The total sample (N = 358) comprised outpatients recruited in collaboration with the Department of Endocrinology and Metabolism at the General University Hospital in Prague and participants in educational camps for patients with DM. The participants were included based on these criteria: the presence of a diagnosed T1D, T2D, gestational diabetes, or another type of diabetes (maturity-onset diabetes of the young, MODY, disease of pancreas, etc.); aged $\ge$18 years; and they could speak Czech and read, write, and understand the questionnaire text in Czech. Participation in the study was voluntary and anonymous for all respondents. Participants completed a self-report test battery comprised of a demographic questionnaire, Diabetes Distress Scale, and a shortened PID-5. The data were collected in a written or online format. Trained administrators (i.e., psychologists, educators, and tutors) conducted the data collection. The respondents were advised to get fully acquainted with all instructions before completing the test, and an administrator was also present to provide help if needed. The exact instructions were given to the participants who completed the online questionnaires. Access to the online form was shared during the educational camps and with a group of patients with DM on social media platforms.

This study’s data were analyzed in a structural equation modeling (SEM) framework, with a set of confirmatory factor analyses (CFAs) for evaluation of the DDS psychometric properties and with multiple regressions within SEM for assessing associations between the DDS and PID-5. All analyses were conducted in R [20]. To verify the factor structure of the DDS, we used a CFA to test the hierarchical model with four first-order and one second-order factor, which represented a general factor of the DDS. Next, the reliability of subscales was verified with Cronbach’s alpha and McDonald’s omega. The results of those analyses are described below. The regression analysis results were bootstrapped with 10,000 nonparametric bootstraps to obtain confidence intervals, estimates, and p-values of natural effects. Additionally, in this analysis, the PID-5 dimensions were inserted into the analysis as observed variables (as arithmetic means) to make the model parsimonious and less complex using item parceling (subset-item-parcel approach) [21]. This approach was necessary because the full PID-5 model, which includes five latent second-order factors, 25 latent first-order factors, and 160 indicators, was not feasible with our sample size. Two separate analyses (one with five broad domains of PID-5 as predictors and the second with all 25 facets of the PID-5 as predictors) were performed. Due to the vast amount of regression analyses, we adjusted the p-values with the Holm-Bonferroni method to reduce the risk of type I errors caused by multiple comparisons. To compare differences in PID-5 domains and facets between groups based on diabetes type (T1 vs T2) and DD level (overall score in DDS $>$2 vs DDS 2), Mann-Whitney tests with Holm-Bonferroni correction were used.

Of the (N = 358) respondents, 219 (61.2%) were diagnosed with T1D, 129 (36%) with T2D, seven (2%) with another diabetes type (MODY, a pancreas disorder), and two (0.5%) were diagnosed with gestational diabetes; 56.2% (n = 201) of the participants were female, the mean (M) age of the sample was 42.33 (SD [standard deviation] = 14.33, Me [median] = 42, IQR [interquartile range] = 26.75) years. The median body mass index (BMI) was 28.1 for men and 26.89 for women. The mean duration of DM was 12.9 (SD = 10.4, Me = 10, IQR = 14) years; 36.3% (n = 130) of the sample had experienced at least one chronic diabetes complication, while 48 participants (13.4%) reported more than one complication. Per cases of T1D, most of the respondents had been using an insulin pump (n = 90; 25.1%). Overall, 11.2% (n = 40) were treated with intensive insulin therapy (besides those with insulin pump) and 85 participants (23.7%) were treated with oral antidiabetics. The rest of the sample reported treatment with a combination of oral antidiabetics and injections (i.e., insulin once a day or incretin mimetics); 17.3% (n = 62) of the patients reported a history of psychiatric treatment and 9.8% (n = 35) were currently taking psychopharmaceuticals. In 28.8% (n = 103) of cases, the respondents were single, 53.6% (n = 191) were married, and 17.3% (n = 62) were divorced or widowed. Most respondents had completed secondary education (n = 183; 51.1%) and 133 (37.2%) had completed higher education or university. The mean (M) level of DD in the sample was 2.18 (SD = 0.95, Me = 1.82).

To analyze the associations between DDS and PID-5, we used the same DDS measurement model (i.e., four latent factors with the two correlated residuals), while the PID-5 factors were inserted into the structural model as the manifest variables. We performed two separate SEM analyses.

