Diabetes Distress, Depressive Symptoms and Type 2 Diabetes Management and Glycemic Control: The Mediating Role of Self-Ecacy

Background Diabetes distress has been distinguished from depressive symptoms in the linkages to type 2 diabetes management and glycemic control. There are few evidences in rural China untangled the underlying pathways among them. The aim of our study was to examine whether self-ecacy reduced the detrimental effects of psychological variables on diabetes outcomes such as self-care behaviors, glycemic control and health-related quality of life (HRQoL). Methods 1574 adults diagnosed with type 2 diabetes from 31 rural clinics in China were assessed for psychological variables, self-management and HRQoL. Hemoglobin A1c (HbA1c) data at 6 months after the survey were extracted from electronic medical records. Hierarchical regression model examined independent effects of psychological variables, self-ecacy on diabetes outcomes. KHB method and bootstrapping condence intervals were employed to test the mediating effect of self-ecacy. Results Hierarchical regression analysis showed both diabetes distress and depressive symptoms were signicantly related to diet and HRQoL, but not related to glycemic control and medication adherence. Depressive symptoms, but not diabetes distress, was directly associated with physical activity. Mediation analysis demonstrated self-ecacy mediated the relationships of both diabetes distress and depressive symptoms on self-care behaviors, glycemic control and HRQoL. Self-ecacy may contribute to better diabetes outcome by ameliorating the effects of diabetes distress and depressive symptoms. Patients with psychological conditions may need tailored intervention that enhance self-ecacy to improve management. symptoms on diabetes self-management, glycemic control and overall quality of life in the same study among adults with type 2 diabetes from primary care settings in rural China. Based on the conceptual model, our analyses reveal three main ndings. First, both diabetes distress and depressive symptoms are independently associated with diet and HRQoL, but not signicant direct associated with HbA1c and medication adherence. Physical activity is signicant affected by depressive symptoms, but not by diabetes distress. Second, self-ecacy was signicant related to self-care behaviors, HbA1c and HRQoL. Third, self-ecacy, as an independent mediator, signicantly mediates the effect of both diabetes distress and depressive symptoms on diabetes self-management, glycemic control and HRQoL. The ndings help elucidate key underlying mechanisms and pathways linking psychological factors and behavioral health with diabetes outcomes.


