The Science of Diabetes Self-Management and Care2023, Vol. 49(6) 438 –448© The Author(s) 2023Article reuse guidelines:sagepub.com/journals-permissionsDOI: 10.1177/26350106231205029journals.sagepub.com/home/tde
Abstract
Purpose: The purposes of this study were to (1) examine the relationships between fatigue, its influencing factors, and diabetes self-management and (2) test the mediation effects of fatigue on the link between the influencing factors and diabetes self-management in adults with type 2 diabetes.
Methods: This cross-sectional, correlational study was guided by the theory of unpleasant symptoms. Data were collected using structured questionnaires. Fatigue was measured by the Fatigue Symptom Inventory and the Multidimensional Fatigue Inventory. Diabetes self-management was measured by the Summary of Diabetes Self-Care Activities. From March to July 2021, a convenience sample of 150 participants was recruited from 2 diabetes outpatient clinics of a regional hospital in Taiwan. Data were analyzed using structural equation modeling.
Results: A more recent diagnosis of diabetes, more depressive symptoms, and lower sleep quality were related to higher fatigue. Higher fatigue correlated with less performance in diabetes self-management. Fatigue mediated the relationship between depressive symptoms, sleep quality, and diabetes self-management.
Conclusions: Fatigue had a mediating effect on the link between psychological influencing factors and diabetes self-management. Future development of fatigue interventions integrating depressive symptoms and sleep management will likely increase the performance of diabetes self-management and improve the health outcomes in adults with type 2 diabetes. The study tested the theory of unpleasant symptoms using empirical data and will assist in building theory-guided fatigue interventions to improve diabetes self-management in people with type 2 diabetes.
Type 2 diabetes mellitus (T2DM), a serious illness that requires consistent daily diabetes self-management (DSM) to delay the onset of disabling complications,1 produces multiple symptoms that impact patient-reported outcomes. Most people with T2DM (95%-97%) experience an average of 4 to 14 symptoms weekly.2-4 Among the symptoms experienced in people with T2DM, fatigue is regularly ranked as one of the most common and bothersome symptoms that affect people’s daily DSM and quality of life.2,5
Fatigue, a subjective feeling of tiredness that impacts the individual’s ability to perform mental and physical tasks,6 is reported to be prevalent among 60% to 68% of adults with diabetes.2,7,8 People with T2DM experience fatigue at a higher intensity compared to the general population.9 Fatigue is related to multiple influencing factors, such as physiological factors (eg, longer diabetes duration,10 having more comorbidities,11 and higher levels of A1C12 ), psychological factors (eg, more depressive symptoms,8,13,14 higher diabetes distress,14,15 and poor sleep quality16 ), and situational factors (e.g., less social support12 and a lower education level10 ). Importantly, fatigue significantly hinders DSM behaviors such as eating healthy meals and doing physical activity.16,17 However, most studies on fatigue in T2DM were not theory-guided, resulting in fragmented knowledge regarding fatigue and limited information on fatigue intervention development for adults with T2DM.5 Specifically, to ensure further intervention development, understanding the extent of all fatigue dimensions, what factors predict fatigue, how fatigue impacts DSM, and the underlying mechanisms of fatigue, its influencing factors, and DSM is crucial. Yet to the best of our knowledge, most studies have examined only 1 of these aspects of fatigue per study (eg, fatigue intensity), and no study has examined whether fatigue plays a mediating role in the relations between the influencing factors and DSM in people with T2DM.
Given that current knowledge on fatigue and its potential mediating effect on the relations between the influencing factors and DSM in adults with T2DM has not been investigated comprehensively, the purposes of this study were to (1) examine the effects of fatigue and its influencing factors on DSM and (2) test the mediation effects of fatigue on the relationship between the influencing factors and DSM in adults with T2DM using an adequate theory.
