BURDEN IN FAMILY CAREGIVERS OF CANCER PATIENTS: THE ASSOCIATION WITH DEPRESSION, RELIGIOSITY AND RELIGIOUS COPING DR. SHIM VUN KONG DISSERTATION SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF PSYCHOLOGICAL MEDICINE DEPARTMENT OF PSYCHOLOGICAL MEDICINE FACULTY OF MEDICINE UNIVERSITY OF MALAYA

Introduction: Cancer caregiving is challenging and often associated with significant burden in family caregivers. Yet, the healthcare providers often overlook the needs of the caregivers. Local data on cancer caregiving is still lacking. In line with the increasing cancer cases worldwide, there is a need to examine factors influencing the caregiver burden, and to find ways to improve the quality of life of the caregivers. 
Objectives: The study aimed to determine the caregiver burden rate among the local family caregivers of cancer patients, and to examine the associations between the caregiver burden and their sociodemographic factors, clinical factors, levels of depression, religiosity, and religious coping patterns. 
Methods: 127 family caregivers of cancer patients in a government tertiary hospital in Sarawak, Malaysia were recruited cross-sectionally using non-random sampling method. The associations between the variables were examined through bivariate analyses (Chi-Square test), followed by multivariate analysis to determine the factor(s) that were significantly associated with the caregiver burden. 
Results: The caregiver burden rate among local cancer caregivers was 55.6%. Two factors remained significant after adjusted in multivariate analysis. Caregivers who were the offspring of cancer patients experienced significantly fewer burden than non-offspring caregivers (OR 0.41, CI 0.18 – 0.94, p = 0.035). Also, the caregiver depression was significantly associated with the caregiver burden (OR 4.26, CI 1.87 – 9.72, p = 0.001). There were no significant associations found between the religiosity and religious coping with the caregiver burden on multivariate level. 
Conclusion: Caregiver burden is common among family caregivers of cancer patients. Caregivers who are the children or grandchildren of cancer patients are less likely to experience caregiver burden, whereas caregivers who have probable depression are more likely to experience caregiver burden. Future studies should assess the effects of specific intervention strategies in helping these caregivers.

(1) I am the sole author/writer of this Work; (2) This Work is original; (3) Any use of any work in which copyright exists was done by way of fair dealing and for permitted purposes and any excerpt or extract from, or reference to or reproduction of any copyright work has been disclosed expressly and sufficiently and the title of the Work and its authorship have been acknowledged in this Work; (4) I do not have any actual knowledge nor do I ought reasonably to know that the making of this work constitutes an infringement of any copyright work; (5) I hereby assign all and every rights in the copyright to this Work to the University of Malaya ("UM"), who henceforth shall be owner of the copyright in this Work and that any reproduction or use in any form or by any means whatsoever is prohibited without the written consent of UM having been first had and obtained; (6) I am fully aware that if in the course of making this Work I have infringed any copyright whether intentionally or otherwise, I may be subject to legal action or any other action as may be determined by UM.     Table 4.14 Bivariate analysis of the association between caregiver burden with depression, religiosity, and religious coping among the study subjects using Chi-square  is postulated to increase to over 20 million by 2025, with the low and middle-income countries bearing most of the cancer burden (Ferlay et al., 2015).
According to the Malaysian National Cancer Registry, a total of 103,507 new cancer cases were diagnosed between 2007 and 2011 in Malaysia. This incidence rate was much lower than the estimated rate reported in GLOBOCAN 2012, as the latter was based on the Penang and Sarawak Cancer Registries with different ethnic distributions (Manan, Tamin, Abdullah, Abidin, & Wahab, 2016). From the same authors, Sarawak documented 9,734 new cancer cases or equivalent to about 9.4% of the national figures, and ranked fourth in the total cancer cases following Johor, Selangor and Penang.
It is well known that cancer causes significant physical and emotional impairments in its sufferers (Stein, Syrjala, & Andrykowski, 2008;Silver, Baima, & Mayer, 2013). Not only the patients, cancer also indirectly affects their significant others, many of whom are the primary informal caregivers (Blanchard, Albrecht, & Ruckdeschel, 1997;Lim, Kim, & Lee, 2013;Wozniak & Izycki, 2014). Cancer patients U n i v e r s i t y o f M a l a y a and their caregivers are frequently exposed to the complications of the disease itself, as well as the side effects of the various cancer treatments available.
According to Family Caregiver Alliance based in the United States, family (or informal) caregiver refers to any person, e.g. spouse, adult children, other relatives, neighbour or friend, who has a personal relationship with, and provides a wide range of unpaid assistance for, an older person or an adult with a chronic or disabling condition (Family Caregiver Alliance, 2014). This is different from a professional (or formal) caregiver, who is either a paid carer or a volunteer with no personal relationship with the person he or she is looking after.
Caring for a loved one can be associated with significant stress or burden, more so in caring for patients with chronic or terminal illnesses such as dementia and cancer.
Among the difficult and time-consuming tasks faced by the family caregivers involve provision of emotional support to the patients, transportation of patients to the hospital for treatment, and management of the behavioural aspects, as well as the disease symptoms of the patients (Bakas, Lewis, & Parsons, 2001).
There is a significant reciprocal relationship between the emotional distress of cancer patients and their caregivers (Hodges, Humphris, & Macfarlane, 2005;Hagedoorn, Sanderman, Bolks, Tuinstra, & Coyne, 2008;Northouse, Katapodi, Schafenacker, & Weiss, 2012). Thus, the management of cancer patients would be compromised if the caregivers' well-being is affected (Mahadevan et al., 2013). Despite caregiving has a significant impact on the caregivers' well-being, the needs of the caregivers are often overlooked or considered secondary to those of the patients (Payne, Smith, & Dean, 1999).

U n i v e r s i t y o f M a l a y a
Recent advancement in the diagnostic and therapeutic modalities has resulted in a paradigm shift in the cancer course and care plan (Rowland & Ganz, 2011;Shekarian, Valsesia-Wittmann, Caux, & Marabelle, 2015). The course of cancer has changed from being an acute condition with rapid or direct consequences, usually fatality, to a chronic illness with variable outcomes (Nijboer et al., 1998). This may translate to a need for long-term and continuous care for the cancer patients, with the integration of both informal and formal care systems.
Consequently, the family caregivers are at increased risk of being burdened with multiple stressful physical and psychosocial problems (Girgis, Lambert, Johnson, Waller, & Currow, 2013). Psychological problems such as anxiety, depression, or loneliness, are the most commonly recognised burdens in caregivers (Stenberg, Ruland, & Miaskowski, 2010). In line with the increasing trend of cancer cases and caregiver burden worldwide, there is now a need to look into the caregiving aspects and to find ways to improve the well-being of the family caregivers.
One suggested solution is through the study of religious or spiritual beliefs and practices, which is present in most human cultures since ancient times. Yet, research on the roles of religion and its relation to the human well-being accounts only a small fraction of the current literature in psychology (Ano & Vasconcelles, 2005). Religiosity, spirituality and religious coping have been shown to play important buffering roles in helping cancer patients to deal with psychological distress (Weaver & Flannelly, 2004).
Thus, it is possible that these benefits might be extended to the cancer caregivers as well.
U n i v e r s i t y o f M a l a y a

RATIONALE OF THE STUDY
There were many studies done overseas on the subject of caregiver burden among the family caregivers of cancer patients (Nijboer et al., 1998;Given et al., 2004;Northouse et al., 2012;Chua et al., 2016). However, most of the studies only identified the caregiver burden in general, but did not specifically look at the factors associated with the high burden among the caregivers (Hsu et al., 2014). In Malaysia, local published data about this subject is still lacking despite the increasing trend of cancer burden nationwide. There is also no validated scale in Malay language suitable for the measurement of cancer caregiver burden in local setting.
As to date, majority of the related studies on cancer caregiver burden in Malaysia were conducted in Peninsular Malaysia (Ambigga, Sherina, & Suthahar, 2005;Mahadevan et al., 2013;Jaafar et al., 2014), with only one qualitative study being carried out in Sarawak thus far (Cheong & Putit, 2011 Thus, it is high time that a local study should be carried out in Sarawak, looking into the aspects of cancer caregiving burden as mentioned above. This research study will hence be able to provide the much needed baseline data for the Sarawak population.
In addition to determining the rate of caregiver burden, the study will help us to explore and to identify the links between the caregiver burden and potential predictors such as the sociodemographic profiles, the level of depression, religious commitment and religious coping patterns in our local setting.
A better understanding of the unique association between all these factors is crucial to shed light on previously unrecognized issues that may affect the caregiving outcomes. It is hopeful that the findings from this study can guide the policy makers and the healthcare professionals on how to implement targeted intervention strategies to help the family caregivers in the future. In addition, this preliminary study might also provide some direction for future research in this area relevant to the local setting.

RESEARCH QUESTIONS
This study was designed to address the following research questions: 5. Is the caregiver burden associated with the level of religiousness in the family caregivers?
6. Is the caregiver burden associated with the level of religious coping in the family caregivers?

