Although all age groups can experience fatigue, it is particularly common in older individuals. Fatigue becomes more burdensome and debilitating in older adults over the age of 60 years compared to younger individuals and children. This is partly due to the coexistence of a number of chronic illnesses in this population with common underlying processes such as inflammation (Hardy & Studenski, 2010), making it one of the most prevalent symptoms reported in older individuals (van Seben et al., 2019). Studies have shown that fatigue in older individuals is associated with biophysiological, sociodemographic, and psychological factors (Torossian & Jacelon, 2021). Fatigue is also associated with a poorer prognosis and higher odds of mortality in this population (Hofer et al., 2018). Because of its prevalence and health-related consequences, fatigue has been a topic of interest for a vast number of researchers, who have contributed to the advancement of scientific knowledge around a number of risk factors of fatigue. However, a comprehensive fatigue management model that could serve as a guide for healthcare providers caring for older individuals experiencing fatigue is lacking.
The theory of unpleasant symptoms (TUS) is a middle-range theory developed by Lenz et al. (1997), with the aim of guiding research and practice in the management of an individual's experience of one or more symptoms. The theory depicts three levels of factors that influence the experience of a symptom. The first category of influencing factors is physiological variables, related to the biological processes that maintain or disrupt normal body function, including nutrition, existence of a pathology, energy levels, trauma, and so forth. Psychological influencers constitute the second category of influencing factors and are variables related to one's state of mind, mood, and affective reaction to a disease or illness, such as depression and anxiety. Lastly, situational influencers are variables related to an individual's physical and social environment, such as place of residence, educational background, social network, and support. In the TUS, the three levels of influencing factors are interrelated and impact the experience of symptoms (Lenz et al., 1997). The experienced symptom, in its turn, affects the individual's functional and cognitive performance (CP). Functional performance is operationalized as an individual's physical and social function, that is the extent to which one is able to carry out activities of daily living, fulfill work- and role-related tasks, and participate in social interactions. Cognitive performance is operationalized as one's ability to carry out cognitive tasks, such as problem-solving, thinking, reasoning, and concentrating. The TUS also highlights feedback loops (Figure 1), which are beyond the scope of this paper.
Examining a fatigue management model based on the TUS, inclusive of multiple risk factors and chronic diseases, is plausible, as the experience of fatigue is associated with common underlying physiological, psychological, and situational risk factors. Hence, the purpose of this research study was to examine a fatigue management model (Figure 2), which highlights the relationship between fatigue and all three influencing factors and the impact of fatigue on physical performance (PP), social performance (SP), and CP, in addition to perceived health and quality of life (QOL). Another purpose of this study was to examine the mediating role of fatigue outcomes in the relationship between fatigue and QOL.
Methods
A nonexperimental, quantitative, cross-sectional exploratory study design was used, which included secondary data analysis. In this study, the source of data was the Patient-Reported Outcomes Measurement Information System-Health Utilities Index (PROMIS-HUI, Version 1.1; Cella, 2017) data set. PROMIS-HUI is a publicly available data set consisting of 150 items reflecting sociodemographic data, comorbidities, PROMIS global form, and PROMIS profile measures on seven domains, consisting of emotional distress, fatigue, physical function, pain, sleep, social participation, and cognition, in addition to the HUI.
Recruitment
Participants in this study were recruited from an online Internet survey company, "Opinions 4 Good" (Op4G). This readily available platform includes a panel of 152,000 respondents of different demographic backgrounds (Op4G, 2020). For PROMIS-HUI, respondents were selected such that they were representative of the 2010 U.S. census, despite analysis that later showed that the selected sample was sicker than the general U.S. population (Hays et al., 2016). Eligibility criteria included being 18 years of age or older, English-speaking, part of the U.S. general population, and enrolled in the Op4G panel. To recruit participants, an e-mail was sent by the online survey company to members of the panel, notifying them of a new survey opportunity. Interested participants filled out a consent form, followed by a survey consisting of nearly 150 items. Participants were compensated with an incentive provided by Op4G, which did not exceed 10 USD. In this study, only participants over the age of 60 years were included. The institutional review board at the University of Massachusetts Amherst determined that this project does not involve an intervention or interaction with individuals or does not use identifiable private information. Thus, institutional review board application was not required.
