Introduction
Majority of patients at the end of life, as well as their family caregivers, choose their homes as the preferred place of care (Dittborn et al. Reference Dittborn, Turrillas and Maddocks2021; Gerber et al. Reference Gerber, Hayes and Bryant2019). With the recent COVID-19 pandemic, there is a growing demand for hospital beds as well as a generalized fear of contracting an infectious disease globally (Fordjour and Chow Reference Fordjour and Chow2022; Shadmi et al. Reference Shadmi, Chen and Dourado2020). Patients with advanced illness therefore spend most of their time at home. Family caregivers are major support systems for this group of patients. In meeting the multidimensional needs of patients, the different demands and responsibilities often induce an enormous level of stress and strain on these caregivers. Family caregivers of terminally ill patients have reported adverse psychological and physical consequences from the burden of care (Onyeneho and Ilesanmi Reference Onyeneho and Ilesanmi2021). A previous study by Oechsle et al. (Reference Oechsle, Ullrich and Marx2020) revealed that about 90% of family caregivers of patients facing advanced cancer have clinically relevant distress with common symptoms of exhaustion, sadness, loneliness, and sleep problems. In response to the stress and strain, some use maladaptive coping, which, in return, may cause complicated health issues for the family caregivers. For instance, a considerable number of caregivers have alcohol use and drug use disorders resulting from high care burdens (Webber et al. Reference Webber, Davies and Leach2020).
Palliative care has been advocated worldwide to improve the quality of life not only for the patient but also for their family members (World Health Organisation 2020). Psychosocial and educational end-of-life care (EoLC) interventions that are caregiver-focused have been made available globally. Caregiver burden or strain is one of the 5 common outcome measures for caregivers under EoLC intervention (Ahn et al. Reference Ahn, Romo and Campbell2020). The UK Medical Research Council has highlighted the importance of the process evaluation of such complex interventions. Particularly, the mechanism of impact outlining the causal assumptions of the intervening variables with the outcomes is required to be spelled out clearly in the design phase (Moore et al. Reference Moore, Audrey and Barker2015). For optimal intervention, addressing the caregiver strain and understanding its mechanism of change are essential. Fundamentally, there is also a need to understand what constitutes caregiver strains considering cultural or geographical sensitivity.
Caregiver strain, stress, and burden are often used interchangeably. Thornton and Travis (Reference Thornton and Travis2003) described stress and burden as the physical and or emotional response to caregiving challenges, whereas strain is defined as the perception of the persistent difficulties that affect the well-being of the caregivers. In most intervention studies for caregivers, strain is considered a unidimensional outcome (Ahn et al. Reference Ahn, Romo and Campbell2020; Chan et al. Reference Chan, Chan and Suen2013); yet, caregiving is complex and multidimensional (Thornton and Travis Reference Thornton and Travis2003). The caregiver strain index was originally developed by Robinson (Reference Robinson1983) as a measure of objective strain to assess caregiver-perceived strain. A study by Rubio et al. (Reference Rubio, Berg-Weger and Tebb1999) assessed the constructs of the caregiver strain and identified 4 different factors, namely inconveniences, adjustment, upsetting, and work adjustments, which were different from the original study. Thornton and Travis (Reference Thornton and Travis2003) also examined the reliability of the earlier version of caregiver strain index by Robinson (Reference Robinson1983) for applications with contemporary long-term family caregivers in USA.
Building on the earlier version, Thornton and Travis (Reference Thornton and Travis2003) amended and developed the Modified Caregiver Strain Index (M-CSI), which consists of 13 items, covering the subjective and objective elements of caregiver strain. They proposed a 3-factor model of the M-CSI: perception of caregiving, care–recipient characteristics, and emotional status (Thornton and Travis Reference Thornton and Travis2003). The study by developers of M-CSI, Thornton and Travis (Reference Thornton and Travis2003), also found 3 components in the measurement, but the analysis of correlation with caregiver strain was done as a unidimensional construct. This M-CSI measurement has been translated and applied in various cultures, such as Portuguese (Ribeiro et al. Reference Ribeiro, Brandão and Oliveira2021), Chinese (Chan et al. Reference Chan, Chan and Suen2013), and Turkish (Ugur and Fadiloglu Reference Ugur and Fadiloglu2010), to assess caregiver strain. Previous studies have also employed the M-CSI, particularly among family caregivers of patients with different illnesses (Lutz et al. Reference Lutz, Kucharska-Newton and Jones2022; Muriuki et al. Reference Muriuki, Oluchina and Mbithi2021; Sorayyanezhad et al. Reference Sorayyanezhad, Nikpeyma and Nazari2022).