Introduction
Diabetes is a complex and burdensome disease that requires the person with diabetes to make numerous daily decisions regarding food, physical activity, and medications [1]. Comorbid diabetes distress and depressive symptoms are prevalent emotional states in people with type 2 diabetes. Psychological problems could arise from stress of management diabetes and worrying about hypoglycemia and the complications of diabetes [2]. Diabetes distress is de ned as a patients' concern about disease management, support, emotional burden and access to care [3], which views as part of diabetes spectrum and not a separate clinical psychopathology [4]. Studies of meta-analysis suggested overall estimated prevalence of diabetes distress among type 2 diabetes patients is 36% [5], but co-occurrence of depressive symptoms is about 20% [6]. Some studies conducted both depressive symptoms and diabetes distress in type 2 diabetes individuals were associated with medication non-adherence and poor self-management [7], worse glycemic control [8] and lower health-related quality of life [9].
However, evidence also suggested diabetes distress had been distinguished from depressive symptoms, in the linkages to self-care activities and glycemic control. Two cross-sectional studies indicated that diabetes distress but not depressive symptoms, was associated with self-management and glycemic control [3,10]. However, another longitudinal study with 253 primary care type 2 diabetes patients suggested only depressive symptoms but not diabetes distress was associated with future diet, physical activity, and glucose testing self-care behaviors [11]. Clearly, the effect of diabetes distress and depressive symptoms on diabetes self-care and glycemic control are well studied but mixed results. Thus, the potential pathways among psychological factors and diabetes outcomes need to be studied further.
Self-e cacy, a critical concept of social cognitive theory (SCT), is related to one's belief and con dence in the ability to successfully perform a task and execute skills effectively [12]. A high degree of self-e cacy is essential for carrying out self-care behaviors. Considerable evidence supported self-e cacy had direct effect of improving diabetes self-management, medication adherence, metabolic outcomes and HRQoL [13,14]. A meta-analysis of 775 studies concluded that self-e cacy was all consistence with all adherence behaviors (e.g. dietary, physical activity) and diet adherence was the most signi cant predictors of HbA1c [15]. Interestingly, several recently research highlighted self-e cacy as a potential mediator in the links between depression or depressive symptoms and diabetes management [16][17][18]. One study with 371 type 2 diabetes patients in Malaysia found both diabetes distress and depressive symptoms had indirect effect on self-care activities via self-e cacy [19]. However, little is known about how psychological variables (e.g. depressive symptoms glycemic control, obesity and quality of life [22]. Another study indicated diabetes distress and self-e cacy were related to HRQoL [23]. Poorer quality of life and depressive symptoms also had been associated with increased out-of-pocket payment for health service utilization and more likely to be impoverished [24]. Understanding how to distribute psychological care to diabetes patients with diabetes distress or depressive symptoms has important implications. Thus, further studies are deserved about the effect of diabetes distress and depressive symptoms on HRQoL and whether self-e cacy alleviate the two variables impact on HRQoL. Diabetes distress, depressive symptoms and self-e cacy are correlated. Their relationships with diabetes management and health outcomes may not be clearly observed unless they are examined in the same statistical model. Yet, to date, no studies to our knowledge have examined the potential pathways and interactions among psychological factors (both diabetes distress and depressive symptoms), behavioral variables (self-e cacy), diabetes management (e.g., healthy diet, physical activity, medication adherence), glycemic control and HRQoL in the same study, especially for primary care type 2 patients in rural China.
The current study examines whether diabetes distress and/or depressive symptoms is directly associated with diabetes outcomes, such as diabetes management, sequent 6-month glycemic control and HRQoL among Chinese primary care adults with type 2 diabetes. In addition, we examine whether selfe cacy mediates both psychological factors and behavioral predictors in relation to diabetes outcomes. Based on social cognitive theory and aforementioned studies, we present a conceptual framework (Fig. 1) and employ regression-based mediation analysis to testify our hypothesis. We hypothesize that: (1) high level of diabetes distress and depressive symptoms would directly decrease diabetes management and health outcomes. (2)self-e cacy affected by diabetes distress and depressive symptoms, which, in turn, determines diabetes self-care behaviors, subsequent diabetes glycemic control and HRQoL.

Study sample
Observational follow-up study design was conducted in 5 rural clinics in Tongzhou district, 10 rural clinics in Dafeng district and 16 rural clinics in Haimen district in Jiangsu province, eastern China. All the clinics performed pilot reforms of family doctor signed contract service with type 2 diabetes patients. The inclusion criteria of clinics included: (1) 2-4 family doctors and/or nurses; (2) the numbers of type 2 diabetes patients signed the contract were more than 50; (3) the clinics agreed to participate in the study for at least 2 years. The inclusion criteria of participants included: (1) had a clinical diagnosis of type 2 diabetes ≥ 12 months and age ≥ 30 years; (2) patients agreed to participated in the full two-year study. Exclusion criteria included severe diabetes complications or functional de cits (e.g., blindness, dementia, psychosis). Eventually, 1574 adults with type 2 diabetes were recruited by cluster sampling from 31 rural clinics. Patients received a telephone call from the family doctors to explain the study purpose and eligible patients were invited and dated a speci c time to complete face-to-face assessment. This study was approved by the ethics committees of Nantong University and informed consent was obtained from each participant. The data were collected between February and March 2019 .

Predictors
The 17-item Diabetes distress Scale 17 (diabetes DDS-17) evaluated distress over the past month that was rated from 1(not a problem) to 6 (a very serious problem). Items responses were averaged to generate a total score, with higher scores indicating greater diabetes distress. A mean item score ≥ 2 (moderate distress) is considered a level of distress worthy of clinical attention [25]. A Chinese version of the diabetes DDS-17 was found to have good psychometric properties for internal consistency(α = 0.90) and test-retest reliability coe cient (α = 0.74) [22].
Depressive symptoms was measured by the 10-item Center for Epidemiologic Studies Depression Scale (CESD-10) short form [26]. CESD-10 assessed the extent to which individuals experienced depressive symptoms during the prior week on a 4-point Likert scale ranging from 1 = rarely or none of the time (0-1 day) to 4 = most of the time (5-7days). The CESD-10 was high reliability and validity in Chinese middle age and elderly population (α = 0.82) [27]. The CESD-10 scores ranged from 1 to 30, with higher scores indicating more severe depressive symptoms. Individuals who had scores 10 or above were classi ed as having depressive symptoms.