RESEARCH OBJECTIVES
In relation to the above research questions, this study has the following objectives in mind: 3. There is a significant association between the cancer caregiver burden and their level of depression; i.e. higher depressive score is associated with higher caregiver burden, and vice versa 4. There is a significant association between the cancer caregiver burden and their level of religiosity; i.e. higher level of religiosity is associated with lower caregiver burden, and vice versa 5. There is a significant association between the cancer caregiver burden and their religious coping; i.e. positive religious coping is associated with lower caregiver burden or negative religious coping is associated with higher caregiver burden

CONCEPTUAL FRAMEWORK
The primary framework for this study was adapted and modified from the model framework of the caregiving process proposed by Nijboer et al. (1998). According to that model, the outcome differences of the caregiver burden can be influenced by exposure to either a stressor or a mediator. Stressors refer to factors that positively induce stress, and thus potentially increase the caregiver burden. On the other hand, mediators refer to factors that negatively reduce stress and may potentially lower the caregiver burden. Figure 1.1 illustrates the conceptual framework of the present study.

Definition and Setting of Caregiver Burden
In practice, the term 'caregiver burden' is frequently used interchangeably with other terms such as 'caregiver strain" or 'caregiving stress'. Across the literature, various studies mentioned about these terms together, but none had attempted to differentiate the meaning of each individual term (Amen, 2010;Merluzzi, Philip, Vachon, & Heitzmann, 2011;Bevans & Sternberg, 2012). According to Kramer, research that utilizes the concept of 'caregiver burden' is essentially focusing on the stress/strain posed by the caregiving role (as cited in Bastawrous, 2013). For the purpose of simplicity and to avoid confusion, only the term 'caregiver burden' will be used throughout this dissertation.
Caregiver burden is an important outcome and measure commonly investigated in both observational and interventional literatures associated with caregiving of patients. Yet, the term is still poorly understood and not well-defined to date (Bastawrous, 2013). One of the earliest interpretations was by Zarit, Reever, & Bach-Peterson (1980), who defined caregiver burden as the extent to which caregivers perceive that caregiving has had an adverse effect on their emotional, social, financial, physical, and spiritual functioning.

U n i v e r s i t y o f M a l a y a
A more comprehensive definition of caregiver burden is "a multidimensional biopsychosocial reaction resulting from an imbalance of care demands relative to caregivers' personal time, social roles, physical and emotional states, financial resources, and formal care resources given the other multiple roles they fulfill" (Given, Kozachik, Collins, DeVoss, & Given, 2001, as cited in Given et al., 2004).
There are many studies conducted on the topic of caregiver burden in different clinical diseases and settings. Alzheimer's disease is the index condition for the study of caregiver burden (Markman, 2014). Over the years, the study scope gradually expanded to include caregivers of other conditions such as cancer, stroke and other mental illnesses (Etters, Goodall, & Harrison, 2008;Rigby, Gubitz, & Phillips, 2009;Girgis et al., 2013). In comparison, the caregiver burden of cancer patients had been reported to be comparable to that of dementia patients, but more than the burden in caregivers of diabetes or elderly patients (Kim & Schulz, 2008). The distress level reported by caregivers can be equal to or even greater than that of the cancer patients (Hodges et al., 2005).

Measurement of Caregiver Burden
The measurement of caregiver burden is a challenging task in view of the cultural, ethical, religious and personal values of different caregivers influencing their understandings on the meaning and consequences of burden (Chou, Chu, Tseng, & Lu, 2003). Many measurement tools, consisting mostly self-administered questionnaires, were invented over the years to assess the level of caregiver burden (Chou et al., 2003;Markman, 2014;Stagg & Larner, 2015 ZBI was the most popular and the most extensively evaluated caregiver burden tool in the literature (Stagg & Larner, 2015). Since the publication of the original 29item version in 1980, various translated versions and shorter revised versions of the ZBI (e.g. ZBI-12 and ZBI-4) have been produced (Bédard et al., 2001). Compared to other measurement tools, the popularity of the ZBI provides the added advantage in which the data obtained across different studies can be easily compared and examined. More details of the ZBI scale are described further in the sub-topic heading 3.8.2 Zarit Burden Interview.
Despite the abundance of measurement tools, there is no uniform approach or consensus achieved on what constitutes a significant caregiver burden (Bastawrous, 2013). It is therefore insufficient to rely on the scores of a single instrument for the diagnosis of caregiver burden without careful consideration of other factors affecting the outcomes of caregiving. Nevertheless, the scores can guide the clinicians in identifying at-risk caregivers who might need further assessment and intervention (Chou et al., 2003).

Factors Associated with Caregiver Burden
Several factors are associated with significant caregiver burden as identified by the many studies conducted on the caregivers of cancer patients. Among these factors, sociodemographic characteristics of the caregivers are important predictors. Higher level of psychological distress has been reported in caregivers who are women, young, U n i v e r s i t y o f M a l a y a spouses of the patients, employed, and those with lower socioeconomic status or poor social support (Nijboer et al., 1998;Kim & Given, 2008). Ethnicity of the caregiver has also been found to be one of the contributing factors (Rivera, 2009;Hsu et al., 2014), although caution should be exercised when generalizing this finding to other ethnic populations.
Additionally, clinical characteristics of the cancer patients and the caregiving process also play significant roles in predicting the caregiver burden. Patient factors such as the cancer types, cancer phases, treatment setting, duration of treatment, patients' quality of life and functional status have all been implicated in previous studies (Gaugler et al., 2005;Pellegrino et al., 2010;Hsu et al., 2014;Lukhmana, Bhasin, Chhabra, & Bhatia;Chua et al., 2016). For example, Hsu et al. (2014) reported higher level of burden in caregivers who cared for patients with solid tumours, in comparison with haematological malignancies. In particular, brain tumour was significantly linked to caregiver stress (Gaugler et al., 2005). Caregiver distress has been demonstrated even in the early phases of cancer when the patient was just diagnosed (Pellegrino et al., 2010).
Compared to curative treatment, caregivers of patients receiving palliative treatment or end-of-life care were reported to have lower scores in quality of life and physical health domains, which corresponded to higher caregiver burden (Weitzner, McMillan, & Jacobsen, 1999;Grov, Dahl, Moum, & Fossa, 2005). This finding is closely related to the patients' poorer functional status, which indicates increase in the care demands and longer caregiving hours per week, thus predicting higher caregiver burden (Weitzner et al., 1999;Hsu et al., 2014). Gaugler et al. (2005) also demonstrated that patients who received shorter duration of treatment were associated with higher burden in their caregivers.

The Local Setting
Locally, Mahadevan et al. (2013) found that sharing of caregiving burden and patient's functional status were significantly associated with caregiver burden among family caregivers of breast cancer patients in Kuala Lumpur Hospital. In that study, caregivers who did not share their caregiving burden were 2.8 times more likely to feel distressed. Shared caregiving was presumed to be protective, as it allows time-off for caregivers to engage in leisure activities as well as to look after their own health (Mahadevan et al., 2013).

Definition and Impact of Caregiver Depression
Caregiver depression refers to the affective disturbance experienced by the caregiver as a result of stress in the caregiving process . Similar to the general population, depression in caregivers may include low mood, insomnia, fatigue, reduced appetite or libido, pessimism, feeling of guilt, hopelessness, and suicidal thought (American Psychiatric Association, 2013). Some individuals with depression may only present with somatic complaints such as headache, joint pain, and gastrointestinal problems (Trivedi, 2004).
If go unrecognized, caregiver depression can be an extremely disabling condition with a negative impact on the physical health of the caregivers (Kurtz, Kurtz, Given, & Given, 2004). For example, depression has been linked to coronary heart U n i v e r s i t y o f M a l a y a disease and even early death in the caregivers (Rivera, 2009). Not only that, depressive symptoms and their consequences have been found to be relatively constant and may persist even after caregiving has ended (Haley et al., 2008).

Measurement of Caregiver Depression
Depressive

The Association with Cancer Caregiving and Caregiver Burden
The association between cancer caregiving and caregiver depression is well documented. In a study done by Rhee et al. (2008)   Relationship between caregiver burden and caregiver depression has been a constant debate among researchers in the caregiving industry. Early researchers believed that both concepts are the same and can be used interchangeably (Baillie, Norbeck, & Barnes, 1988, as cited in Stommel, Given, & Given, 1990. Nevertheless, recent literatures have concluded that they are two distinct constructs explained by different variables, thus should be targeted separately by individual intervention strategies Butler, Turner, Kaye, Ruffin, & Downey, 2005).
Many studies have found that caregiver depression is highly correlated with caregiver burden (Hérbert, Bravo, & Préville, 2000;Butler et al., 2005;Grov, Fossa, Sorebo, & Dahl, 2006;Rivera, 2009). In fact, care burden may be the best predictor for the development of depression (Rhee et al., 2008). This relationship can be bidirectional, in which the presence of depression invariably increases the risk for perceived burden in caregiving (Stommel et al., 1990). It is however noteworthy that causal relationship between these two constructs is still not established to date as majority of the studies done were cross-sectional in design (Pirraglia et al., 2005).
A review by Rivera (2009)

The Local Setting
In the context of local setting, Jaafar et al. (2014) reported that 17.7 percent of the family caregivers were depressed while caring for breast cancer patients in Kuala Lumpur Hospital. In that study, patient's functional status and caregiver's education level were found to be significantly associated with caregiver depression. The authors explained that the increased risk for depression might be due to the anticipatory grief and the impending loss of loved one associated with advanced stage of the cancer. Low education in the caregivers may lead to patients' late presentation to the healthcare services, thus indirectly increases the caregiving demands expected from the more severe symptoms in the patients (Jaafar et al., 2014).
Another earlier local study by Ambigga et al. (2005) reported a much higher prevalence of caregiver depression at 48.6%. Education level of the caregivers and duration of illness of the patients were found to have significant associations with the caregiver depression in this study. Substantial difference in the rate of caregiver depression between the two local studies may be due to the different instruments being used to assess depressive symptoms. In the latter, HADS, a screening tool, was used to measure the levels of depression; as compared to the use of Mini International Neuropsychiatric Interview (MINI), a diagnostic instrument, in the study by Jaafar et al.