The Turkish version by Ugur and Fadiloglu (Reference Ugur and Fadiloglu2010) identified 4 factors of M-CSI, namely adaptation, upsetting, inconvenience, and overwhelmed. The Portuguese version by Ribeiro et al. (Reference Ribeiro, Brandão and Oliveira2021) reported a 2-factor model, namely individual experiences of burden and repercussions on the caregiver’s life. Contradictorily, a validation study of the Chinese version by Chan et al. (Reference Chan, Chan and Suen2013) among Hong Kong caregivers distinguished a single-factor model of M-CSI. However, in their study, no confirmatory factor analysis (CFA) was conducted to further validate the structure and compare the findings with different structure models.
From the construct validity, the items in the M-CSI represent very different dimensions. For guiding the development of complex support for caregivers of patients with advanced illnesses, a more rigorous investigation of the measurement leading to the conceptualization of the core outcome construct is needed. This study aims to explore (1) the structural dimension of strain among family caregivers of patients with end-of-life care in Hong Kong and (2) examine the differences in the identified factors between demographic groups. We hypothesized that the structure of family caregivers’ strain is multifactorial. Based on the findings from the study, we can propose a clearer concept of caregiver stain, which is distinguishable from caregiver stress or caregiver burden. A multifactorial caregiver strain identified in this study can provide broader domains for screening of high risk sociodemographic groups. Based on the findings from this study, service directions and plans for tailored caregiver-focused psychosocial EoLC interventions can be developed, which conform to the United Nations’ health-related Sustainable Development Goals of ensuring healthy lives and promoting well-being for all persons.
Methods
Procedures and study participants
The data of this study were extracted from a larger evaluation study of a home-based end-of-life care project in Hong Kong. Data from the participants were collected through a self-administered standardized questionnaire. The baseline clinical assessment data from the project inception (January 2016) to December 2018 were utilized, which are also more comparable with the Hong Kong Chinese version by Chan et al. (Reference Chan, Chan and Suen2013), in terms of the time frame and sample size. Family caregivers of patients diagnosed with a terminal illness with a prognosis of less than 12 months were included. The inclusion criteria of family caregivers were those (1) aged 18 years old and above and (2) identified by the patient as the primary caregiver who provides the main care and thus at the risk of caregiving burden. Paid caregivers such as domestic helpers were excluded from the study as they are not family caregivers.
Measures
Demographic characteristics of family caregivers explored in this study encompass caregivers’ age (younger age <60 years vs. older age ≥60 years), gender (0 = male and 1 = female), relationship with the patient (spouse vs. adult child), living with the patient (yes vs. no), average caregiving time per day in hours (≤8 hours daily vs. >8 hours daily), and major diagnosis of the patient (cancer vs. non-cancer).
The Chinese version Modified Caregivers Strain Index (C-M-CSI), validated in Hong Kong by Chan et al. (Reference Chan, Chan and Suen2013), was used to assess caregivers’ strain. The C-M-CSI consists of 13 items on various caregiving challenges, such as “caregiving is a physical strain.” Each item was answered by a caregiver on a 3-point response scale: 0 = no, 1 = yes, sometimes, and 2 = yes, always. When used as a scale the total scale score ranged between 0 and 26.
Data analyses
The Kolmogorov–Smirnov test confirmed the data to be normally distributed (p > 0.05). Split-half method of data distribution was employed to split the total sample into 2 halves for use separately in the exploratory factor analysis (EFA) and CFA of the 13 items of C-M-CSI. The 2 subsample sizes met the commonly used sample size criterion of “10 subjects per item” for EFA and CFA (Bollen Reference Bollen1989; Goretzko et al. Reference Goretzko, Pham and Bühner2021; Watkins Reference Watkins2018). Using the R (version 3.5.3) statistical package, the EFA was estimated by maximum likelihood methods, with oblique Crawford-Ferguson (CF)-varimax rotation used for the first half of the participants (n = 227). Given the nested structure of the data, that is, caregivers nested within 4 nongovernmental organizations, sandwich standard error, known as robust standard error, was used (Zhang et al. Reference Zhang, Preacher and Hattori2019). Parallel scree plot was employed to ascertain the number of factors to be extracted, and items with a factor loading greater than 0.30 were assigned to each factor (Arcolin et al. Reference Arcolin, Godi and Corna2022).