To guide this study, the authors adapted the theory of unpleasant symptoms (TOUS) based on an integrated literature review5 to form a conceptual model of fatigue in adults with T2DM (Figure 1). The concept of fatigue consists of intensity, timing, distress, and quality domains, shown in the center of Figure 1. Depicted on the left side of Figure 1 are influencing factors related to fatigue, including physiologic factors (ie, diabetes duration, comorbidities, and A1C), psychological factors (ie, depressive symptoms, diabetes distress, and poor sleep quality), and situational factors (ie, social support and education). The primary outcome is DSM, which is depicted on the right side of the conceptual model depicted in Figure 1.
This was a cross-sectional, correlational study using convenience sampling. The study was approved by the Institutional Review Board (IRB) of a university and the IRB at the regional hospital where the data collection took place.
A sample of 150 participants was recruited from 2 diabetes outpatient clinics of a regional hospital in northern Taiwan. The regional hospital was located in Hsinchu City and had about 500 medical-surgical beds with about 30 to 70 outpatient visits to the diabetes outpatient clinic per session. People who were diagnosed with T2DM, were age 20 or above, could speak or read in Mandarin or Taiwanese, and were willing to participate and sign the informed consent were included in the study. People who were pregnant, had a fever, were receiving treatment for cancer, had chronic fatigue syndrome, or were diagnosed with fibromyalgia were excluded.
Data were collected from March 2021 to July 2021. A nurse researcher with 3 years of experience as a registered nurse in a diabetes ward in a medical center in Taiwan approached potential participants at the diabetes outpatient clinics in person. After obtaining written informed consent from the participant, the nurse researcher collected data using Qualtrics survey software (Qualtrics, Provo, UT) on a tablet by reading the survey to the participants and recording their responses. The most recent A1C data (ie, within 3 months) and body weight and height in the medical records were recorded in the survey. Participants received a small incentive in the form of a gift card after completing the survey.
This study used the Chinese versions of structured questionnaires, which includes a demographic and clinical characteristic form, Center for Epidemiological Studies Depression Scale (CES-D), Diabetes Distress Scale (DDS), Pittsburgh Sleep Quality Index (PSQI), Multidimensional Scale of Perceived Social Support (MSPSS), Fatigue Symptom Inventory (FSI), Multidimensional Fatigue Inventory (MFI), and Summary of Diabetes Self-Care Activities (SDSCA).
Demographic and clinical characteristics. Participants self-reported their demographic information (ie, age, gender, ethnicity, marital status, and education). Clinical information (ie, body mass index [BMI], diabetes duration, diabetes treatment, number of comorbid conditions, the Charlson comorbidity index [CCI] score, and the most recent A1C data within the past 3 months) were recorded from the medical records.
Depressive symptoms. The CES-D is a 20-item scale that measures the frequency of depressive symptoms for the past 7 days.18 Each item is measured on a 4-point scale (0 = rarely or none of the time, 3 = all of the time), with the possible total sum score ranging from 0 to 60. Higher scores indicate more frequent depressive symptoms.18 The Chinese version of CES-D had good convergent validity.19 The Cronbach’s α for the current study was .92.
Diabetes distress. The 17-item DDS measures 4 aspects of diabetes-related distress: emotional burden, physicianrelated distress, regimen-related burden, and diabetes-related interpersonal distress. The items are measured on a 6-point scale (1 = not a problem, 6 = a serious problem) with a possible total score range of 17 to 102; a higher score indicates higher levels of diabetes distress.20 A total diabetes distress score is calculated, and the mean of the total score is used.21 The Chinese version of the DDS demonstrated good convergent validity.22 The Cronbach’s α was .83 for the present study.
Sleep quality. Sleep quality was measured by the 19-item PSQI. The sum of the PSQI components is a global sleep quality score, with a possible range of 0 to 21. Higher scores indicate worse sleep quality.23 The Chinese version of the PSQI has been tested in community-dwelling adults in Taiwan, with Cronbach’s αs ranging from .71 to .83 and good construct validity.24 The Cronbach’s α was .71 for the present study.