Definition and Roles of Religiosity
In the historical context, religiosity is commonly equated to another popular term 'spirituality', and the distinction between these two terms is often unclear (Miller & Thoresen, 2003;King & Crowther, 2004). According to Koenig, McCullough, & Larson, religiousness is considered "a measure of an individual's belief in, adherence to, and practice of a given religion", whereas spirituality "corresponds to the personal quest for meaning and purpose in life, which may or may not be associated with religions" (as cited in Paiva, Carvalho, Lucchetti, Barroso, & Paiva, 2015). In this dissertation, 'religiosity' will be used interchangeably with 'religiousness'.
Religiosity (or religiousness) refers to the multidimensional aspects of the religious beliefs and involvement, encompassing subjective, cognitive, behavioural, social and cultural dimensions (Bergan & McConatha, 2000). Ironically, religion was once viewed as a unidimensional construct and considered as neither helpful nor functional by early researchers or psychologists, most notably Sigmund Freud and Albert Ellis (Ward, 2010 (King & Crowther, 2004). All these measures have their own strengths and limitations as they cover different components of the religion, such as involvement in religious activities (organizational and nonorganizational), intrinsic orientation, extrinsic orientation, religious attitudes and practices, and religious identification and affiliation (King & Crowther, 2004;Ng, Mohamed, Sulaiman, & Zainal, 2016).

The Association with Caregiver Burden
The potential influence of religion on caregiver burden is increasingly being recognized. Many family caregivers of cancer patients reported that religion and spirituality help them to cope better with their stress (Weaver & Flannelly, 2004;Delgado-Guay et al., 2012;Paiva et al., 2015). A prospective cohort study involving 175 caregivers of cancer patients found that the caregivers with high religiousness score were found to be significantly less likely to have major depression at 13-month followup (Fenix et al., 2006).
Nevertheless, majority of studies reported no or a mixed association (i.e. a combination of positive, negative, or non-significant results) between religion and wellbeing of informal caregivers (Hackney & Sanders, 2003 , 2006). Intrinsic religiosity and organizational religious activity were associated with lower perceived burden, while non-organizational religious activity was associated with poorer mental health (Herrera, Lee, Nanyonjo, Laufman, & Torres-Vigil, 2009).
Variation in the outcomes of religious studies may be explained by the use of nonuniform and single-item measures in these studies, which failed to address the multidimensional nature of religiosity (Stefanek et al., 2005).

The Local Setting
To date, there is yet any local study that specifically examines the relationship between religiosity and caregiver burden in Malaysia. Nevertheless, local data on religiousness is available among cancer patients and psychiatric patients. In particular, Nurasikin et al. (2012) Tix and Frazier (1998) defined religious coping as "the use of cognitive or behavioural techniques, in the face of stressful life events, that arise out of one's religion or spirituality". In simple terms, religious coping is the use of religious means to deal with stress, and the coping strategies can be either positive or negative (Pargament, Smith, Koenig, & Perez, 1998). In a review of literature by Harrison, Koenig, Hays, Eme-Akwari, & Pargament (2001), religious coping was noted to be widely used, especially in medically ill, hospitalized patients to cope with their illness or with life in general.
According to Pargament et al. (1998), positive religious coping strategies include appraising a secure relationship with a benevolent God, a belief that there is meaning in life, and seeking support from clergy/church members; while negative coping strategies include attributions of situations to a punishing God and feelings of abandonment by God, a less secure relationship with God, a tenuous and pessimistic view of the world, and religious struggle in the search for significance. In response to a particular stressful situation for example, a person who uses positive religious coping strategies may perform prayer, confess his or her sins, or seek strength and comfort from God (Tix & Frazier, 1998).
Religious coping and religiosity are two separate constructs, although they are closely related to each other (Tix & Frazier, 1998;Ng et al., 2016). Holland et al. (1999) have found a significant association between greater reliance on religious and spiritual beliefs, and the use of active-cognitive coping style. This kind of coping is defined as a U n i v e r s i t y o f M a l a y a person's acceptance of his or her illness, and the attempt to view its effects in a positive, meaningful manner (Fawzy et al., 1990, as cited in Holland et al., 1999. Religiosity components such as religious affiliation and religious orientation have also been identified as important moderators of the effects of religious coping during stressful events (Tix & Frazier, 1998).

Measurement of Religious Coping
The measurement of religious coping has gradually expanded and gained more precision over the years (Harrison et al., 2001). In the past, religious coping was measured based on the frequency of religious prayer or congregational attendance, but this method did not address the functional roles of religion in coping (Pargament, Koenig, & Perez, 2000). Subsequent scales such as Ways of Coping Scale and Religious Coping Activities Scale were refined to become more comprehensive, multidimensional, empirically based and yet clinically valid (Pargament et al., 2000).
This eventually led to the development of a short measure of religious coping (Brief RCOPE), which is now widely used in most research studies on religious coping (Pargament, Feuille, & Burdzy, 2011).

The Association with Caregiver Burden
Despite a growing body of evidence suggesting the increase use of religious coping in stressful events, research on the efficacy of religious coping in general has yielded mixed results (Harrison et al., 2001;Ano & Vasconcelles, 2005). A review by Harrison et al. (2001) reported that negative religious coping has consistently been U n i v e r s i t y o f M a l a y a associated with more depressive symptoms, whereas positive religious coping strategies have been consistently related to self-esteem, life satisfaction and quality of life. These findings were confirmed quantitatively in a meta-analysis of 49 studies, which concluded that positive and negative styles of religious coping were associated with positive and negative psychological adjustment to stress respectively (Ano & Vasconcelles, 2005).
Similar mixed results were also seen in the studies of the effects of religious coping pattern on the caregiver burden in particular. In a study of hospice caregivers, those who appraised their situation in a positive outlook with God reported positive mental health outcomes, while caregivers who appraised their situation as punishment or abandonment by God had negative outcomes (Mickley, Pargament, Brant, & Hipp, 1998). Pearce, Singer, & Prigerson (2006) reported that the use of negative religious coping was associated with more burden and less satisfaction; while surprisingly, the use of positive coping was also associated with more burden, but with more satisfaction in the caregiving. Increased use of positive religious coping strategies and diminished use of negative coping strategies were associated with posttraumatic growth among the family caregivers of cancer patients in India (Thombre, Sherman, & Simonton, 2010). Herrera et al. (2009) however found that only negative religious coping was significantly associated with higher levels of caregiver burden and worse mental health outcomes. These findings were similar to a study done by Hebert, Zdaniuk, Schulz, & Scheier (2009) in women with breast cancer. In contrast, other studies reported that only positive religious coping showed significant correlation with caregivers' psychological well-being, whereas negative religious coping had no effect on any outcomes (Gholamzadeh, Hamid, Basri, Sharif, & Ibrahim, 2014;Heo, 2014).

The Local Setting
In Malaysia, the two local studies that examined the religiosity component also studied the association between religious coping with psychological distress level (Nurasikin et al., 2012;Ng et al., 2016). Both studies have discovered and agreed that negative religious coping was significantly associated with higher distress level, but there was no significant correlation found between positive religious coping and the level of distress. Cancer patients with depression were reported to be using more negative religious coping mechanisms, however, the causal relationship cannot be established as the study was limited by a cross-sectional design and a relatively small sample size (Ng et al., 2016). This was a hospital-based, cross-sectional, and association study on the burden of the cancer caregivers; and also looking at its association with caregiver depression, religiosity, and religious coping styles.

STUDY SUBJECTS
The source population of this study was the family (or informal) caregivers of any cancer patients, either outpatients or inpatients, in a local government hospital.

STUDY SETTING
This study was carried out at the Department of Radiotherapy, Oncology and

U n i v e r s i t y o f M a l a y a
The department consists of outpatient clinics, chemotherapy daycare centre, radiotherapy unit (RTU), and three inpatient oncology wards (Male RTU ward, Female RTU ward and Palliative ward). Based on the SGH Annual Report in 2014, total numbers of new oncology cases seen in outpatient clinics and daycare centre were 1725, while total numbers of follow-up cases were 14786. As for inpatient cases, total admissions in 2014 were 2623 cases, with the palliative care consisting of 1354 cases, and this corresponded to a bed occupancy rate of 74% (Sarawak General Hospital, 2015). This information indicated the suitability of SGH to be the study site for data collection in this study.
The study involved family caregivers of any cancer patients that were receiving oncology treatments at the chemotherapy daycare or the outpatient clinics; as well as those that were admitted to the inpatient wards either for active chemotherapy or radiotherapy treatment, or to receive palliative care. The study was also conducted at the hematology daycare and hematology ward of SGH for recruitment of caregivers of patients with hematological malignancies.