Mediating variables
Diabetes self-e cacy was measured using Lorig's 8-item diabetes self-e cacy scale (α = 0.85) and test-retest validity of 0.80 [28]. Participants were requested to indicate their con dence in completing speci c self-management tasks on a 5-point Likert scale, ranging from 1 (not at all con dent) to 5 (totally con dent).
The scores of the scale is the mean of 8 items, with higher scores indicating greater self-e cacy.

Dependent variables
Diabetes self-management was assessed using the 11-item revised Summary of Diabetes Self-Care Activities (SDSCA) [29]. The SDSCA measures frequencies of speci c self-care activities during the last week for six aspects of the diabetes regimen: general diet, speci c diet, exercise, foot care, blood-glucose testing, taking medication and smoking. Respondents indicated the number of days during the past week that they followed their diets and physical activity regimens. All scale scores range from 0 to 7, with high scores suggesting better self-management. The SDSCA has shown high reliability in measuring self-care behaviors of Chinese diabetes (α = 0.82) [30].
Adherence to diabetes medication was assessed using the Chinese version 8-item Morisky Medication Adherence Scale (α = 0.65) (MMAS-8) [31]. Response choices are yes/no for item 1-7 and 5-point Likert response for last item. The total score of MMAS-8 ranged from 0-8, with higher scores indicating of greater medication adherence. The scores are then categorized as low (less than 6), medium (6 to less than 8) and high (8) medication adherence subsequently.
HRQoL was assessed using EQ-5D-3L scale self-report questionnaires and has been validated Chinese population [32]. EQ-5D questionnaire provides utility values for 243 health states described using a system comprising ve dimensions (e.g.mobility, self-care, usual activities, pain/discomfort, and anxiety/depression) and three levels (1 = no problems, 2 = some/moderate problems, 3 = extreme problems) for each dimensions. Chinese Time trade-off value sets for EQ-5D was used to calculate EQ-5D index scores [33].
Hemoglobin A1c (HbA1c) data at 6 months after the survey were extracted from electronic medical records. HbA1c values less than 7% (53 mmol/mol) was regarded as optimal glycemic control.

Data analysis
We rst conducted descriptive analyses of all variables (Table1), and examined the correlations among predictors, mediator of self-e cacy, and diabetes outcomes ( Table 2), and then applied hierarchical ordinary least squares (OLS) linear regression models to explore the mediator self-e cacy on diabetes selfcare behaviors, glycemic control and HRQoL (Table 3a, 3b). KHB method was lastly used to estimate the total effect of the predictor variables diabetes distress and depressive symptoms on the outcome of diabetes management [34]. The total effect was divided into a direct effect of the predictor variables and an indirect effect of the mediator self-e cacy (Fig. 1).   Note: *P < 0.05; † P < 0.01; ‡ P < 0.001.
Hierarchical regression models introduced variables according to our proposed conceptual model. In the models, psychological variables (diabetes distress and depressive symptoms) were predictor variables, diabetes outcomes (self-care behaviors, HbA1c and HRQoL) were dependent variables, while self-e cacy was the mediator. Age, gender, education, marital status, diabetes duration, numbers of chronic diseases were covariates with impact on predictors, the mediator and diabetes outcomes. They were controlled for in all the models. while the predictors and the mediator are added in the models (path b). Model 1 is used to test whether the mediator self-e cacy is affected by psychological variables (Fig. 1, path a). Models 2 and Models 3 are used to investigate the total effects (path c) of psychological variables on diabetes outcomes. From Models 2, Models 3 to Models 4 check whether the effect of psychological variables act through the mediator self-e cacy (indirect effect) or are unexplained by the mediator (direct effect). For mediation occurs, when the mediator is added (Models 4), the coe cient of mediator should be signi cant (path b), the coe cients of predictor variables on diabetes outcomes should diminish or disappear (direct effect, path c'). The Hierarchical estimation approach allows us to test how the impact of psychological variables may change once the mediator is applied for diabetes outcomes variables in the models.
In order to check the robustness of mediation effects, the bootstrapping mediation approach is used, with 95% bias-corrected con dence intervals (CIs) based on a bootstrapping sample with 5000 [35]. If the con dence interval of indirect effect does not contain zero, it indicates that the mediating role of self-e cacy is in fact different from zero. Analyses were performed using Stata software version 14. All statistical tests were two-tailed, with α set to 0.05.