Social support. The MSPSS is a 12-item scale measuring the support from family, friends, and significant others. The items are measured on a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree), with higher total scores representing higher perceived social support.25 The mean score of the total scale is calculated.25 The MSPSS has been used in Taiwanese adults with chronic illness, demonstrating good reliability with a Cronbach’s α of .89 and good structural validity.26,27 For the present study, the Cronbach’s α was .93.
Fatigue. The 14-item FSI includes 4 aspects of fatigue: fatigue intensity, interference, duration, and pattern. Items are measured on an 11-point Likert scale (0 = no fatigue at all, 10 = extreme fatigue), except for Item 12 (measuring how many days in the past week the participant felt fatigued) and Item 14 (evaluating fatigue patterns using 4 categories). A higher score on the FSI indicates a higher level of fatigue.28 The Chinese version of the FSI has good convergent and divergent validity.29 The Cronbach’s α for the present study was .91.
Fatigue quality was measured by the 20-item MFI. This questionnaire evaluates fatigue with 5 dimensions (ie, general fatigue, physical fatigue, mental fatigue, reduced motivation, and reduced activity) using a Likert-type scale ranging from 1 (strongly agree) to 5 (strongly disagree). A total MFI score is calculated. The possible range for the sum of the MFI items ranges from 20 to 100, with higher scores representing more fatigue.30 The Cronbach’s αs of the Chinese version of the MFI ranged from .73 to .92 for people in Taiwan.31 The Cronbach’s αs for the present study ranged from .74 to .84 across the 5 dimensions.
Diabetes self-management. The SDSCA includes 10 items measuring diabetes self-care activities, including following a healthy diet, engaging in physical activity, taking prescribed medication, monitoring of blood glucose, and performing foot care.32 Higher scores indicate more diabetes self-care activities done during the past 7 days. The SDSCA has good validity across 7 studies.32 The Chinese version of SDSCA has an acceptable internal consistency in people with diabetes in Taiwan (Cronbach’s α = .71).33 The Cronbach’s α for the current study was .68.
The authors examined the data using SPSS version 24.0 to ensure that all required statistical assumptions were met. Descriptive statistics and the correlation matrix were examined.34,35 The conceptual model of fatigue in adults with T2DM was tested by path analysis in a structural equation modeling framework using Mplus version 8.6 software.36 A statistically nonsignificant chi-square test, root mean square error of approximation (RMSEA) and its 90% CI of less than 0.08, and a comparative fit index (CFI) above 0.95 were considered indicators of a good model fit.37 The authors used full information maximum likelihood to handle missing data.34 In addition, fatigue intensity, duration, disturbance, and quality were the indicators for fatigue as a latent variable. An alternative model (Figure 2) was tested to lower the risk of confirmation bias.34
A Monte Carlo simulation was conducted to estimate the sample size needed to reach an adequate power of 80%.34 The population parameters for the latent variable indicator loadings were set at 0.7, and the direct path loadings were set at 0.4. With a proposed sample size of 150, the power to detect significant effects was above 80% for all direct and indirect paths specified from the structural model of fatigue in adults with T2DM.
The participants’ demographic and clinical information is presented in Table 1. Most of the 150 participants were male and married and self-reported as Hoklo Taiwanese. On average, the participants were about 59 years old with a BMI of about 27 and had about 2 to 3 comorbidities other than T2DM. Most of the participants were prescribed oral hypoglycemic agents (77%).