STUDY PERIOD
Data for this research study was collected within four months from early January 2017 until end of April 2017. The sample was accessed from the study site on random days from Monday to Friday based on the convenience of the investigator. In circumstances where there was presence of more than one family caregiver for a particular cancer patient, the core or principal caregiver was selected. Those caregivers who gave their consent to participate in the study were screened for the following inclusion and exclusion criteria. Based on the formula proposed by Kish (1965), the sample size of this crosssectional study (Phase II) was calculated as follow:

Inclusion Criteria
where α is the level of significance (α = 0.05) Zα/2 is the z-statistic for 95% confidence interval (Zα/2 = 1.96) D is the absolute precision (D = 10%) P is the estimated prevalence rate of caregiver burden in cancer patients (P = 43.5%, based on the study by Lukhmana et al., 2015) Therefore, n = 1.96 2 [0.435 (1 -0.435) / 0.1 2 ] = 94.4 After taking into account the dropout or non-response rate of 20%, a reasonable sample size of at least 115 subjects was planned for this study. was made similar to the English version (range 0 to 4), so that direct comparison of the total scores can be made between these two versions of ZBI.

STUDY PROCEDURE
After the above corrections, the finalized version of the MZBI was used to interview 50 family caregivers who were bilingual (knowing both English and Malay language). At the same time, the caregivers were also given the sociodemographic and Phase II, a cross-sectional association study, was the main phase of this study.
The aim of this phase was to examine the objectives and research hypotheses as In view of time and resource constraints, the two phases were carried out concurrently throughout the study period using convenience sampling method. A single investigator carried out the data collection procedure in this study. The sample was collected from the family caregivers that were conveniently available at the study site and were willing to participate. Phase I involved recruitment of the first 50 subjects who could comprehend both the English and Malay languages. At the same time, Phase II recruited those subjects who comprehend mainly the Malay language regardless of their level of English language proficiency. The summary of the study flow chart is shown in The family caregivers of cancer patients from the above setting were approached and invited to participate in the study. Participation was based on voluntary basis.
Explanation was given to the study subjects regarding the indication of this study and what information would be required from them. Subjects were told that the entire task would take approximately half an hour, and that their decision on whether to participate in the study or not would in no way affect the treatment of the patients under their care.
The subjects were then given ample time to read and understand the information sheet (refer to Appendix C & D) which further explained the details of the study. Any queries from the subjects were addressed accordingly by the investigator. Once they U n i v e r s i t y o f M a l a y a have consented for their participation, the subjects were required to fill in the informed consent form (refer to Appendix E & F) attached together with the information sheet.
Thereafter, the subjects were administered with the relevant questionnaires and measures by the investigator. All data were collected independently with full confidentiality assured. Out of the total 177 family caregivers who consented for this study, 50 caregivers were recruited into Phase I and 127 caregivers were recruited into Phase II.

Sociodemographic and Clinical Questionnaire
The On the other hand, the clinical questionnaire consists of data on the types of cancer, cancer severity, treatment setting, and functional status of cancer patients; together with caregiving parameters such as duration as a caregiver, duration of caregiving in a week, and presence of any shared caregiving and/or external support. In the presence of shared caregiving, the number of other family caregivers assisting in caregiving was also assessed. As far as possible, the clinical information was obtained U n i v e r s i t y o f M a l a y a from the caregivers directly. In case there was ambiguity of the information provided by the subjects, patients' medical record would be referred for confirmation.

Further explanations of the individual sociodemographic and clinical variables
are available under the topic heading 3.9 Operational Definitions. Please refer to Appendix G to view sample of the sociodemographic and clinical questionnaire used in this study.

Zarit Burden Interview
Zarit Burden Interview (ZBI) is the oldest and the most widely used measure for assessment of caregiver burden (Hérbert et al., 2000). It was originally a 29-item questionnaire used to assess subjective burden in caregivers of impaired elderly (Zarit et al., 1980). The measure was later revised to the current self-administered, validated, 22item scale that evaluates caregiver's physical health, psychological and social wellbeing, finances, and relationship between caregiver and care recipient (Zarit, Orr, & Zarit, 1985 The factor structure of ZBI is still not clear with many different models existing in the literature (Knight, Fox, & Chou, 2000). In this study, the 3-factor model by Bianchi, Flesch, da Costa Alves, Batistoni, & Neri (2016) was adapted. The three factors emerged from the exploratory factor analysis were "Tensions related to the role" (Factor 1), "Intra-psychic tensions" (Factor 2), and "Competencies and expectations" (Factor 3) respectively. These factors explained 44% of the total variability and showed good internal consistency (Cronbach's alphas ranging from 0.72 to 0.83) when tested on the senior caregivers in Brazil (Bianchi et al., 2016).
The EZBI was translated to the Malay version (MZBI) for local use by Rosdinom et al. (2011)

Center for Epidemiological Studies -Depression Scale
Center for Epidemiological Studies -Depression (CES-D) scale was initially designed by Radloff (1977) to screen for depressive symptoms in the general population, but was later recommended by the author to be used as This scale is unique in a way that not all the items are worded as negative statements. In fact, the item no. 4, 8, 12, and 16 are worded as positive statements to control for response bias. Therefore, these four items are rated in a reverse order from 3 to 0 compared to the rest of the items. The scores on the items are then summed for the total score, with higher scores indicating more depressive symptoms. The total score ranges from 0 to 60. A cut-off score of 16 is suggested for identification of potential clinical depression in the respondents (Radloff, 1977).

Duke University Religion Index
The Duke University Religion Index (DUREL) is designed to measure religious commitment or religiosity of the respondent (Koenig, Parkerson, & Meador, 1997;Koenig & Büssing, 2010). It consists of five items covering three major dimensions of religiosity: organizational religious activity (ORA, 1 item); non-organizational religious activity (NORA, 1 item); and intrinsic religiosity (IR, 3 items).
ORA consists of public religious activities e.g. frequency of attending religious services or participation in group-related religious activities. NORA involves religious activities performed in private, e.g. prayer, reading the religious scripture or watching religious television show. On the other hand, IR refers to the degree of personal religious commitment or motivation, whereby persons with high IR regard religion as the ultimate significance compared to other personal needs (Koenig & Büssing, 2010).
ORA and NORA are scored from 1 to 6, whereas the 3 items in IR are scored from 1 to 5. Thus, the total score for DUREL ranges from 5 to 27. However, the authors of the original DUREL recommended the measure to be interpreted separately within the three subscales of ORA, NORA and IR, corresponding to the three distinct religious dimensions measured by them (Koenig et al., 1997). Koenig and Büssing (2010) reported that the DUREL has high internal consistency (Cronbach's alpha ranged from 0.78 to 0.91) and high test-retest reliability (intra-class correlation 0.91).

U n i v e r s i t y o f M a l a y a
According to Koenig and Büssing (2010), the DUREL scale is available in up to ten different languages and has been used in over 100 published studies since its introduction. The scale has been translated into the Malay language (DUREL-M) and was validated by Nurasikin, Aini, Aida, & Ng (2010) (Nurasikin et al., 2010;Nurasikin et al., 2012;Ng et al., 2016). Please refer to Appendix L to view the sample for this scale.

OPERATIONAL DEFINITIONS
Throughout this dissertation as well as in the study questionnaires, the following terms and operational definitions were applied to ensure uniformity and to facilitate clear understanding.
1. Family caregiveralso known as informal caregiver, defined as any person who has a personal relationship with, and provides a broad range of unpaid assistance and informal care for, a person with a chronic or disabling condition (Family Caregiver Alliance, 2014). Therefore, based on this definition, the caregivers in this study do not necessarily need to be related to the cancer patients by blood or marriage, and may include neighbours or close friends of the patients. However, the patients must identify the caregivers as their primary carers who provide majority of the informal care.
U n i v e r s i t y o f M a l a y a 2. Burden -similar to 'Strain' or 'Load', defined as a duty or responsibility borne by a person, in this context, the caregiver of cancer patient. In this study, the burden of the caregivers is measured quantitatively with the Zarit Burden Interview scale as mentioned above.
3. Depressiondefined as a mental condition associated with low or depressed mood, which covers a broader spectrum of depressive disorders under the DSM-5 criteria.
In the context of this study, the condition is experienced by the caregiver and not the cancer patient. MCES-D scale was used in this study to measure the level of depression objectively.
4. Copingdefined as "the cognitive and behavioral efforts made to master, tolerate, or reduce external and internal demands and conflicts among them". (Folkman & Lazarus, 1980). In short, coping is a method to reduce the stress level. In the setting of this study, the use of religious means for coping is measured objectively among the caregivers using the M-RCOPE scale.
5. Agea continuous numerical variable expressed in years. The lower limit of age in this study was 18 years old. There was no upper limit of age set in this study.
6. Gender -a categorical variable grouped into either 'Male' or 'Female' only. 'Low income' is defined as total household income of less than RM 3,000 a month.
'Mid-range income' ranges between RM 3,000 to RM 10,000, whereas 'High income' refers to total household income of more than RM 10,000 per month. The above cut-off numbers were derived from the statistics of the Malaysian middle class incomes in 2012 (Chi, 2014  This was carried out with the assumption that caregiver burden is closely related to    The commonest religion was Christianity (54%), followed by Islam (24%) and Buddhism (16%). One of the caregivers (2%) had no religion. Majority of the caregivers were married (56%), employed (54%), and from middle-income group with household income between RM 3,000 and RM 10,000 per month (54%). Most of them stayed in terraced houses (48%), semi-detached houses (20%) and village houses (20%).
A substantial portion of the caregivers was well-educated with 66% had tertiary education and 32% were educated up to secondary level. Only one caregiver (2%) had primary education and none was without formal education. Majority were either the children or grandchildren of the cancer patients (62%), followed by the spouses (24%) and the parents (6%).