Following the EFA, a CFA was conducted on data from the second half of the participants (n = 226) to confirm the factor structure suggested by the EFA. Using Mplus (version 8), the CFA model was estimated by robust maximum likelihood methods, which take clustering cases into account (Padgett and Morgan Reference Padgett and Morgan2021). The model fit of the CFA model was assessed by multiple fit indices, including model Chi-square, comparative fit index (CFI), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR). For RMSEA and SRMR, a value of 0.06 or lower indicates a good fit, while a CFI value of 0.95 or higher is accepted as a good fit (Shi et al. Reference Shi, DiStefano and Maydeu-Olivares2022). Additional CFAs were conducted to examine and compare the model fits with the C-M-CSI models identified by previous studies (See Table 3). The internal consistency of the total scale and extracted subscales were tested by Cronbach’s alpha for an acceptable coefficient of 0.70 and above (Goretzko et al. Reference Goretzko, Pham and Bühner2021). Finally, after the ideal factor structure was determined, generalized linear models (GLM) using the entire sample were performed to explore the demographic correlates with the identified factors of C-M-CSI.
Results
Descriptive results
Table 1 presents the descriptive statistics of the total sample of study participants as well as the 2 separate samples used for EFA and CFA. Of the 453 family caregivers, 76.6% were female, with a mean age of 57.57 years (SD = 14.14). They spent approximately 50 hours (per week) on caregiving. About 50.6% of the caregivers had spousal relationships with the patients, while 48.95% of caregivers were adult children of the patients. More than half (55.4%) had cancer patients, while 71.5% lived with the patients. No significant difference in their demographic variables was detected for the EFA and CFA sample.
Table 1. Participants’ demographic information
a Patients with chronic lower respiratory disease, heart failure, renal disease, motor neurone disease, kidney disease, Parkinson’s disease, stroke, and dementia.
EFA results
Initially, the factorability of the C-M-CSI items was investigated. The Kaiser–Meyer–Olkin measure of sampling adequacy was 0.91, which is well above the commonly recommended threshold of 0.60 (Watkins Reference Watkins2018). Bartlett’s test of sphericity was also significant (χ 2 (78) = 2499.12, p < 0.001). Examination of the scree plot and eigenvalues (greater than 1.0) yielded a 3-factor solution explaining 50% of the variance for the entire set of variables. After CF-varimax rotation, items with a factor loading greater than 0.30 were considered significant to load on a factor (Watkins Reference Watkins2018).
As shown in Table 2, 3 factors emerged from the 13 items with no double-loading. The first factor was labeled “Perception of Caregiving,” which included 6 items, and it explained 22% of the variance. The second factor, which explained 15% of the variance, was labeled as “Empathetic Strain,” which only included 2 items. The last factor labeled “Adjustment Demand” involved 3 items, with 13% explained variance. The validity of the 3 factors extracted was assessed using Promax rotation method, which showed good validity with a factor correlation matrix of less than 0.70. Internal consistency for each of the factors, as well as the overall scale, was examined using Cronbach’s alpha. The alpha was good, with coefficient above 0.70 for the overall scale and the subscales.
Table 2. EFA factor loadings: 3-factor solution (n = 227)
CFA results
The CFA for the 3-factor solution from the EFA indicated the model as acceptable (χ 2 [62, N = 226] = 125.68, p < 0.001, CFI = 0.94, Tucker–Lewis Index [TLI] = 0.93, SRMR = 0.05, and RMSEA = 0.07). Modification indices showed that model fit would improve if residuals between items 7 and 8 correlate. Moreover, both items were loaded on the same factor (i.e., adjustment demand) and were semantically similar. Thus, the revised model added the residual correlation between items 7 and 8, with model fit indices showing a good fit of the data (χ 2 [61, N = 226] = 108.86, p < 0.001, CFI = 0.96, TLI = 0.95, SRMR = 0.04, and RMSEA = 0.06 [for factor loadings and factor correlations], see Figure 1).