The descriptive statistics and the correlations of the study variables are presented in Table 2. On average, the participants had formal education for about 11 years, were diagnosed with T2DM for about 7 years, scored about 1.7 on the CCI, and had a mean A1C level of 7.32% (56 mmol/mol). The participants had low levels of depressive symptoms (CES-D mean score = 5.25) and diabetes distress (DDS mean score = 1.51), but they were having poor sleep quality (PSQI mean score = 5.08) according to the >5 cutoff score of the PSQI.26 Participants also reported moderate social support (MSPSS mean score = 4.9). For the fatigue characteristics, the participants reported an average fatigue intensity of 1.41, with about 3 days of feeling fatigued in the past week. The participants reported a mean score of 3.61 on the FSI disturbance subscale and had a mean score of 37.75 on the MFI. Overall, the participants performed diabetes self-management for about 3 to 4 days per week (SDSCA mean score = 38.56). The correlation matrix showed that the exogenous variables of the model were mild to moderately correlated with each other (rs = −.48 to .48) with the magnitude of the correlations less than .90, indicating low levels of collinearity.35
The measurement model (ie, the fatigue latent variable) fit the data well (χ2 = 1.28, P = .257; RMSEA = 0.04, RMSEA 90% CI, 0.00-0.06; CFI = 0.99). The standardized indicator loadings on the factor (Figure 3) were strong in this model (ranging from 0.75 to 0.92) and were all statistically significant at P < .001.
The full structural model had a good fit (χ2 = 37.27, P = .071; RMSEA = 0.05, RMSEA 90% CI, 0.00-0.09; CFI = 0.98). The unstandardized path coefficients for the relationships among influencing factors, fatigue, and DSM are presented in Table 3, and the standardized coefficients are shown in Figure 3. For the direct effects of the influencing factors on fatigue, a more recent diagnosis of T2DM (b = −0.002, P = .018), having more depressive symptoms (b = 0.10, P < .001), and reporting lower sleep quality (b = 0.07, P = .004) were related to higher levels of fatigue. Fatigue was also associated with DSM such that higher fatigue was related to less performance of DSM activities (b = −3.25, P = .005). The direct effects of the CCI score, A1C, diabetes distress, social support, and years of education on fatigue and DSM were nonsignificant. Overall, the influencing factors explained 54% of the variances in fatigue. The influencing factors and fatigue explained 28% of the variances in DSM (Figure 3).
The results of the mediation analysis are shown in Table 4. Although the direct and total effect of depressive symptoms on DSM were not significant, there was a significant indirect effect of fatigue on the relation between depressive symptoms and DSM (standardized indirect effect = −.17; 95% CI, −.36 to −.06), indicating that those with more depressive symptoms also report higher fatigue, and higher levels of fatigue are related, in turn, to fewer DSM activities. Fatigue also had a significant indirect effect on the link between poor sleep quality and DSM (standardized indirect effect = −.07; 95% CI, −.15 to −.02) such that those with poorer sleep quality reported more fatigue, which was then associated with DSM. The authors observed no significant mediated effects of fatigue on the links between diabetes duration, CCI, A1C, diabetes distress, social support, or education with DSM.
The authors tested the alternative model (Figure 2), and the results showed that the model fit of the alternative model (RMSEA = 0.21, 90% CI, 0.17-0.26; CFI = 0.43) was not as good as the original model (RMSEA = 0.05, 90% CI, 0.00-0.09; CFI = 0.98). Because the alternative model was nonnested, a statistic test to compare the fit of the original and the alternative model could not be conducted.18 However, the results from the alternative model provide some support for the arrangement of the constructs in the original model.
This study examined the relationship between fatigue, its influencing factors, and DSM using a structural equation modeling approach. The results showed that a more recent diagnosis of T2DM, more depressive symptoms, and lower sleep quality were related to higher fatigue. Higher fatigue was related to less performance of DSM for the past week. Importantly, fatigue mediated the relationship between depressive symptoms and DSM and the relationship between sleep quality and DSM. This study contributes to the literature by testing the TOUS using empirical data and measuring all aspects of fatigue proposed in the TOUS simultaneously.