Clinical Characteristics
Majority of the caregivers in this study were looking after relatives with solid cancer (92%), while the rest were caring for relatives with hematological malignancies.
Among the solid cancer group, breast cancer (n = 11) was the commonest cancer suffered by relatives of the caregivers. This was followed by colorectal cancer (n = 10), lung cancer (n = 4) and cervical cancer (n = 3). As for those with hematological malignancies (n = 4), two had leukemia, and another two had aplastic anemia. The distribution of types of primary cancer is depicted in Figure 4.3.
In terms of cancer stages or severity in relatives of the caregivers, twenty (40%) were of moderate severity, or in Stage 2 and 3 of their illness. This was followed by severe cancer or Stage 4 (38%), and mild cancer or Stage 1 (16%). About two-thirds of the relatives were receiving oncology treatment as outpatient (n = 32), and the rest as inpatient (n = 18). Majority of the relatives were on active treatment (82%), while only two (4%) were on palliative care. Some of the relatives were on neither active nor palliative treatments, as 10% of them were still awaiting treatment plan and 4% of them already completed their treatments.
As for the functional status of the relatives based on ECOG scoring, most of them scored 1 (48%) and 2 (38%), whereas the rest scored 4 (8%), 3 (4%), and 0 (2%) respectively. The ECOG scores of 0 and 1 correspond to relatives that were still able to work or perform house chores, whereas a score of 2 or less indicates the relatives were still capable of caring for themselves. In this study, half of the relatives of the caregivers were able to work or do house chores (50%), but most were still capable of self-care (88%).

U n i v e r s i t y o f M a l a y a
Looking at the caregiving aspects, majority of the caregivers have shared caregiving responsibilities with other family caregivers (86%), with average having two other caregivers assisting in the care of their relatives. In contrast, however, only three caregivers (6%) have shared caregiving in the form of external support such as housemaids or private nurses. The median duration as caregiver in this study was ten months (range 1 -96 months, IqR 14 months). As for the total time spent on caregiving in a week, the median time was 84 hours (range 5 -168 hours, IqR 113.8 hours). Table   4.2 summarizes the clinical characteristics of the study subjects.

Reliability and Internal Consistency of MZBI Scale
The split-half reliability of the MZBI scale was measured using the Spearman-Brown coefficient for equal length-items scale. For split-half of the first 11 items versus the last 11 items, the correlation coefficient was 0.873. As for split-half of the odd items versus the even items, the correlation coefficient was even higher at 0.912, indicating an excellent internal consistency for the MZBI scale.
The high internal consistency of the MZBI scale was further confirmed with a Cronbach's alpha value of 0.898. The highest inter-item correlation coefficient (0.780) was found between item 20 and 21, whereas the lowest (-0.152) was found between item 12 and item 21. The corrected item-total correlations ranged from 0.228 to 0.760 (Table 4.3).
For item 20 and item 21, the corrected item-total correlation value was 0.261 and 0.228 respectively. These low values (i.e. less than 0.30) indicated weak correlations for these two items with the rest of the items in the MZBI scale. Also, Cronbach's alpha coefficient would increase from 0.898 to 0.900 if item 20 was deleted, or to 0.902 if item 21 was deleted from the scale (Table 4.3). The rest of the items were essential as they had higher corrected item-total correlations, and their removal would result in lower Cronbach's alphas.

Factor distribution was based on a study by Bianchi et al. (2016): a Tensions related to the role; b Intra-psychic tensions; c Competencies and expectations EZBI, English version of the Zarit Burden Interview scale; MCES-D, Malay version of the Center for Epidemiologic Studies -Depression scale; MZBI, Malay version of the Zarit Burden
Interview scale; ** p < 0.01; * p < 0.05

Determining Cut-off Score for MZBI Scale
The ROC curve ( indicated that the MZBI scale has fairly good accuracy in differentiating those with significant caregiver burden from those without. The MZBI score with optimal sensitivity and acceptable specificity was chosen as the cut-off score. At MZBI score of 22, the sensitivity and specificity of the scale were 70.8% and 69.2% respectively (Table 4.5). The positive predictive value and the negative predictive value were 68% and 72% respectively, calculated based on the 2 X 2 contingency table as demonstrated in Table 4.6. These values were reasonable in minimizing both false negative and false positive rates. Therefore, a MZBI score of 22 was determined to be the cut-off score in this study, and would be used in subsequent bivariate and multivariate analyses in Phase II.

Clinical Characteristics
Majority of the caregivers in this study were looking after relatives with solid cancer (92.9%), while the rest were caring for relatives with hematological malignancies. Among the solid cancer group, nasopharyngeal cancer (n = 27) was the commonest cancer suffered by relatives of the caregivers. This was followed by colorectal cancer (n = 23), breast cancer (n = 22) and lung cancer (n = 11). As for those with hematological malignancies (n = 9), five had leukemia, three had lymphoma and one had aplastic anemia. The distribution of types of primary cancer among the cancer patients in Phase II is depicted in Figure 4.6.
In terms of cancer stages or severity in relatives of the caregivers, about 54 (42.5%) were of moderate severity, or in Stage 2 and 3 of their illness. This was followed by severe cancer or Stage 4 (32.3%), and mild cancer or Stage 1 (9.5%).
Majority of the relatives were receiving oncology treatment as inpatient (n = 72), and the rest as outpatient (n = 55). Most of the relatives were on active treatment (73.2%), while only ten (7.9%) were on palliative care. Some of the relatives were on neither active nor palliative treatments, as 11.8% of them were still awaiting treatment plan and 7.1% of them already completed their treatments.
As for the functional status of the relatives based on ECOG scoring, most of them scored 1 (43.3%) and 2 (37.8%), whereas the rest scored 3 (10.3%), 4 (4.7%), and 0 (3.9%) respectively. The ECOG scores of 0 and 1 correspond to relatives that were still able to work or perform house chores, whereas a score of 2 or less indicates the relatives were still capable of caring for themselves. In this study, slightly more than half of the relatives of the caregivers were not able to work or do house chores (52.8%), but most were still capable of self-care (85%).    For DUREL-M, the total scores in this study ranged from 7 to 27 with a median score of 23 (interquartile range 5) and a mean score of 22 (standard deviation 4.03). The scale was further described in three separate domains (ORA, NORA, IR) with their respective median and mean scores as shown in Table 4.10. Based on the median score of 23, the study subjects were divided into two groups: those who scored 23 and above U n i v e r s i t y o f M a l a y a the median score (high religiosity), and those who scored below the median score (low religiosity). As shown in Table 4.11, the group with high religiosity (56.7%) constituted the majority over the low religiosity group (43.3%).

Descriptive Statistics for MCES-D, DUREL-M and M-RCOPE Scores
The M-RCOPE was interpreted separately for two of its main subscales (PRC and NRC), as the total scoring is not applicable for this scale. For PRC, the median score was 25 (interquartile range 7) and the mean score was 23.6 (standard deviation 4.54) within a range of scores from 7 to 28. Compared to PRC, the median and mean scores for NRC were considerably lower (median 10, interquartile range 5; mean 11.4, standard deviation 4.52) with scores ranged from 7 to 27 (Table 4.10).
Similarly, the study subjects were divided into two separate groups based on the median scores for PRC and NRC respectively: High PRC group (scored 25 and above) versus low PRC group (scored below 25); and high NRC group (scored 10 and above), versus low NRC group (scored below 10). Both high PRC group (53.5%) and high NRC (57.5%) group were more in numbers than the low PRC group (46.5%) and low NRC group (42.5%) respectively (Table 4.11).  In other words, caregivers who were the children or grandchildren of the cancer patients were significantly less likely to have caregiver burden than caregivers who were not (OR 0.46, CI 0.22 -0.94, p = 0.032). Table 4.12 illustrates the associations between the caregiver burden and the sociodemographic characteristics from bivariate analysis using the Chi-square test.

Association between Caregiver Burden and Clinical Characteristics
Clinical characteristics in relatives of the caregivers such as types of primary cancer, severity of cancer, treatment settings, treatment types, and functional status (ability to work and capacity for self-care); as well as caregiving factors such as presence of shared caregiving, external support, and caregiving duration were examined for associations with the caregiver burden. Two factors in the relatives of the caregivers were found to be significant: Treatment setting and the ability to work.

U n i v e r s i t y o f M a l a y a
On bivariate analysis, significant caregiver burden was more likely among family caregivers who cared for cancer patients that were treated as inpatients than those that were treated as outpatients (OR 2.21, CI 1.08 -4.55, p = 0.03). Also, caregivers who looked after cancer patients that were unable to work experienced significantly more burden than those who cared for relatives that were still able to work (OR 2.43 CI 1.18 -4.99, p = 0.015). Table 4.13 illustrates the associations between the caregiver burden and the clinical characteristics from bivariate analysis using the Chisquare test.