Study Measures
As previously mentioned, physiological variables are those related to the biological processes that maintain or disrupt normal body function, whereas psychological variables are related to one's state of mind, mood, and affective reaction to a disease or illness. Lastly, situational influencers pertain to an individual's physical and social environment, such as place of residence, educational background, social network, and support.
Physiological Influencing Factors
These included age (Bevilacqua et al., 2018; Salter et al., 2019), number of comorbidities (Horne et al., 2019; W. Q. Lin et al., 2015), pain (Crowe et al., 2017; Teshale, 2019), sleep (Barak et al., 2020; Loh et al., 2018), and gender (Bevilacqua et al., 2018; Salter et al., 2019).
Gender, unlike all other physiological factors, is a categorical variable, which included males (1) and females (2). Age was based on a self-report of the number of years lived, whereas number of comorbidities was based on the presence (1) or the absence (0) of the listed chronic diseases (sum of present chronic conditions), including hypertension, rheumatoid arthritis, asthma, migraines, diabetes, angina, cancer, lung, heart, liver, and/or kidney disease. Pain was measured using the PROMIS short form v1.0-pain intensity 3a (three items) and the PROMIS item bank v1.1-pain interference (nine items). Sleep was measured as the sum of scores of sleep disturbance and sleep-related impairment items selected from their respective item banks (16 items in total), the validities of which have been supported in different chronic conditions (Yu et al., 2012). Higher scores on both instruments reflect worse pain and higher sleep disturbance/impairment.
Psychological Influencing Factors
Psychological risk factors of fatigue included depression and anxiety (Barak et al., 2020; Crane et al., 2016; Kim & Son, 2020; Salter et al., 2019). Eight items from the PROMIS item bank v.1.0 emotional distress-depression and PROMIS emotional distress v.1.0 emotional distress-anxiety were used to measure depression and anxiety, respectively. The depression item bank, from which the items of the depression scale in this study were chosen, has shown adequate internal consistency ([alpha] = .988; Nolte et al., 2019).
Situational Influencing Factors
Situational factors associated with fatigue included education (Jing et al., 2015; Kessing et al., 2016; W. Q. Lin et al., 2015; Muszalik et al., 2016), marital status (Jing et al., 2015; W. Q. Lin et al., 2015), race (Chou, 2013; Franklin & Harrell, 2013; F. Lin et al., 2013), hospitalizations (Bhalla et al., 2014), and sensory impairment (SI), including hearing (Alhanbali et al., 2017; Hornsby & Kipp, 2016) and vision (Berthold Lindstedt et al., 2019). Unlike other chronic conditions that independently impact the individual's experience of symptoms, the repercussions or the burden of hearing problems are dependent upon social interactions. For this reason, despite being linked to physiological factors, SI was categorized as a situational factor in this study.
Race, marital status, educational level, and number of hospitalizations were nominal-level variables. Race was operationalized as White, Black, and others. Education was categorized as "no high school," "high school, GED, or technical/associate's degree," and "college degree." The number of hospitalizations was operationalized as self-reports of the frequency of hospital stays in the last 12 months and was dichotomized into <=2 or >2 hospitalizations per year, based on the average rate of hospitalization per community-dwelling older adult (Gjesten et al., 2018). Lastly, SI was based on the score of four items reflecting vision impairment (two items) and hearing impairment (two items). Scores for SI ranged between 0 and 4, with higher scores reflecting a higher level of SI.
Fatigue
This was the main outcome of interest in this study and was measured based on the scores of eight items on fatigue impact and eight items on fatigue experience selected from the PROMIS item bank v.1.0-fatigue (16 items total). The reliability and construct validity of PROMIS-fatigue short forms have been supported in individuals with heart failure, chronic obstructive pulmonary disease, back pain, and others (Cella et al., 2016; Flynn et al., 2015).
Fatigue Outcomes
Hypothesized outcomes included PP, SP, CP, perceived health, and QOL. Although not highlighted in the TUS, perceived health (Silva et al., 2011) and QOL (Schmidt et al., 2018; Tolstrup Larsen et al., 2018) have been shown to be consequences of fatigue and were thus included in the model. Physical performance was operationalized as the total sum of scores on 18 items selected from PROMIS v2.0 item bank-physical function, the psychometric properties of which have been supported (Crins et al., 2018; Oude Voshaar et al., 2015). Cognitive performance was measured by eight items on cognitive function and eight items on cognitive function-abilities. The questions were chosen from the PROMIS item bank-cognitive function v2.0 and PROMIS item bank-cognitive function-abilities v2.0, respectively. The reliability of the PROMIS cognition short form, which includes some of these items, has been supported (Fieo et al., 2016; Saffer et al., 2015). Social performance was operationalized as the sum of scores on eight items selected from PROMIS, the ability to participate in social roles and activities v.2.0 item bank and eight items from PROMIS satisfaction with social roles and activities v2.0 item bank. Both item banks have shown to have good internal consistency in individuals with rheumatoid arthritis (Bartlett et al., 2015). Perceived health was based on the score of two items that assessed subjective ratings of physical and mental health. Each of these items was based on a 5-point Likert scale, resulting in a summative score that ranged from 0 to 10. Lastly, QOL was based on one item only, a response ranging from "poor" to "excellent."
