion in the short form, resulting in a study population of 2,016 participants. Three weeks after the first survey, 512 participants who agreed to be contacted again completed a retest survey. Data collection has been described in detail elsewhere (15).

The survey included the previously validated (15) Danish version of the T1-DDS-28, originally developed in U.S. (5), in which response options for each item were on a six-point Likert scale from 1 (“not a problem”) to 6 (“a very serious problem”). The total diabetes distress score was calculated as the mean of the 28 items and ranged from 1 to 6. Moreover, subscale scores were calculated for the seven sources of diabetes distress (5). Moderate diabetes distress is defined as a mean score of ≥2, indicating the need for support (5).

General well-being was measured using the World Health Organization-Five Well-Being Index (WHO-5) (21), which contains five items assessing different aspects of quality of life experienced in the prior 2 weeks. Furthermore, the survey included the PAID-20 instrument (4), from which scores on the PAID-5 (17) could be derived. Participants using an insulin pump (N = 684) answered a version of the survey omitting the PAID-20 (15). Finally, the survey also included questions about current insulin delivery modality (pump or pen), current glucose monitoring modality (continuous glucose monitor or test strips), and diabetes duration.

Data on age, sex, highest attained and registered educational level, household composition, most recent recorded HbA_1c, and diabetes complications were collected from Danish national registries (22), as close in time as possible to completion of the initial survey.

Characteristics of the study population were summarized using descriptive statistics. To assess the factorial validity of the proposed T1-DDS-7, we conducted an exploratory factor analysis (EFA) of the seven items retained in the short-form measure to identify the number of possible factors. Principal factor solution was used as the estimation method to account for the nonnormal distribution of data. We deemed factor loadings for individual items as acceptable if >0.40 and good if >0.70 (23).

To evaluate the reliability of the proposed T1-DDS-7, internal consistency was assessed with Cronbach’s α (α), with a level of 0.7 (24) predefined as acceptable because of the reduced number of items. Test-retest reliability was assessed as the correlation between T1-DDS-7 scores at baseline and 3 weeks. Diabetes distress has previously been found to be relatively stable over time (25), and we expected to observe highly correlated scores over a 3-week interval.

We assessed convergent validity by examining correlations between the proposed T1-DDS-7 and T1-DDS-28, as well as other available psychosocial characteristics that previous research has indicated were potentially related to diabetes distress. We used the Pearson correlation coefficient to measure relationships between the short-form measure and the full T1-DDS-28 scale and subscales and with PAID-20, PAID-5, WHO-5, HbA_1c levels, sex, and diabetes duration. Based on previous findings (e.g., 2,5,14,26), we expected higher T1-DDS-7 scores to be correlated with higher PAID scores, lower WHO-5 scores, higher HbA_1c levels, female sex, and shorter diabetes duration. Based on the cut point of mean T1-DDS-7 and T1-DDS-28 scores of ≥2, indicating moderate diabetes distress, we examined receiver operating characteristics: area under the curve (AUC), sensitivity, and specificity. To evaluate whether the T1-DDS-7 retained the multidimensionality of the T1-DDS-28, we also calculated sensitivity and specificity, compared with a score of ≥2 on each of the T1-DDS-28 subscales and a cut point of moderate diabetes distress of ≥2 for any subscale score.

Statistical analysis was carried out in R studio version 4.1.

Participants in both surveys gave informed consent to participate before commencing the survey. The validation survey among adults with type 1 diabetes was approved by the Danish Data Protection Agency, Valby, Denmark (ID: P-2020-34 and P-2019-812). More details are described elsewhere (15).

Data supporting the findings of this study are not publicly available, to maintain the privacy of participants. Please contact the corresponding author for further information.

In general, expert evaluations of the influence and importance of T1-DDS-28 items were highly consistent (Supplementary Table 1). We constructed T1-DDS-7 as the seven items rated by participating experts as having the greatest influence and importance (Table 1). The average influence rating for retained items was 5.4–8.6, and 46–86% of experts ranked each as the most important to retain. The level of agreement was especially high within some subscales, for example, management distress and hypoglycemia distress, and lower within others, for example, friend/family distress and negative social perceptions. Yet all retained items met the predefined criteria for retention. In addition, Pearson correlation coefficients for individual T1-DDS-7 items and corresponding subscale scores were 0.79–0.86 (Supplementary Table 1).

The average age of participants completing the T1-DDS-28 was 52 years, 51% were women, and most had a higher education and were married or living with a partner (Table 2). Respondents had had type 1 diabetes for an average of 29 years, 28% had at least one diabetes complication, and 60% used a continuous glucose monitor.