Religious Coping
After creating the dummy categorical variables for depression, religiosity, and religious coping (Table 4.11), the associations between these variables and the caregiver burden variable were then tested using Chi-square test. The finding from this bivariate analysis is summarized in Table 4.14.
Significant positive association was found between caregiver burden and caregiver depression (OR 5.20, CI 2.41 -11.21, p < 0.001). This finding can be interpreted in two possible ways: depressed caregivers are more likely to have significant caregiver burden than non-depressed caregivers, or burdened caregivers are more likely to be depressed than non-burdened caregivers. Caregivers with high negative religious coping (NRC) also experienced significantly more burden than those with low NRC (OR 2.21, CI 1.08 -4.55, p = 0.03). However, no significant relationship was found between caregiver burden with the religiosity level or positive religious coping pattern of the caregiver.

Factors Associated with Caregiver Burden: A Multivariate Analysis
Significant factors obtained from the bivariate analyses of the earlier data were included in the multiple logistic regression analysis to determine the associated factor(s) for caregiver burden after controlling for the confounders. The findings from this multivariate analysis were summarized in Table 4.15.
According to this regression model, the caregivers' relationship of being the children or grandchildren of cancer patients (OR 0.41, CI 0.18 -0.94, p = 0.035), as well as the presence of caregiver depression (OR 4.26, CI 1.87 -9.72, p = 0.001) remained significantly associated with caregiver burden after adjusted for other significant associated factors.
In other words, the odds of having significant caregiver burden in caregivers who were the children or grandchildren of cancer patients were 2.44 times less than the odds in non-children or non-grandchildren caregivers in this study. Also, the presence of depression in the caregivers increased the odds of having significant caregiver burden up to four-folds.
With five significant variables, this logistic regression model explained 27.9% (Nagelkerke R 2 = 0.279) of the variance on caregiver burden and correctly classified 73% of the caregivers. The model fits well to the data with good calibration, as suggested by the Hosmer-Lemeshow goodness-of-fit test that yielded a χ 2 (8) of 3.599 and was insignificant (p = 0.891).  ). This may be due to the increasing number of aging population in developed countries where better healthcare services for elderly citizens are readily available (Baider & Surbone, 2014).

U n i v e r s i t y o f M a l a y a
In this study, 60% of the caregivers in Phase I and about two-thirds of the caregivers in Phase II were women. Majority of them were married. Almost all the previous studies on cancer caregiving showed that female caregivers and married caregivers were more in numbers than male caregivers and unmarried caregivers respectively Meecharoen et al., 2013;Lukhmana et al., 2015;Chua et al., 2016;National Alliance for Caregiving, 2016). The higher number of female U n i v e r s i t y o f M a l a y a caregivers may be due to the differential survival rates for gender that resulted in more women than men in the general population (Romito et al., 2013).
Another plausible explanation is the difference in basic personality traits between men and women, where women in general tend to be more nurturing, and have a greater sense of responsibility and self-sacrifice (Sharma, Chakrabarti, & Grover, 2016). Traditionally, women are also less likely to be employed outside and more likely to stay at home, which lead to women devoting more time to caregiving than men (Sharma et al., 2016). Correlatively, a substantial amount (53.5%) of the female caregivers in the main phase of the study are homemakers.
The ethnicity of the cancer caregivers in the main phase of the study represents the overall ethnic distribution in Sarawak, where the natives or Bumiputera Sarawak are the majority ("State statistics", 2014). Similarly, Christianity was the commonest religion in the study as majority of the Bumiputera Sarawak, especially Ibans and Bidayuhs are Christians. Despite a considerable portion of the caregivers in Phase II of the study were from lower socioeconomic group with total household income less than RM 3,000 per month, and mostly stayed in village houses, none of the caregivers were illiterate. In fact, more than half of them had at least secondary education.
Nevertheless, in Phase I, Chinese caregivers constituted the majority, followed by Bumiputera Sarawak. Majority of the caregivers were from middle socioeconomic group with total household income between RM 3,000 to RM 10,000 per month. Most of them were also employed, had tertiary education, and stayed in terraced houses compared to the caregivers in Phase II. The overall younger age group and higher socioeconomic status among the caregivers in Phase I may be due to the inclusion criteria in which the Phase I caregivers were required to be bilingual. English language U n i v e r s i t y o f M a l a y a proficiency is generally more prevalent among the younger generation and those with higher socioeconomic status.
Phase I also reported a great number of caregivers are looking after their elderly parents, followed by the spouses. The reverse was true for Phase II in which spousal caregivers were more common. Similar findings were found in other caregiver studies done previously (Chua et al., 2016;National Alliance for Caregiving, 2016). This finding may be related to the fact that cancer is predominantly a chronic disease affecting older people (Baider & Surbone, 2014). Thus, it makes sense that most of the cancer caregivers were either the children or the spouses of the elderly cancer patients.

CAREGIVERS
The commonest cancer in patients cared for by the caregivers in this study was different for Phase I and Phase II. In Phase I, breast cancer was the commonest, followed by colorectal cancer and lung cancer. This data depicted well the cancer trend in Malaysia where breast, colorectal and lung cancers were the top three cancer diagnoses between 2007 and 2011 (Manan et al., 2016). In Phase II, however, nasopharyngeal cancer was the commonest, followed by colorectal and breast cancers.
Nasopharyngeal cancer is unique as its rate in Sarawak was twice the rate in Peninsular Malaysia. In fact, nasopharyngeal cancer was the most common cancer type diagnosed among Sarawakian males (Manan et al., 2016).
The cancer stages reported in the present study also followed the national cancer trend in which stage 2 and 3 combined were the most common, followed by stage 4 and U n i v e r s i t y o f M a l a y a stage 1 respectively (Manan et al., 2016). Majority of the cancer patients were on active treatment, especially chemotherapy, due to the study being conducted in a tertiary hospital setting with oncology services. This finding is similar to another hospital-based study by Lukhmana et al. (2015). There were more outpatients than inpatients in Phase I, but the opposite was true in Phase II, presumably due to differences in the sampling process that were based on convenience sampling.
Despite more cancer cases were in moderate to severe stages and needing active chemotherapy, majority of the patients in both phases of the study were still functional and capable of self-care (ECOG score 0 to 2). This finding is comparable to the functional status of cancer patients in another local study that also used the same ECOG scoring (Mahadevan et al., 2013). Nevertheless, there was a slight difference in the number of patients who were still able to work between Phase I and Phase II. In Phase II, cancer patients who could not work were slightly more in numbers than patients who could still work, but the numbers were equal for patients in Phase I.
As for the caregiving characteristics, a great number of the caregivers (86% in Phase I; 78.7% in Phase II) shared their caregiving roles with other caregivers. This seemed to be a universal trend both locally as well as abroad (Jaafar et al., 2014; National Alliance for Caregiving, 2016). On the contrary, only a small number of the local caregivers (6% in Phase I; 0.8% in Phase II) in the present study engaged in the service of housemaids or paid nursing services to look after their relatives, compared to 41% of the caregivers in the United States (National Alliance for Caregiving, 2016).
This trend may be explained by the overall lower socioeconomic status of the local caregivers, as majority of them could not afford the cost of formal care services. The lack of such specialized services locally may be another possibility (Yi, 2013).

U n i v e r s i t y o f M a l a y a
The median duration of caregiving in this study was almost similar for both phases: ten months in Phase I and nine months in Phase II. These durations are much shorter than the median duration of 24 months in a study conducted in Singapore (Chua et al., 2016), as well as the average duration of 22.8 months in the United States (National Alliance for Caregiving, 2016). The longer duration of caregiving in Singapore and the United States may reflect the longer survival rates of cancer patients owing to the medical advances in these two developed nations. However, this notion needs further examination and review in order to establish more evidence.
Comparing the total time spent on caregiving in a week, there is a marked discrepancy between the results from the present study (median time 84 hours in Phase I and 128 hours in Phase II) and the findings (median time ranged from 10 to 20 plus hours) reported in other studies (Hsu et al., 2014;Jaafar et al., 2014;National Alliance for Caregiving, 2016). In fact, about 37.3% of the total 177 caregivers in the current study reported spending full time on caregiving. This finding should be interpreted with caution as the definition and criteria on caregiving hours might be different from one study to another. Nevertheless, the result suggests that cancer caregiving is a labourintensive task.  item 21 were discovered as well in the Chinese version of ZBI, of which the authors suggested for the two items to be considered for deletion (Lu, Wang, Yang, & Feng, 2009).

MZBI SCALE AS A RELIABLE AND VALID TOOL FOR CAREGIVER BURDEN
Nevertheless, both item 20 and item 21 were still retained in the MZBI scale used in this study for two reasons. Firstly, the Cronbach's alpha value in this study was already satisfactory high and did not increase by a large degree from deleting the two items. Statisticians have suggested in keeping an item in a scale if the alpha value only increases by miniscule amount after dropping the said item (Tolmie, Muijs, & McAteer, 2011). Secondly, the squared multiple correlation values for both item 20 and 21 were fairly acceptable, suggesting that these items were still able to predict some of the variability of the other items. Therefore, there is no strong statistical reason to drop item 20 or item 21 from the MZBI scale, although revision of the wording used in the two items may be considered in the future to further improve the reliability of the scale.