Fig. 1. Three-factor CFA of the 13-item M-CSI.
Model fit comparison analyses with models from previous studies are summarized in Table 3. Models 3, 4, and 5 from previous studies indicated poor fit (CFI < 0.90 and RMSEA > 0.08), while Model 2, a 4-factor model, had an acceptable fit. Given that Model 1 (i.e., the 3-factor model of this present study) and Model 2 are not nested, parametric tests, such as Chi-square difference tests, were not used.
Table 3. Comparative studies analysis of model goodness of fit indices
Chi-square, Satorra–Bentler scaled Chi-square; df, degrees of freedom; CFI, comparative fit index; TLI, Tucker–Lewis Index; SRMR, standardized root mean square residual; RMSEA, root mean square error of approximation; CI, confidence interval.
*** p < 0.001.
A simple comparison of model fit indices showed that Model 1 extracted in this study had a better fit in terms of RMSEA, SRMR, TLI, and CFI. In addition, the 4-factor model in Model 2 includes a single-item factor (work adjustment), which makes it impossible to compute internal consistency reliability using Cronbach’s alpha. Furthermore, the adjustment factor of the 4-factor model was found to have low levels of internal consistency with a Cronbach’s alpha of 0.66, while, as noted earlier, all the 3 factors of the present study had good internal consistency above 0.70. As a result, the 3-factor model of the present study model was determined to be the best fitting model of C-M-CSI.
Exploration of correlates
The influential role of caregiver’s gender, age, and other factors, such as whether living with the patient, patient health condition, relationship with the patient, and caregiving time were assessed on the identified 3 subscales of the C-M-CSI. Table 4 presents the result of the GLM regression.
Table 4. The GLM regression result
* p < 0.05,
** p < 0.01,
*** p < 0.001.
The GLM regression analysis revealed that gender had no significant influence on the experience of the various dimensions of caregiver strain (p > 0.05). Negative perception of caregiving was positively related to family caregivers of cancer patients (β = 0.70, p < 0.05), who live with the patient (β = 0.46, p < 0.05), who are spouses of the patient (β = 0.14, p < 0.05), and who provide longer caregiving time of more than 8 hours daily (β = 0.28, p < 0.001). Empathetic strain was negatively related to younger family caregivers of ages below 60 years (β = −0.22, p < 0.05), who do not live with the patients (β = −1.01, p < 0.05), who have no spousal relationship with the patient (β = −0.35, p < 0.05), and who provide few hours of care daily (<8 hours) (β = −0.37, p < 0.001). Adjustment demand had positive relation to caregivers of cancer patients (β = 0.82, p < 0.01) and younger family caregivers aged below 60 years (β = 0.14, p < 0.05).
Discussion and implications
Constructing the structural dimension of strain
This study’s primary goal was to examine the structure of the C-M-CSI by Chan et al. (Reference Chan, Chan and Suen2013) using a sample of Hong Kong Chinese family caregivers of patients with terminal illness. The factor analyses discovered a new 3-factor model, namely the perception of caregiving, empathic strain, and adjustment demand. The reliability tests confirmed the total scale (0.89) and the 3 subscales (perception of caregiving – 0.82; adjustment demand – 0.80, and empathetic strain – 0.79) have good internal consistency reliabilities. The single factor of C-M-CSI for caregivers of chronically ill patients identified by Chan et al. (Reference Chan, Chan and Suen2013) also found a Cronbach’s alpha of 0.91. Consistent with the original validation study of caregiver strain index (Robinson Reference Robinson1983) and the modified version by Thornton and Travis (Reference Thornton and Travis2003), the result from this study suggests that caregiver strain is a multidimensional concept. Comparison of the model fit indices with previous studies gave evidence that in the Hong Kong context, the three-factor model captures family caregivers’ strain more appropriately than the unidimensional model proposed by Chan et al. (Reference Chan, Chan and Suen2013).