Sample Characteristics
The sample was more female (67.60%) and most were married (82.97%), with a mean age of 66.33 (SD 8.97) years (

Correlations of psychological variables, behavioral variables and diabetes outcomes
Correlations analyses conducted on between psychological variables, behavioral variables and diabetes outcomes ( Table 2). Diabetes distress and depressive symptoms were signi cantly correlated (r = 0.26, P < 0.001). Both DDS17 scores and depressive symptoms were signi cantly negative correlated with selfe cacy, physical activities, medication adherence and EQ-5D. However, psychological and behavioral variables were no signi cant correlated with HbA1c.
Self-e cacy was signi cantly correlated with all psychological variables and diabetes outcomes (all P < 0.001). Hence, self-e cacy was identi ed as a potential mediator.

Hierarchical regression analysis
Hierarchical regression analyses were conducted to test the effects of psychological variables on diabetes outcomes. As shown in Table 3, Models 1 and Models 2 supported Hypothesis 1 that both diabetes distress and depressive symptoms were negatively associated with physical activity and HRQoL, but not related to HbA1c (path c in Fig. 1). However, only diabetes distress but not depressive symptoms was signi cantly and negatively affected diet and medication adherence, suggesting that the total effect of diabetes distress on healthy diet and medication adherence distinguished from depressive symptoms.
Fortunately, self-e cacy was signi cantly affected by diabetes distress and depressive symptoms (Model 1). While predictors and self-e cacy were included in the equations (Models 4), self-e cacy was signi cantly and positively associated with self-care behaviors and HRQoL, negatively associated with HbA1c, which indicated self-e cacy was a mediator, providing support for Hypothesis 2.
The changes in the estimated coe cients from Models 2 to Models 4 (path b and c' in Fig. 1), suggest the self-e cacy could compensate for the impact of diabetes distress on diabetes outcomes. The coe cients of diabetes distress in relationship to diet (from β=-0.69, P < 0.001 to β=-0.61, P < 0.001) and HRQoL (from β=-0.06, P < 0.001 to β=-0.02, P < 0.01) became smaller but still signi cant, which indicated self-e cacy partially medicated the relationships among them. Diabetes distress failed to predict physical activity and medication adherence after the self-e cacy was added (Models 4), suggesting self-e cacy played completely mediating role on the associations between depressive symptoms and the two self-care behaviors (physical activity, medication adherence). Likewise, the results in Models 4 showed self-e cacy had partial mediating effect of depressive symptoms in relation to physical activity (from β=-0.68, P < 0.001 to β=-0.37, P < 0.01) and HRQoL (from β=-0.13, P < 0.001 to β=-0.11, P < 0.01), only mediating impact of depressive symptoms on diet, medication adherence. However unexpectedly, both diabetes distress and depressive symptoms only had indirect effect on HbA1c through self-e cacy.