Overall, the results found in this study were consistent with the assumptions of the TOUS. The authors found that the 4 aspects of fatigue (ie, intensity, duration, disturbance, and quality) fit the fatigue measurement model well, which indicated the importance of measuring all aspects of fatigue in future studies. Regarding the relationships between the influencing factors and fatigue, we found that only diabetes duration, depressive symptoms, and sleep quality significantly predicted fatigue. In other words, the effects of the psychological influencing factors (ie, depressive symptoms and sleep quality) on fatigue were stronger than the effects of physiologic (eg, CCI and A1C) and situational factors (eg, social support and education) on fatigue. The nonsignificant findings could be because of the complex relationship among the influencing factors.38 For example, Park et al15 found that A1C had an indirect effect on fatigue through diabetes symptoms, diabetes distress, and depression. Future studies with a larger longitudinal sample could examine a more complex model incorporating the multimediation effects among the influencing factors.
In addition, this study found that more depressive symptoms and lower sleep quality were related to more fatigue, which is consistent with previous studies in adults with T2DM.11,13-15,39 This study further adds to the literature that suggests fatigue mediates the effects of both depressive symptoms and sleep quality on DSM. These findings imply that managing all aspects of fatigue and depressive symptoms and improving sleep quality could likely improve the performance of DSM behaviors. Future interventional studies could develop fatigue management programs based on this finding and test the effectiveness of the intervention on improving fatigue and DSM in adults with T2DM.
This study examined fatigue and DSM using a theoretically grounded framework and incorporated all aspects of fatigue, both of which are key strengths. Nevertheless, this study has some limitations that need to be considered. The Cronbach’s α for the SDSCA in this study was only .68. The internal consistency of the SDSCA had been lower across various language versions, and researchers proposed that the lower internal consistency reflects the multidimensional nature of diabetes self-management.40,41 Although the reliability of the SDSCA meets established thresholds, future studies could consider evaluating different aspects of DSM separately.
Although all data were collected during the preventive shutdown for the coronavirus disease 2019 (COVID-19), the rates of new cases spiked in Taiwan during mid-May 2021.42 The authors conducted independent t tests with bootstrapping43 to compare the mean fatigue scores before and after the spike, and the results showed that before the spike, the participants reported more mental fatigue (95% CI, 0.57-3.06; P = .003) and more reduced motivation (95% CI, 0.18-2.11; P = .016) than those recruited after the COVID-19 spike. This result suggests that there could be a selection bias of people who had lower fatigue after the spike of COVID-19 cases in Taiwan and that the levels of fatigue in this study could be underestimated.
Fatigue mediates the effect of depressive symptoms on DSM and the link between sleep quality and DSM. Overall, the data supported the concepts and links proposed in the TOUS, and the conceptual model adapted from the TOUS adequately guided the study. Future development of fatigue interventions integrating depressive symptoms and sleep management while measuring and monitoring the intensity, duration, disturbance, and quality of fatigue will likely increase the performance of DSM and improve the health outcomes in adults with T2DM.
The authors declare that there is no conflict of interest.
Hsuan-Ju Kuo https://orcid.org/0000-0001-6227-8911
Julie A. Zuñiga https://orcid.org/0000-0003-3778-8148
Heather Cuevas https://orcid.org/0000-0003-4314-6686
Supplemental material is available online with this article.
From School of Nursing, College of Medicine, National Taiwan University, Taipei City, Taiwan (Dr Kuo); School of Nursing, The University of Texas at Austin, Austin, Texas (Dr García, Dr Zuñiga, Dr Cuevas); School of Nursing, Texas A&M University, College Station, Texas (Dr Huang); Department of Human Development and Family Sciences, The University of Texas at Austin, Austin, Texas (Dr Benner); Division of Endocrinology and Metabolism, Department of Internal Medicine, National Taiwan University Hospital Hsin-Chu Branch, Hsin-Chu City, Taiwan (Dr Fan); and Taipei City Hospital Renai Branch, Taipei City, Taiwan (Dr Hsu).
Corresponding Author:Hsuan-Ju Kuo, National Taiwan University, College of Medicine, School of Nursing, No.1, Sec. 1, Jen Ai Road, Taipei, 10051.Email: hsuanjukuo@ntu.edu.tw