The EFA yielded a one-factor solution (eigenvalue, 2.83) accounting for 40% of the variance. Standardized factor loadings (Table 3) were acceptable for all individual T1-DDS-7 items (range 0.45–0.86), and Cronbach’s α was satisfactory (0.82). In addition, the 3-week test-retest results were highly correlated (r, 0.90). A sensitivity analysis showed similar factor loadings and α-values in subpopulations stratified on age (<50 and ≥50 years), diabetes duration (≤10 and >10 years), and education (unknown, basic, or vocational and higher education) (data shown in Supplementary Table 2).

Total T1-DDS-7 scores were strongly correlated with total T1-DDS-28 scores (r, 0.95) and moderately correlated with subscale scores (Table 4). In addition, total T1-DDS-7 scores were correlated with other frequently used diabetes distress measures. Correlations with PAID-20 (r, 0.89) and PAID-5 (r, 0.83) were comparable to those between T1-DDS-28 and the same measures (PAID-20 r, 0.91; PAID-5 r, 0.84) (Supplementary Table 3). Higher T1-DDS-7 scores were also correlated with lower WHO-5 scores (r, −0.56) and higher HbA_1c levels (r, 0.24). These were also comparable to correlations between T1-DDS-28 and the same measures (WHO-5 r, −0.58; HbA_1cr, 0.24) (Supplementary Table 3) and between PAID-5 and the same measures (WHO-5 r, −0.65; HbA_1cr, 0.23).

Furthermore, women had a higher average T1-DDS-7 score (mean, 2.1; SD, 0.9) than men (mean, 1.8; SD, 0.7; P value [t test], <0.001). Diabetes duration was negatively associated with total T1-DDS-7 score (r, −0.15; P value, <0.001). A sensitivity analysis showed similar correlation coefficients in subsamples stratified on age, diabetes duration, and education, all with the same direction of association and level of statistical significance (data shown in Supplementary Table 4). However, correlations between T1-DDS-7 score and HbA_1c were higher in the subpopulations below 50 years of age. Overall, all correlations were in the expected directions, demonstrating convergent validity of the T1-DDS-7.

The prevalence of diabetes distress as assessed by total T1-DDS-7 scores was high, with 42.2% of respondents reporting at least moderate diabetes distress (mean score, ≥2). Compared with the full T1-DDS-28 scale, the AUC was 0.91 (95% CI, 0.90–0.93), while sensitivity was 93% and specificity was 89%. Comparing the T1-DDS-7 with each T1-DDS-28 subscale score, we found sensitivity between 66% (for powerlessness and hypoglycemia distress) and 86% (physician distress) and specificity between 70% (physician and friend/family distress) and 97% (powerlessness) (Supplementary Table 5). With a cut point of diabetes distress for any subscale score indicating moderate distress, sensitivity was 53% and specificity was 100% for T1-DDS-7, compared with sensitivity of 48% and specificity of 100% for the original T1-DDS-28 cut point (Supplementary Table 5).

The T1-DDS-28 is a preferred assessment tool for detecting diabetes distress among adults with type 1 diabetes. However, its length may restrict its use in time-limited clinical settings, such as clinical consultations, and thus, we aimed to identify a short-form scale.

Through systematic content validation based on clinical relevance evaluated by experts in diabetes care and research, we selected seven items to retain in a short form of the T1-DDS-28. We found a high level of agreement about the most influential and important item within each diabetes distress source. Participating experts particularly agreed on the most important items within specific sources, such as management distress, while reaching a more moderate level of agreement for other sources, such as negative social perceptions. The proposed T1-DDS-7 showed satisfactory psychometric properties of factorial validity, convergent validity, and reliability in a larger sample of adults with type 1 diabetes. The psychometric properties (e.g., eigenvalue of the EFA and correlation coefficients for convergent validity) of the T1-DDS-7 were comparable with another commonly used short-form measure, PAID-5 (17). Furthermore, by retaining the conceptual model of the original T1-DDS-28, which encompasses various distinct sources of diabetes distress among adults with type 1 diabetes (5), the seven-item short-form measure continues to embrace the multidimensionality of the concept of diabetes distress. Therefore, we propose the T1-DDS-7 as a valid and reliable measure of diabetes distress among adults with type 1 diabetes (a template of the T1-DDS-7 questionnaire is available in Supplementary Material S6).