Mediation Analysis
Results of KHB mediation analysis and bootstrapping con dence intervals testing indirect effect of psychological variables on diabetes outcomes are presented in Table 4. KHB regression results revealed self-e cacy signi cantly mediated the effect of both diabetes distress and depressive symptoms in relation to self-care behaviors, HbA1c and HRQoL. The 95% bias corrected bootstrap CIs for indirect effects of psychological variables on diabetes outcomes through self-e cacy all excluded zero (Table 4), consisting with the KHB results and hierarchical regression results, which showed self-e cacy had a signi cant mediating role on diabetes management. The estimated coe cients of indirect effects indicated higher diabetes distress and severe depressive symptoms among type 2 diabetes patients were associated with lower self-e cacy, which in turn were associated with unhealthier diet, less exercise, poorer medication adherence, worse glycemic control and lower HRQoL. In addition, patients with higher diabetes distress had a signi cant direct effect for lower diet behavior (direct effect: β=-0.61, P < 0.001) and HRQoL (direct effect: β=-0.02, P < 0.001), but no signi cant direct effect for exercise, medication adherence and HbA1c. Those with depressive symptoms displayed signi cantly poor physical activities (direct effect: β=-0.37, P < 0.05) and worse HRQoL (direct effect: β=-0.01, P < 0.001). 95%CI= 95% bias-corrected bootstrapping con dence interval.

Discussions
Our study is the rst to our knowledge to examine the mediating role of self-e cacy in the relationships of both diabetes distress and depressive symptoms on diabetes self-management, glycemic control and overall quality of life in the same study among adults with type 2 diabetes from primary care settings in rural China. Based on the conceptual model, our analyses reveal three main ndings. First, both diabetes distress and depressive symptoms are independently associated with diet and HRQoL, but not signi cant direct associated with HbA1c and medication adherence. Physical activity is signi cant affected by depressive symptoms, but not by diabetes distress. Second, self-e cacy was signi cant related to self-care behaviors, HbA1c and HRQoL. Third, self-e cacy, as an independent mediator, signi cantly mediates the effect of both diabetes distress and depressive symptoms on diabetes self-management, glycemic control and HRQoL. The ndings help elucidate key underlying mechanisms and pathways linking psychological factors and behavioral health with diabetes outcomes.
Diabetes self-management is crucial for metabolic control and quality of life. Consistent with previous studies [14], diabetes distress but not depressive symptoms had direct effect of healthy eating behavior, patients with depressive symptoms but not diabetes distress is related to poorer physical activities. Yet, it was not able to test whether self-e cacy mediated the relationships between the two affective variables and self-care behaviors. Devarajooh et al reported evidence to support self-e cacy as a mediator of the relationship between depressive symptoms, diabetes distress and the total score of self-care activities among type 2 diabetes using the structural equation model [19]. Neither the direct effect of both two variables with diabetes self-management nor the links between speci c self-care activities (e.g. diet, physical activity) and distress or depression was found. Our new ndings con rmed both diabetes distress and depressive symptoms had signi cant indirect effect of diet, physical activity through self-e cacy. Adherence to the prescribed medical regimen is an important goal in the management of diabetes and prevented complications. Gonzales et.al found diabetes distress, but not depression severity, was signi cantly related to medication adherence and mediation analyses demonstrated a signi cant indirect effect for depressive symptoms and medication adherence through self-e cacy and perceived control [16]. In contrast, the current ndings found self-e cacy is convinced predictive of medication adherence and signi cant fully mediated the effect of both diabetes distress and depressive symptoms in related to medication adherence. Thus, the main contribution of our study is explaining possible pathway that how diabetes distress and depressive symptoms impact on self-care behaviors via self-e cacy.
In lines with previous reports [18,36], both diabetes distress and depressive symptoms were not signi cantly direct linked to glycemic control in the study. Different types of psychological variables measures (such as PHQ-9, CESD-10, diabetes DDS-17, PAID-5) may partly contribute to no signi cant association between depressive symptoms and glycemic control [36]. Another possible explanation was lower prevalence of diabetes distress (12.46%) for patients in primary care settings couldn't detect the glycemic control difference in the cross-sectional study. Finally, potential mechanisms may explain the linkages between psychological variables and diabetes outcomes. For example, depression may decrease motivation to complete and maintain self-management behaviors that are protective against worsening of metabolic control [37]. Self-e cacy had been shown to signi cantly in uence diabetes self-care behaviors, which in turn affected HbA1c [14,30]. Fortunately, our ndings also suggest self-e cacy plays signi cant mediating roles of psychological variables (diabetes distress and depressive symptoms) in relation to HbA1c.
Both diabetes distress and depressive symptoms were directly related to EQ-5D index scores in current study, similar to previous studies [9]. A longitudinal study of quality of life trajectory patterns among 455 type 2 diabetes from ve clinics in Taiwan had shown diabetes distress and HbA1c level were important for determining participants in the QoL trajectory development [31]. Our study also found poor psychological conditions decreased adherence to self-care behaviors, especially for physical activities. A Bangladesh out-patient diabetes study had shown no-adherence to diet and exercise self-care practices were signi cant associated with lower quality of life [38]. A meta-analysis of 20 comparisons across 1892 subjects supported that self-management interventions improved health-related quality of life [39]. Furthermore, new evidence was added to understand how both diabetes distress and depressive symptoms impact on HRQoL through self-e cacy. In current study we found self-e cacy had signi cant mediating effects of psychological factors on HRQoL among type 2 diabetes patients. Self-e cacy enabled to maintain healthy behaviors, improve diabetes outcomes and promote positive well-beings. A recent meta-analysis study of ten trails with 1308 participants found self-e cacy-focused education in adults with diabetes probably reduced A1C, enhanced self-e cacy, regulated self-management behaviors and improved the QOL [40].