With reference to the 3 different dimension items included in the factors, perception of caregiving is the appraisal which was value laden. This finding might be affected by past and current experiences or deliberation of family or cultural values. The second factor, empathetic strain, describes the emotional reactions to witnessing the changes or behaviors of patients, which is specific to the caring recipient and is more contingent. The last factor, adjustment demand, though it is also contingent, portrays the purposive accommodation of the caring role by changing daily functioning. More important of all, all these are carrying a negative connotation. The identified factors share some commonalities with previous studies about caregivers’ strain by capturing the perception, the emotional aspects (Robinson Reference Robinson1983), and adaptation issues (Rubio et al. Reference Rubio, Berg-Weger and Tebb1999; Ugur and Fadiloglu Reference Ugur and Fadiloglu2010). Moreover, compared to a general claim of the emotional aspect, empathetic strain is more accurate in summarizing the related items and more specific to the group of caregivers. In our study, there are no overlapping components, which makes our generated 3-factor caregiver strain model a good fit.
Differentiation of the concept of stress, burden, and strain
Concepts of stress and strain have been defined and applied in the field of mechanical engineering for a long time. Stress is described as the force applied per unit area, whereas strain is the amount of deformation caused by the stress, which is divided by the initial dimension (Collins Reference Collins2019). Stress and strain are thus different, though they have related concepts. When Thornton and Travis (Reference Thornton and Travis2003) reviewed the measurements of stress, strain, and burden used in caregiving research, they found that the terms stress and strain have been used interchangeably without a commonly accepted definition. They attempted to define stress and burden as the physical or emotional response to challenges in the caregivers, while strain was described as the caregiver’s perception of the enduring problems or an altered state of their well-being. The researchers then developed the M-CSI (Thornton and Travis Reference Thornton and Travis2003), which has been explored in various cultures as mentioned earlier.
The M-CSI has still been used interchangeably to measure caregiving stress, strain, or burden. For example, in the study by Saimaldaher and Wazqar (Reference Saimaldaher and Wazqar2020), the M-CSI was used as the measure of caregivers’ stress, instead of caregivers’ strain. There is also the concern of whether stress or strain should be defined in terms of the person, the environment, or both (Hart Reference Hart, Cooper, Anderson, Ones and Sinangil2001). Despite the theoretical congruence in some literature, strain, burden, and stress are distinct concepts.
Redefining the concept and measurement of caregiver strain
Borrowing from the engineering concepts of stress and strain, we propose to define stress as the external demands in caregiving and strain as the responses or changes of the caregivers after the exposure to stress. This new conceptualization, depicted in Figure 2, offers a clearer process of change, with caregiver stress as the external force, which brings about the changes in various forms of caregiver strain and consequently affects the health outcomes and quality of life of the caregiver.
Fig. 2. Conceptualization of caregiver strain.
Following the new conceptualization, the factor of adjustment demand is considered as the caregiver burden or stress, whereas the empathetic strain is the caregiver strain, which is emotional in nature. The factor of perception of caregiving is considered as an element that affects the caregiver strain but being affected by the caregiving stress. The relation between perception of caregiving and actual strain may be bidirectional. Moreover, the demands of the caregiver tasks, including the adjustment factor, are considered as stress, which are the external forces exerted on the caregivers. Some studies have also reported positive outcomes of stress in caregiving among family caregivers (Lindeza et al. Reference Lindeza, Rodrigues and Costa2020; Zhong et al. Reference Zhong, Wang and Nicholas2020).
If the definition of caregiver strain is the change in the individual after resuming the caregiver role, then this study postulates that caregiver strain would have negative, neutral, and positive dimensions. These can be valuable directions for future research.
Identifying sociodemographic factors associated with caregiver strain
Our study found that the experiences of the 3 subscales of C-M-CSI differed among family caregivers’ sociodemographic characteristics. This study revealed that the level of empathic strain, adjustment demand, and negative perception of caregiving did not differ by caregivers’ gender. Notwithstanding, significant differences were revealed in some caregivers’ sociodemographic characteristics. Notably, family caregivers for patients with terminal cancer exhibited higher levels of negative perception and adjustment demand than their counterparts who cared for non-cancer patients. Indeed, previous studies (Lippiett et al. Reference Lippiett, Richardson and Myall2019; Thana et al. Reference Thana, Lehto and Sikorskii2021) indicated that cancer patients have complex needs for treatment, and thus caregiving for cancer patients is demanding. A study by Starr et al. (Reference Starr, Washington and McPhillips2022) also postulated that the vast majority of family caregivers with cancer patients experienced moderate to severe sleep disturbance but were hesitant to take the prescribed sleep medication due to their caregiving responsibilities. Similarly, other studies attributed high caregiving demands on family caregivers of cancer patients to the decline in health status and financial well-being of the caregivers (Junkins et al. Reference Junkins, Kent and Litzelman2020; Zavagli et al. Reference Zavagli, Raccichini and Ostan2022). These give adequate reasons why the level of the negative perception and adjustment demands subscales were greater for family caregivers of cancer patients.