The short-form measure is useful in clinical settings to identify adults with type 1 diabetes for whom support may be appropriate and to monitor change over time (27). As with the full scale, the mean score of T1-DDS-7 items can be interpreted dichotomously, with ≥2 indicating moderate diabetes distress. In addition, the short-form scale enables a rapid overview of the seven distinct diabetes distress sources, and an elevated score on an individual item must also be interpreted as a cause for concern. T1-DDS-7 assessment is also intended to function as a psychosocial dialogue tool in routine consultations, increasing person-centered care by using the rapid overview of the individual diabetes distress sources as a guide to conversation about psychosocial aspects of living with diabetes (28).

We found that 42% of survey respondents experienced moderate to high diabetes distress assessed by the T1-DDS-7, in keeping with previously reported rates internationally (3,5). The accuracy of the short-form measure was acceptable, with 93% sensitivity and 89% specificity compared with the original scale. We recommend that a high score on the total T1-DDS-7 or any individual item(s) should prompt open dialogue and the administration of the full T1-DDS-28 scale or at least the corresponding subscale(s), for a more comprehensive identification of individual diabetes distress sources. This is needed to guide consultations and future interventions of support. In addition, any assessment using the short-form or full scale should be followed up by open-ended questions, given findings that, despite low scale scores, some adults can experience significant levels of diabetes distress or experience emotional problems arising from unmeasured sources (27,29).

This study has several strengths, including content validation by relevant experts in diabetes care and research. Other strengths include a large study population, the use of several psychosocial assessments to evaluate convergent validity, and the inclusion of a 3-week test-retest assessment. Limitations include the lack of male participants and, despite the involvement of several diabetes experts with extensive experience in practice and research, participation by other professions, such as medical doctors, would have been preferable. However, this is not likely to have influenced our findings. Compared with nonresponders (n = 3,575), responders differed in terms of higher age (mean age in years: 48.6 vs. 52.4) and by having a higher proportion of women (percentage women: 45.0 vs. 50.7) (15), so we cannot rule out the possibility that our results may be less generalizable to other population groups.

Sensitivity analysis revealed that the main findings did not differentiate in subsamples stratified on low or high age, diabetes duration, and educational attainment, demonstrating the robustness of the findings. However, we found that the association between diabetes stress and HbA_1c was stronger among younger participants (<50 years of age). We have not found any published literature that documented the association between distress and HbA_1c being stronger for younger adults. Existing literature indicates that relatively younger adults have higher distress and HbA_1c (30), and this linkage is intriguing and worthy of future study. Finally, as the study was conducted in a Danish setting, priorities may be different in other cultural settings with different health care and social structures and among other populations with different ethnicities.

In conclusion, we recommend the adoption of the T1-DDS-7 in clinical practice as a valid and reliable measure to identify diabetes distress levels among adults with type 1 diabetes, potentially followed by full-scale assessment and open-ended questions and, as needed, appropriate support. The most significant strengths of the T1-DDS-7 are its applicability in time-limited clinical settings, minimal burden of data collection, and ease of rapid review in consultations.

Acknowledgments. The authors sincerely thank Lawrence Fisher, University of California, San Francisco, CA, for his constructive and valuable guidance regarding the development and evaluation of the short-form measure, the experts within diabetes care and research who contributed valuable information for this study, the people with type 1 diabetes who participated in the survey, and Jennifer Green, Caduceus Strategies, for editorial assistance.

Funding. This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Duality of Interest. M.N., L.E.J., P.L., K.O., V.S., K.N., and I.W. are employed at Steno Diabetes Center Copenhagen, a public hospital and research institution under the Capital Region of Denmark, which is partly funded by an unrestricted grant from the Novo Nordisk Foundation. The funders had no role in this work. V.S. received lecture fees from Sanofi. K.N. received funding to her institution for participating in advisory boards from Medtronic, Novo Nordisk, and Convatec and for lecturing from Sanofi, Novo Nordisk, Medtronic, and Dexcom. Her institution received funding for studies she performed from Zealand Pharma, RSP Systems, Novo Nordisk, Medtronic, and Dexcom. U.P.-B. has served on advisory boards and received lecture fees from Novo Nordisk and Sanofi. P.L. and I.W. own stocks in Novo Nordisk. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. M.N., I.W., L.E.J., V.S., D.H., and K.O. conceptualized the study. P.L., K.N., and U.P.-B. contributed to the study’s design, curation, and critical review. M.N. and K.O. undertook the data analysis and interpreted the data. M.N., I.W., and K.O. drafted the first manuscript. All authors critically reviewed, edited, and approved the final version of the manuscript. K.O. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.