We analyzed the associations between the five PID-5 domains and four DDS factors in the first model. In the second model, we used all 25 PID-5 facets as predictors and one general second-order factor of DDS as an outcome (due to many regression analyses). Both models showed satisfactory fit indices, X² (176) = 347.415, p 0.001, RMSEA = 0.062 [90% CI: 0.052, 0.072], SRMR = 0.046, CFI = 0.947, TLI = 0.933; and X² (561) = 4256.204, p 0.001, RMSEA = 0.046 [90% CI: 0.040, 0.052], SRMR = 0.040, CFI = 0.914, TLI = 0.917, respectively. Complete analyses with 25 facets and four subscales of the DDS are shown in the Appendix (see Appendix Table 7).

Our findings (see Tables 3,4) suggest a strong association between the PID-5 domains of NEF and emotional burden (EB) ($\beta$ = 0.852, p 0.001) and a moderate association between the domains of NEF and regimen distress (RD) ($\beta$ = 0.435, p = 0.006). Even though some other associations showed weak to moderate strength associations (see Fig. 1), they were statistically significant only without Holm-Bonferroni correction and were, therefore, not further investigated.

Fig. 1.

Association between PID-5 domain and DDS subscale scores.

On the general DDS factor-level predicted by 25 PID-5 facets there was no association that was statistically significant after applying Holm-Bonferroni correction. The strongest association existed between anxiousness and the total level of DDS ($\beta$ = 0.371, p = 0.075). There was also a moderate association between depressivity and DDS level ($\beta$ = 0.443, p = 0.0432). The model with 25 facets of PID-5 and four factors of DDS yielded satisfactory fit indices, X² (436) = 716.766, p 0.001, RMSEA = 0.047 [90% CI: 0.041, 0.053], SRMR = 0.032, CFI = 0.927, TLI = 0.906. The only result that remained statistically significant after Holm-Bonferroni correction was between anxiousness and emotional burden (EB) ($\beta$ = 0.559, p 0.001; see Appendix Table 7).

Additionally, we conducted Mann-Whitney tests to explore the differences in the distribution of PTs between T1D and T2D. The sample consisted of 219 people with type 1 diabetes (PWT1D) and 129 PWT2D. Statistically significant differences with small effect sizes were found between PWT1D and PWT2D in their NEF (p 0.001, r = 0.28), Antagonism (p = 0.035, r = 0.14), Disinhibition (p = 0.001, r = 0.20), and Psychoticism (p = 0.001, r = 0.20) per the broad PID-5 domains (see Table 5). PWT1D scored higher in all domains. The differences in the level of facets are shown in the Appendix Table 8.

Following, again using the Mann-Whitney tests, we checked for differences in the groups based on DD level: participants with DD 2 (n = 219) versus participants with elevated DD $>$2 (n = 129). Differences were found among all the broad PID-5 domains with medium effect sizes (all ps 0.001, rs varied between 0.29 and 0.50; see Table 5). In all cases, participants with DD $>$2 scored higher on all domains. Differences among all facets are shown in the Appendix Table 8.

The verification of the psychometric properties of the Czech version of the DDS was the first important step because, to the best of our knowledge, there has been no previous study using the Czech DDS reported. The Czech DDS version showed satisfactory psychometric properties in its factor structure, internal consistency, and measurement in variance between sex, diabetes types, and age. It is, therefore, approved for further clinical and research purposes in the Czech population with DM. Cronbach’s alpha of a 17-item scale was the same as that obtained by authors of the original version (0.93 [1]) and the Polish version [22].

One of the study’s main goals was to identify connections between subjectively-perceived DD and maladaptive PTs using the PID-5. Our findings suggest that the NEF broad domain of the PID-5 was significantly associated with two subscales of the DDS – EB and RD. To our knowledge, there has been no previous research on the DDS using the PID-5. Nevertheless, the AMPD trait model was aligned with the FFM, which was included in several studies [8, 9, 12]. The PID-5’s broad domain of NEF comprises three facets (anxiousness, emotional lability, separation insecurity) which might be linked to the FFM’s neuroticism domain. From this perspective, our results confirmed the findings from a MILES study [9] and a study from Sanatkar et al. [12]. In the former, the tendency to experience negative emotions was negatively connected with treatment demands (i.e., burden of diabetes). Conversely, the second study referred to the level of neuroticism and its relation to the level of DD. In Sanatkar et al.’s [12] study, the level of neuroticism affected more than one component of DD (including physician distress, EB, and RD). In our research, NEF affected all DD components. However only EB and RD were affected significantly.