Cronbach's alpha for all continuous variables was also calculated, and findings showed that instruments used in this sample were reliable. That is, all Cronbach's alpha coefficients were greater than .7. Results pertaining to internal consistency reliability are presented in Table 1.
Data Analysis
A number of statistical analysis techniques were used to examine the hypothesized pathways. To examine the relationship between fatigue and the three categories of influencing factors, bivariate analyses were conducted. Significant physiological, psychological, and situational influencing factors were then entered simultaneously into multiple linear regression as independent variables, and fatigue was entered as the dependent variable.
For the second part of the model (fatigue outcomes), path analysis was used to examine the consequences of fatigue and the mediating effects of fatigue outcomes on the relationship between fatigue and QOL. The four paths examined are depicted below:
1. Age, gender, number of comorbidities, pain, sleep, depression, anxiety, marital status, education, SI, race, hospitalization [heavy wide-headed rightwards arrow] fatigue
2. Fatigue [heavy wide-headed rightwards arrow] PP, SP, CP, perceived health
3. Fatigue, PP, SP, CP, perceived health [heavy wide-headed rightwards arrow] QOL
4. Age, gender, number of comorbidities, pain, sleep, depression, anxiety, marital status, education, SI, race, hospitalization [heavy wide-headed rightwards arrow] PP, SP, CP, perceived health, QOL
The PROMIS-HUI data set was examined for missing data points. Descriptive analyses showed that missing data were not systematic and did not represent a threat to the internal validity of study findings. Thus, participants with missing data were excluded. Statistical assumptions were checked, and a "robust" option was used wherever needed. All statistical procedures in this study were carried out using Stata IC (2019, Version 16), and significance was set at [alpha] = .05.
Results
Data for 20 participants were deleted due to inaccurate data entry (unrealistic numbers). Of the remaining sample of n = 2,980, those below the age of 60 years and two participants who had missing data points for QOL were excluded. The final sample size was n = 725.
Sample Characteristics
The average age was 70.18 years, with an age range of 60-88 years. Of the total sample size (n = 725), the majority of participants were White (n = 440, 60.69%), and the rest were either Black, Hispanic, Asian, or of other ethnic backgrounds. About half of the participants were female (n = 400, 55.17%); the majority of participants had either a technical/associate degree or a college degree (n = 304, 41.76%) and were married (n = 453, 62.48%). Participants had an average of two to three comorbidities, and the majority (n = 481, 66.34%) had not been admitted to the hospital in the past 12 months (Table 2).
In terms of the descriptive statistics of the study variables, participants had mild pain with a mean of 27.38 (PROMIS range: 12-60) and moderate levels of sleep problems (M = 39.09; PROMIS range: 16-80) and fatigue (M = 38.32; PROMIS range: 16-80). Regarding the psychological variables, participants reported low levels of depression and anxiety on average. Most participants had no visual or hearing impairments (Table 3). In what relates to fatigue outcomes, most participants in this study had above average physical, social, and CP levels and "good" perceived health and QOL on a spectrum ranging from "poor" to "excellent."
Influencing Factors of Fatigue
Bivariate analyses showed that all variables, except for marital status, were significantly associated with fatigue. Hence, all significant variables were simultaneously entered into multiple linear regression using the "robust" option to account for heteroscedasticity of some variables (discussed above). The results of the regression are presented in Table 4.