Family caregivers of ages below 60 years were more likely to experience adjustment demand, as they often have to fulfill other family duties (e.g., childrearing) or work-related responsibilities, which may generate inter-role conflicts (Boumans and Dorant Reference Boumans and Dorant2021). Unlike the perception and adjustment subscale, the result indicates that irrespective of the types of illness of the patients being cared for, caregivers share a universal empathetic strain. This sheds more light on the care of caregivers. Instead of focusing on the physical and practical supports to the caregivers, which most caregiver support programs do (Hudson et al. Reference Hudson, Morrison and Schulz2020), there is universal emotional suffering faced by the caregivers of patients with a terminal illness, which should not be overlooked. The results also revealed that longer caregiving time of more than 8 hours daily leads to higher levels of empathetic strain. However, younger family caregivers of ages below 60 years who do not live with the patient have less level of empathetic strain.
We recommend that health professionals in Hong Kong and to some extent globally should pay critical attention to these higher risk sociodemographic groups and assess their caregiving strain considering the domains of their perception of caregiving, empathetic strain, and adjustment demand. This will protect the higher risk family caregivers from adverse mental, psychosocial, and physiological consequences resulting from caregiving.
Conclusion
This study critically examined the slippery concept of caregiver strain and offered a precise and distinct new definition ideologized from the engineering discipline. It offers a new differentiation between caregiver stress, burden, and strain, which provides a logical explanation of the controversies of outcomes of caregiver strain in the available studies. Based on the new conceptualization proposed in this study, the caregiver strain is defined as the reactions to the caregiving stress or demands. It can include other emotional reactions, such as the feeling of being torn, or other positive ones, such as contentment. Additionally, physical reactions such as fatigue and hypervigilance can be included. Further qualitative studies of the reactions of caregivers toward caregiving demands and stress can offer more insights in this area.
The new conceptualization advances future studies in examining the consequences of caregiver strain that will offer new directions for handling the negative consequences. The advancement in this theoretical conceptualization and empirical study may inform the development of EoLC interventions tailored to different subgroups of family caregivers to ensure their health and well-being, which are in conformity with the United Nations’ health-related Sustainable Development Goals. More research should be carried out to explore the emotional suffering as well as the changes in emotion or emotional strain of the caregivers. Specifically, empathetic emotional pain resulting from the suffering and change in patients’ conditions should be critical targets for emotional caring practice or research.
The study has a limitation. The sample of this study involved a homogenous sample of Chinese family caregivers of patients with terminal illness, which will limit the generalizability of the factor structure of M-CSI found in this study to caregivers of other target groups with different cultural backgrounds, as portrayed by the different factor structures of M-CSI reported in previous studies. Despite the limitations, the study sheds light on future research to further refine the concept of caregiver strain and its relationship with stress by taking into account family caregivers with diverse cultural backgrounds.
Acknowledgments
The authors would like to thank all nongovernmental organizations’ partners in the project that provided services to patients and their family members.
Author contributions
A.Y.M.C. designed the study, conceptualized the major arguments, and wrote the paper. G.A.F. revised the manuscript and assisted with the data analysis. K.J. assisted with the writing of the paper and performed the data analysis. N.J. assisted with earlier data analysis. I.K.N.C. supervised the data collection stage and offered comments on the draft. C.N.K. assisted with earlier data analysis.
Funding
This work was supported by funding from the Hong Kong Jockey Club Charities Trust.
Conflicts of interest
No conflicts of interest exist.
Ethical approval
Ethical approval of the study was obtained from the human research ethics committee of the University of Hong Kong (approval number: EA1512009). Verbal or written consent was obtained from all study participants.
Open access