Our findings were also supported by differences found between groups based on their overall levels of DD, as participants with elevated DD scored higher on pathological PTs. Moreover, previous findings regarding the suitability of the PID-5 for personality assessment in patients with primarily endocrinological illnesses support these findings to some extent, as people with chronic diseases may experience feelings of loss and guilt related to a sense of self-negativity associated with a body that is symbolically damaged by chronic illness [23]. Some studies have supported the association between self-blame and poorer quality of life in PWD, as well as between increased levels of guilt and a tendency to cope through behavioral withdrawal, with negative implications for health outcomes [24].

In the present study, PWT1D reached statistically significantly higher levels of personality pathology among all broad domains of the PID except detachment, and most facets (17 out of 25) except anhedonia, withdrawal, grandiosity, manipulativeness, unusual beliefs and experiences, callousness, restricted affectivity, and risk-taking. T1D patients who score higher compared with T2D are usually those diagnosed with diabetes in childhood; the disease becomes part of their self-concept and affects their self-esteem, which is associated with emotional stability, EB of diabetes, or depressive symptoms [25]. Therefore, a negative self-concept may be associated with DD levels in adulthood. We also discovered that participants with higher DD showed statistically significantly higher levels on all broad domains of the PID-5 and across almost all facets (23 out of 25), except for risk-taking and attention-seeking.

Conversely, higher levels of risk-taking and intimacy avoidance in T2D were consistent with the findings of Riegel et al. [15], where these facets were elevated at the clinical level in a sample of pre-bariatric patients. As T2D is a common comorbidity of obesity, we might have expected some similarities in levels of personality pathology in our sample. Scores of maladaptive PTs were lower in our sample, but this may have been influenced by the fact that the average BMI in our sample was in the overweight range.

In considering the results from a systematic review and meta-analysis of effective interventions for reducing DD [26] where the effect sizes were low, we might answer the question of why a well-educated patient is still non-compliant. Sometimes, more practical tools are insufficient to achieve the desired result. In addition, according to Samadi et al. [27], behavioral interventions or education can increase patients’ self-concept and self-esteem. However, there are still cases in which systematic psychotherapy should be considered. It was proven by using the AMPD in the study of Huprich et al. [14] that receiving therapeutic assessment and guidance to understand how the level of personality functioning and pathological PTs might influence the diabetes management of PWT2D helped to decrease the level of HbA1c.

Understanding the emotion regulation skills associated with NEF according to the PID-5 might be helpful in DD reduction programming, as confirmed by Coccaro et al. [28]. Moreover, it could explain DD’s stability over time as described earlier [3]. Our results also showed that DD is associated with PTs and those with elevated DD scored higher overall on PTs. DD and other negative feelings such as shame, anger, self-blame, or fear have also been linked to a state of damaged ego, creating a transference-countertransference dynamic in which elements personal to both client and therapist interact, which may be helpful in recognizing relationships with healthcare providers when behavioral interventions are not working [16, 23]. Using the AMPD and giving feedback to the PWD can help them understand themselves and bring therapeutic effects, as shown by Huprich et al. [14].

The present findings should be viewed in the context of certain limitations, which may inspire further research. From the perspective of the overall level of personality psychopathology within DM patients, the battery of tests should be enriched with specific instruments measuring the AMPD’s criterion A (i.e., the level of personality functioning), as well. However, each added inventory makes the battery of tests more demanding and time-consuming. This could partially be solved by using a shorter PID-5 form, which has proved sufficient in capturing AMPD criterion B (i.e., PTs) per the level of broad domains [29], and by adding a screening tool focused on the level of personality functioning (LPFS), like the LPFS-Brief Form [29]. It also seems essential to consider the subjectivity of the participants’ statements, as both tools used were self-report measures. However, the presented study was exploratory and pioneering regarding DD in the Czech Republic. Future investigations might use larger samples and batteries of tests that focus on treatment adherence outcomes, patient compliance, and assessment for symptoms of depression and anxiety, as elevated scores on the DDS are usually associated with being depressed and having poorer self-care and a lower quality of life [2]. Nevertheless, in line with Ehrenthal et al. [30], our results support the finding that the dimensional models of personality pathology and its screenings can help detect specific groups of “difficult to treat” patients who might benefit from psychological support in the diabetes treatment process.