Findings from the simultaneous entry of all significant risk factors of fatigue into multiple linear regression showed that three of the examined physiological factors, both psychological factors, and two of the situational factors were significant predictors of fatigue. The number of comorbidities (B = 0.49, p = .003), pain (B = 0.37, p < .001), sleep (B = 0.35, p < .001), depression (B = 0.20, p = .015), anxiety (B = 0.20, p = .020), having a high school degree/GED/technical or associate's degree compared to a college degree (B = 1.93, p < .007), and more than one SI (B = 2.04, p = .015) were significant predictors of fatigue, whereas all other variables rendered insignificant after simultaneous entry. Overall, the effect size of this model was moderate. Variables entered into the regression accounted for 69% of the variance of fatigue scores.
Fatigue Outcomes
As mentioned earlier, a path analysis consisting of four paths was conducted to examine the research question. Findings from the path analysis regarding the direct effect of fatigue on the selected outcomes showed that fatigue was a significant predictor of poorer physical (B = -0.18, p < .001), social (B = -0.51, p < .001), and CP (B = -0.24, p < .001), as well as poorer perceived health (B = -0.03, p < .001). Higher scores of these outcome variables, with the exception of CP, were shown to be significant predictors of a better QOL. Surprisingly, a better CP was a predictor of a poorer QOL (B = -0.01, p = .014). In addition, fatigue was not directly associated with QOL (B = 0.00, p > .05). The effects of all performance-based outcome variables on QOL were generally weak, but that of perceived health was moderate.
To support the examination of whether fatigue outcomes mediate the relationship between fatigue and QOL, the investigators calculated the indirect effect of fatigue on QOL. The results showed that the indirect path was significant (B = -0.01, p < .001), suggesting the presence of a mediating effect. Individual paths for mediators were not examined, because all mediators were significantly correlated, and examining them separately would not have been very informative. Results from path analyses are presented below (Figure 3).
Further analyses were conducted to investigate the relationship between fatigue outcomes and QOL, especially the unexpected negative association between CP and QOL. Social performance, CP, and perceived health were divided according to their domains, which, in turn, were entered into a regression equation. That is, SP was divided into "social ability" and "social satisfaction," CP was divided into "cognitive function" and "cognitive function-abilities," whereas perceived health was divided into "perceived physical health" and "perceived mental health." Physical performance was entered as one domain (per PROMIS item banks).
Findings from this regression analysis suggested that neither cognitive ability (B = -0.00, p = .430) nor cognitive function (B = 0.00, p = .974) predicted QOL, with beta coefficients of zero in both cases. Better social satisfaction (B = 0.01, p = .003), but not social ability (B = -0.00, p = .918), significantly predicted better QOL (Table 5). The relationship between PP and QOL was weak but significant (B = 0.01, p < .01), which was also the case for mental health and QOL (B = 0.21, p < .001). However, the association between perceived physical health and QOL was moderate (B = 0.60, p < .001).
Discussion
The aim of this study was to examine a fatigue management model as a first step to guide the care of older individuals experiencing fatigue through the early detection and management of risk factors of fatigue. Findings suggested that the number of comorbidities, sleep, pain (physiological), depression, anxiety (psychological), education, and SI (situational) are significant predictors of fatigue. The latter, in its turn, influences PP, SP, CP, and perceived health, which mediate the relationship between fatigue and QOL.
Although the association between fatigue and QOL was supported in other studies (Abrahams et al., 2018; Fan et al., 2020), this was not the case in this study. This might be due to the inclusion of mediators, such as PP, SP, CP, and perceived health. The significant association between QOL and PP, SP, and CP, respectively, in this study was almost negligible. Statistical significance in this case might be due to a large sample size alone, rather than an actual clinical significance, which is further supported by the fact that CP was insignificant in the post hoc analysis described above. Thus, it can be concluded that the relationship between all three performance outcomes (physical, social, and cognitive) and QOL is mediated by perceived health, the inclusion of which makes the association between these outcomes and QOL very weak.
Findings of this study are of value to older individuals experiencing fatigue, their family members, rehabilitation nurses, and researchers interested in improving the management of fatigue and the improvement of QOL of older individuals. These findings highlight the importance of screening older individuals for the identified risk factors of fatigue and implementing multicomponent interventions that address different risk factors. Several interventions have been developed to address the different risk factors of fatigue. For example, cognitive behavioral therapy, mindfulness, and sleep hygiene training (MacLeod et al., 1980; Wennberg et al., 2013) have been shown to be effective interventions for improving sleep. Furthermore, the application of cold and/or hot packs and relaxation breathing techniques have been shown to effectively minimize pain (Fouladbakhsh et al., 2011). However, such interventions and others address one outcome only and thus are not holistic. Findings of this study suggest the need for multifactorial interventions, as fatigue is linked to interrelated risk factors that contribute to the experience of fatigue. A study presented promising findings regarding yoga therapy in the management of depression, pain, and sleep, with high participant adherence rates (Cartwright et al., 2020). Hence, rehabilitation nurses should implement similar interventions that address all or most risk factors of fatigue. This would not only manage fatigue but also improve PP, SP, CP, perceived health, and QOL.