U n i v e r s i t y o f M a l a y a
In this study, the MZBI score was positively and significantly correlated to the MCES-D score (rs = 0.58, p < 0.01), which supported the convergent validity of the MZBI. Again, this finding is similar to those found in Hérbert et al. (2000) and Wang et al. (2008), which indicated a strong association between depression and caregiver burden. The high correlation between the MZBI score with EZBI score (rs = 0.84, p < 0.01) further confirmed the validity of the MZBI, as respondents who were bilingual tended to select similar score range for the related items in both scales.
A MZBI cut-off score of 22 derived from this study is suggested for use in local population. This reading is slightly higher than the score 21 recommended in the original scale (Zarit et al., 1985). It is however noteworthy to mention that the score 21 from the original ZBI was not a statistically derived cut-off score, but instead an arbitrary value proposed by the original author (Hérbert et al., 2000). As different populations have unique cultural values and distinct demographic characteristics, it is not unusual to obtain a different cut-off value in this study (Wang et al., 2008).
Higher ZBI cut-off scores ranging from 24 to 26 have previously been identified for different caregiver groups, although the scores were used as predictors for risk of depression rather than caregiver burden in general (Schreiner, Morimoto, Arai, & Zarit, 2006). As caregiver depression has been identified as separate construct from caregiver burden , it is not surprising the cut-off scores used to identify them will be different. Caregiver burden being a multidimensional entity that encompasses various emotional states  will be a wider construct than caregiver depression, thus requiring a much lower range of cut-off scores for identification.
In addition, the use of MZBI scale as a screening tool would mean that a false negative rate (i.e. below the burden cut-off but depressed) is less desired than a false U n i v e r s i t y o f M a l a y a positive rate (i.e. above the burden cut-off but not depressed), thus a cut-off score that results in higher sensitivity rate would be preferred at the expense of specificity (Schreiner et al., 2006). This explains the rationale in choosing a lower cut-off score of 22 in this study, as the value would give a higher sensitivity of 70.8% with a lower but reasonable specificity of 69.2%, compared to higher cut-off values (Table 4.5).

CAREGIVER BURDEN AND DEPRESSION RATE
Using the MZBI cut-off score of 22, the rate of caregiver burden among family caregivers of cancer patients in SGH was estimated at 55.6%. This rate is lower than the cancer caregiver burden rates (ranged from 67.3 to 75%) reported from previous studies conducted in the United States (Kim & Schulz 2008;Hsu et al., 2014), but much higher than the rate (24.6%) reported in another local study conducted in Kuala Lumpur (Mahadevan et al., 2013).
The difference in rates may be explained by the distinct caregiver demographics from each study site, as well as the use of non-standardized individual caregiver burden assessment tools in these studies. For instance, Depression, Anxiety, Stress Scale (DASS) was used by Mahadevan et al. (2013), while Hsu et al. (2014) employed the Caregiver Strain Index (CSI) to assess the caregiver burden. These scales have different scoring systems from the ZBI scale, thus producing vast difference in the burden rates.
There was also no uniform criteria or definition on caregiver burden across the studies.
Example is the use of 'caregiving stress' instead of 'caregiver burden' in Mahadevan et al. (2013).

U n i v e r s i t y o f M a l a y a
A more appropriate study for comparison of the caregiver burden rate with the present study is a similar cross-sectional, hospital-based study by Lukhmana et al. (2015). This study was conducted among 200 family caregivers of cancer patients from Delhi, India, using a validated Hindi version of ZBI that follows the same scoring system as the original English version of ZBI. The rate of caregiver burden reported in this study was 43.5%, a much closer value to the rate found in the current study.
As for the rate of caregiver depression, the present study reported a rather high rate of 59.1% among the family caregivers of cancer patients in SGH, using the MCES-D scale with the cut-off score of 16. This rate was very similar to the caregiver burden rate in the present study as reported above. Compared to other studies with documented prevalence rates of depression between 48.6% and 67% (Ambigga et al., 2005;Rhee et al., 2008), the rate reported in the present study may be considered as the average rate for caregiver depression. However, it is important to note the potential variation in rates due to the different scales used across these studies e.g. Hospital Anxiety and Depression Scale (HADS) used by Ambigga et al. (2005), and Beck Depression Inventory (BDI) used by Rhee et al. (2008).
Careful consideration should be exercised when interpreting the high rate of depression among the local cancer caregivers in the present study. It may be best to view the depression identified in this study as probable depression rather than true depression so as to avoid unnecessary psychiatric diagnosis and unwarranted treatment. Cancer Network as a multifactorial, unpleasant experience of an emotional, psychological, social or spiritual nature that interferes with one's ability to cope with a debilitating condition such as cancer (Holland, 2003). It encompasses a broad spectrum of feelings ranging from normal stress, sadness, and fear to psychiatric conditions such as clinical depression and anxiety (Holland, 2003). Thus, it is entirely possible that some of the cancer caregivers might actually be distressed rather than depressed.
Due to the above findings, it can be deduced that MCES-D scale should be regarded as a screening tool rather than used as an isolated diagnostic measure for depression (Vilagut, Forero, Barbaglia, & Alonso, 2016). Nevertheless, the high rate of caregiver depression present in the current study proved that it is worthwhile to screen for depression among all cancer caregivers. Caregivers who screened positive for depressive symptoms should warrant further thorough assessment, preferably by psychiatrists, to identify those with notable depressive disorders that need psychiatric interventions (Rivera, 2009).
In summary, both caregiver burden and depression rates reported in the present study are high, indicating that cancer caregiving is a relatively burdensome and challenging task, and the well-being of the local cancer caregivers is at stake due to the huge care demands placed on them. The high rates found in this study may in part explained by the vulnerabilities faced by the local caregivers, who were mostly of lower socioeconomic status, and were less equipped with the knowledge and support in caregiving compared to the caregivers from developed countries. Furthermore, the local healthcare systems often neglect the needs of the caregivers, and do little to assist them U n i v e r s i t y o f M a l a y a in looking after their own welfares in addition to those of their loved ones (Cheong & Putit, 2011). The overall high religiosity of the cancer caregivers in the study was also observed in other studies on caregivers of impaired elderly and Alzheimer's patients (Shah, Snow, & Kunik, 2001;Herrera et al., 2009

U n i v e r s i t y o f M a l a y a
Similar to the religiosity level, most of the caregivers (53.5%) in the current study demonstrated high positive religious coping (PRC), although high negative religious coping (NRC) was also common (57.5%). Nonetheless, the median score for PRC was notably higher than the score for NRC (25 versus 10), indicating that more caregivers practiced positive coping strategies rather than negative coping strategies when looking after their relatives. This trend was also present among the cancer caregivers in India, and likely conforming to the religious belief systems of the majority of the Indian population (Thombre et al., 2010).

FACTORS ASSOCIATED WITH CAREGIVER BURDEN
Among the various sociodemographic factors being examined in this study, caregiver's relationship as the children or grandchildren of cancer patients was the only factor found to be significantly associated with caregiver burden. This factor remained significant after controlling for the confounding variables in multivariate analysis (OR 0.41,p = 0.035). The level of significance was not observed however, when examining the association between spousal caregiving and caregiver burden.
According to Wolff and Kasper (2006), nearly 80% of the informal caregivers who care for an older adult aged 65 or older are spouses or adult children, as being demonstrated in the present study where the spouses and the offspring of the cancer patients constituted 85.9% of the total 177 caregivers in both phases of the study.
Spouses are usually the first to provide care (primary caregiver) to an older adult who is sick. Adult children generally function as the secondary caregiver and only step into the main caregiving roles when spouses are not available (Lowenstein & Gilbar, 2000;Revenson et al., 2015).

U n i v e r s i t y o f M a l a y a
The influence of the caregiver's relationship to cancer patients has been studied previously and found to be related to the caregiver burden (Lowenstein & Gilbar, 2000;Ferrario, Cardillo, Vicario, Balzarini, & Zotti, 2004;Given et al., 2004;Hagedoorn et al., 2008). Majority of these studies reported that spousal caregivers have greater caregiver burden compared to adult children caregivers. A notable exception is a study done by Given et al. (2004) who reported the opposite where adult children caregivers were found to have high levels of depressive symptoms and feeling of abandonment (a portion of caregiver burden) when caring for patients with cancer at the end of life.
In the present study, it was discovered that offspring caregivers of cancer patients were significantly less likely to experience caregiver burden than non-offspring caregivers, even though spousal caregivers were not found to be significantly experiencing more burden than non-spousal caregivers as reported in the studies mentioned above. The lack of significant association between spousal caregiving and caregiver burden in this study may be due to the relatively small sample size and the use of non-random sampling method.
There are a few possible explanations on why adult children who care for the cancer patients may experience fewer burdens than other caregiver groups. Firstly, as mentioned before, majority of the offspring caregivers are secondary caregivers as compared to the spouses who are the primary caregivers. As the children only involve in assisting the primary caregiver, the caregiving expectations and workloads are understandably lower in the former. Thus, the adult children are generally less vulnerable to the physical and psychological distress of caregiving (Revenson et al., 2015).