Strengths And Limitations
The present study has multiple strengths. The study provides a comprehensive framework to elucidate the relationships among psycho-social factors, behavioral factors and diabetes outcome based on social cognitive theory, and empirically explored multiple factors and pathways link these variables to one another. This is the rst study to examine the mediation role of self-e cacy in the relationship among these variables in Chinese adults with type 2 diabetes and verify the indirect effect by KHB method. The KHB method decomposes effect of both discrete and continuous variables, applies average partial effects and provided analytically derived statistical tests. Second strength of this study is a relatively large sample size with high response rate in rural adults with type 2 diabetes. Finally, well-validated and speci c instruments were used to measure multiple psychological and behavioral variables.
Several limitations should be kept in mind. First, cross-sectional design data prevents conclusions about the causal mediation and the ability to test the conceptual model among psychological variables, behavioral variables to diabetes management. Although we have described direct and indirect effects of psychological factors on diabetes outcomes, this causal mediation analysis is a convention of the statistical approach and should not be taken to imply that these cross-sectional analyses can address questions of directionality. Second, this study used a cluster sampling method, which added bias and decreased the representativeness of the sample. The study sample was relatively less educated and rural residents. Thus, the ndings may not generalize to more diverse populations. Third, although we used well-validated and speci c instruments, self-report measures may have introduced recall and social desirability bias. Finally, the prevalence of diabetes distress in this study is very low (12.46%) and relatively higher optimal glycemic control of the participants (43.18%). Thus, limiting direct effect of diabetes distress on glycemic control detected. Future studies should consider employing longitudinal design with patients from primary and tertiary hospitals and estimate changes in psychological conditions in relation to changes in diabetes management, glycemic control and HRQoL.
Further studies also could test whether interventions to increase self-e cacy from existing primary health care to improve glycemic control and HRQoL among patients who experience high levels of psychological distress.

Conclusions
Self-e cacy may contribute to better diabetes self-care behaviors, glycemic control and HRQoL both directly and by mediating against the negative effects of diabetes distress and depressive symptoms. Diabetes distress and depressive symptoms had both direct and indirect effect on HRQoL, but only indirect effect of glycemic control through self-e cacy. The ndings suggest that increased self-e cacy contribute to better diabetes management, metabolic control and HRQoL, particularly among individuals with elevated diabetes distress and depressive symptoms. Healthcare professionals should take consideration of psychological factors across the life span among diabetes adults, make assessment of affective factors, and addressed psychological problems upon identi cation. Mental health treatment in the context of primary care should be strengthened. Integrating behavioral and psychological interventions at primary care level could be prioritized for the right and most needy adults with type 2 diabetes to improve their diabetes management and health outcomes.

Declarations
Ethics approval and consent to participate This study was approved by the ethics committees of Nantong University and informed consent was obtained from each participant. Study conceptual model for proposed the relationships between psychological variables, behavior variables and diabetes outcomes. Dotted lines show a potential pathway mediated by self-e cacy.