In addition, given the prevalence of hearing, vision, and dual SI in older individuals (Elliott et al., 2015; Roets-Merken et al., 2014), consistent vision and hearing screening protocols for older individuals should be implemented in the community and nursing homes. In a recent study, researchers reported that less than half of the nursing home staff used hearing and vision screening tools to detect SI in residents (Andrusjak et al., 2021). They also identified a lack of routine assessments for SI and a lack of access to vision and hearing assistive devices (Andrusjak et al., 2021). A number of easy-to-administer SI screening tools are available for nursing home staff and nurses. The Severe Dual Sensory Loss screening tool (Roets-Merken et al., 2014) is one of many inexpensive and easy-to-administer SI screening tools and has been shown to be reliable and valid in detecting SI in older adults.
Future research recommendations include examining a number of associations that were not examined in this study and exploring new fatigue-related risk factors. Cognitive performance was negatively (although negligibly) correlated with QOL, which warrants further investigation on the moderating role of neurodegenerative diseases, such as dementia, in the association between CP and QOL. In addition, the feedback loops depicted in the TUS were not investigated in this study. That is, examining the extent to which fatigue influences its predictors (sleep, pain, and depression) and the extent to which PP, CP, and SP influence fatigue and its predictors would illustrate a better picture of the interplay between these variables.
Study findings should be viewed in lieu of this study's limitations. First, this was a secondary data analysis, which renders the researcher unaware of the circumstances in which participants responded to questionnaires. Second, a number of variables, including social support, physical activity, medications, and income, which have been shown to influence fatigue levels, were not included in the data set. Third, QOL, one of the main outcomes in this study, was operationalized as the score to a single item, which does not accurately represent the multifaceted aspects of QOL. Fourth, there were no data on neurodegenerative diseases, such as mild cognitive impairment or Alzheimer's disease, which have been shown to influence the association between CP and QOL (Chuang et al., 2016; Hsiao et al., 2016; Stites et al., 2017). Lastly, path analysis does not infer causality, the establishment of which calls for randomized controlled trials.
The strengths of this study lie in its large sample size, the recruitment of study subjects from multiple states; the inclusion of older individuals with different chronic conditions representative of the 2010 U.S. census; the use of reliable, valid PROMIS measures; and the simultaneous examination of multiple risk factors of fatigue. Another strength of this study is the examination of outcomes of fatigue as mediators in the relationship between fatigue and QOL using the TUS, which helps plan multifactorial interventions to improve QOL. Furthermore, all the relationships depicted in the TUS, except for feedback loops, were examined in this study, which provides a better understanding of the interplay between different variables associated with fatigue, as opposed to the examination of fatigue in relation to either its risk factors or consequences.
Conclusion
Findings in this study supported the association between fatigue and pain, sleep, SI, education, and depression among older individuals. This association highlights the importance of routine screening for these risk factors in older individuals and the implementation of multifactorial interventions for timely and effective fatigue management. Doing so would improve CP, PP, and SP, as well as QOL. Findings suggested that CP, SP, PP, and perceived health mediate the relationship between fatigue and QOL. Future research should be directed toward examining the characteristics of the relationship between fatigue and the above-mentioned outcomes, exploring other risk factors of fatigue, examining feedback loops depicted in the TUS, and identifying factors (such as neurodegenerative diseases) that moderate the relationship between fatigue outcomes and QOL.
Key Practice Points
* Rehabilitation nurses and other healthcare providers should screen older individuals for three levels of influencing factors of fatigue.
* Nurses should consider the theory of unpleasant symptoms in planning interventions to address patients' fatigue.
* Nurses should examine the effectiveness of the examined model in clinical practice and propose modifications as needed.
Conflict of Interest
The authors declare that there is no conflict of interest.
Funding
The authors declare that there is no funding associated with this project.
Acknowledgments
The authors extend their immeasurable appreciation and gratitude to Dr. Jessica Pearlman for her significant input in selecting the appropriate statistical techniques to answer all four research questions in this research study.
References