U n i v e r s i t y o f M a l a y a
Secondly, the nuclearization of families following rapid urbanization of Asian societies might also influence the perceived burden by the adult children caregivers. A nuclear family usually comprises of married couple staying together with or without their dependent children (Sharma, 2013). Relevant to the increasing work requirements, many adult children start to live separately from their parents to form a nuclear family of their own. As a result, they are only able to provide distance caregiving when a parent becomes ill, relying much on the help of other family members or formal nursing services (Mazanec, 2012).
Consequently, distance adult children caregivers with adequate social support may experience less direct burden on caregiving. Lowenstein and Gilbar (2000) demonstrated this whereby the children caregivers that stayed in separate households from the cancer patients were found to experience significantly fewer burden than the spouses who resided together with their ill partners. Nevertheless, distance caregivers or adult offspring caregivers may still expose to psychological distress, especially when they lack social support, as they often have multiple commitments such as child-rearing, domestic chores, full time employment and various financial responsibilities (Gaugler et al., 2005;Mazanec, 2012).
Thirdly, the concept of filial piety, a cultural belief that influences adult children's attitudes and behaviours toward their parents, may play an important role in the buffering of care burden faced by the adult children caregivers. In many Asian cultures, Malaysian included, children are often expected to care for their parents in their aging years, more so when the elderly parents are suffering from chronic medical illnesses (Chan et al., 2012). The practice of filial piety includes showing respect and affection toward parents, fulfilling responsibilities, preserving the family harmony, and making payments and other sacrifices (Sung, 1998, as cited in Lai, 2010.

U n i v e r s i t y o f M a l a y a
Various studies have demonstrated the positive effects of filial piety on the caregiving appraisal and subsequently on the caregiver burden (Lai, 2010;Khalaila & Litwin, 2011). According to Lai (2010), filial piety may even serve as protective factor against caregiver burden, which is mediated through other factor such as caregiving appraisal as there was no direct relationship observed between these two variables.
Thus, offspring caregivers who identify more with filial piety are more likely to find caregiving tasks to be positive, beneficial and not burdensome (Khalaila & Litwin, 2011). This will help to reduce the subjective care burden as the caregivers have a more positive evaluation and understanding of their roles and responsibilities.
Coming to the clinical and caregiving characteristics, two factors related to the cancer patients were found to be associated with caregiver burden from bivariate analysis: treatment settings (inpatient versus outpatient) and ability to work (able to work versus unable to work). Generally, cancer patients requiring inpatient hospital treatment, often due to more serious presentation or complications, are more dependent and difficult to care for ( Girgis et al., 2013;Mahadevan et al., 2013). Thus, it is understandably that their caregivers would experience more burden than the caregivers who care for patients from outpatient setting.
Similarly, cancer patients with good functional status (capable of self-care and still able to work) would be relatively easier to care for compared to patients with low functional abilities, as the latter require more time and assistance in their activities of daily living (Jaafar et al., 2014). Various studies have indicated that cancer patients with poor functional status would predict higher caregiver burden due to the increase in the care demands and the longer caregiving hours needed (Weitzner et al., 1999;Gaugler et al., 2005;Mahadevan et al., 2013;Hsu et al., 2014). Nonetheless, most of the above U n i v e r s i t y o f M a l a y a studies did not examine further which aspects of the patients' functionality would best predict the caregiver burden.
In the present study, it was discovered that the patient's ability to work is more important than the patient's capacity for self-care in terms of association with the caregiver burden. The results from bivariate analyses would suggest that employed cancer patients might be associated with fewer burdens in their caregivers, and vice versa. Employed patients can be an indicator that patients have higher functionality and mobility, thus requiring less physical assistance from the caregivers. This would translate to less restriction in caregivers' ability to socialize and participate in valued activities, which in turn reduce their perceived burden (Cameron, Franche, Cheung, & Stewart, 2002). In addition, working patients could contribute financially to ease the economic strain of their caregivers as well (Ghaswalla & Min, 2016). However, after controlling for the confounders in multivariate analysis, both patient factors (treatment settings and ability to work) were no longer significant in the present study.
Presence of depression in cancer caregivers was another important factor that significantly associated with caregiver burden in this study (OR 4.26, CI 1.87 -9.72, p = 0.001). The strong association between these two variables was also indicated by the presence of significant correlation between the MZBI score and the MCES-D score in Phase I of the study. Consistent results were found in other caregiver studies, which support the significant relationship between caregiver depression and caregiver burden (Stommel et al., 1990;Butler et al., 2005;Grov et al., 2006;Rivera, 2009).
As mentioned in previous chapter, the significant relationship between caregiver burden and depression can be bi-directional. In general, caregiver depression has been viewed as a reaction or consequence of the caregiving processes (Rhee et al., 2008).

U n i v e r s i t y o f M a l a y a
Thus, many researchers established that both subjective and objective caregiver burdens, especially following long-term caregiving, might contribute to depression in the caregivers (Rhee et al., 2008;Rivera, 2009). In other words, caregivers with high burden are more likely to be depressed, and vice versa.
However, Stommel et al. (1990) put forward an interesting perspective that the opposite might be true where caregiver depression can predict the caregiver burden.
Simply put, depressed caregivers are more likely to have high caregiver burden. The authors argued that depression, being a rather stable condition, would be a significant influence on the caregivers' perceptions and thus able to explain most of the variation in the perceived burdens (Stommel et al., 1990). This intriguing perspective was confirmed in a study by Lim, Sahadevan, Choo, & Anthony (1999), who concluded that depressed mood in the caregivers of dementia patients was predictive of the caregiver burden and explained about 40% of the variance.
In addition to the direct association, it is possible that caregiver depression may also act as a mediator of indirect influence on the caregiver burden (Grov et al., 2006).
One such example is the finding by Utne, Miaskowski, Paul, & Rustøen (2013), which suggested that depression possibly mediates caregiver burden through the feeling of hope, as hopelessness is known to be an important component of and highly correlated with depression. Nevertheless, the causal relationship between caregiver burden and depression was still not established in the present study due to the relatively small sample size and the cross-sectional design of the study (Pirraglia et al., 2005).
To the best understanding of the author, there is yet any local study done on the association between religious factors and caregiver burden. The role of religion in caregiving can be studied in a multiracial and multireligious environment like Malaysia.

U n i v e r s i t y o f M a l a y a
Unexpectedly, the current study failed to demonstrate any significant relationship between the religiosity and religious coping strategies with the caregiver burden on a multivariate level, although there was a significant association between the negative religious coping (NRC) with the caregiver burden from bivariate analysis. Similarly, there were no significant associations found between the caregiver burden and each of the religion groups in the study. The results were inconsistent with those of the previous studies reporting the association between these variables (Mickley et al., 1998;Pearce et al., 2006;Herrera et al., 2009;Gholamzadeh et al., 2014).
The present study showed that high NRC use, but not positive religious coping (PRC), was associated with higher caregiver burden on a bivariate level. This finding is in concordance with the results from previous literature supporting the association between the NRC, but not the PRC, with the levels of psychological distress or mental health outcomes (Hebert et al., 2009;Herrera et al., 2009;Nurasikin et al., 2012;Ng et al., 2016). Caregivers who felt that God has abandoned them, questioned God's power or considered their caregiving responsibilities as a punishment from God were more likely to report less satisfaction and more burden from caregiving (Pearce et al., 2006).
It is postulated that individuals who struggle with their religious beliefs may not be able to use their faith as a resource for coping, as they see their faith as distant and weak, thus leading to an existential crisis (Ng et al., 2016). Despite the presence of significant association between the caregiver burden and the NRC in bivariate analysis, the association did not remain significant in multivariate analysis, possibly due to the small sample size in the current study.
The relatively high religiosity level and use of PRC strategies among the cancer caregivers in the present study were found to have little or no significant impact on the caregiver burden and depression rate. These findings may be explained by the presence U n i v e r s i t y o f M a l a y a of multiple concomitant stressors, e.g. lower socioeconomic status, lack of caregiving resources or supports, severity of cancer presentation, etc., which greatly overshadow the buffering effects of religiosity and PRC in the current study. Interestingly, the use of PRC strategies may actually be associated with greater caregiver burden as reported by Pearce et al. (2006). The authors hypothesized that PRC may build up one's burden by increasing the possibility that one takes on more number of or more difficult caregiving tasks out of a feeling of religious duty. Also, it is possible that those who are the most burdened may be more likely to turn to religion as a way to cope and to find sacred meaning (Pearce et al., 2006).
Another plausible explanation is the use of non-random sampling method and cross-sectional design in the present study, which limit the understanding on the multidimensional nature of the religious elements (Stefanek et al., 2005). Furthermore, the religious pluralism in Malaysia may pose a challenge for the effective study of religious factors in this respect (Samian, 2010). Consequently, the current study was not robust enough to detect the complex relationships between the religious variables and the caregiver burden or depression.
In general, the findings from the current study supported that the influence of religiosity and religious coping on caregiver burden is still vague and inconsistent (Hackney & Sanders, 2003;Hebert et al., 2006). However, this does not mean that religious factors have no value or positive role at all in the caregiving process. More local research is needed in this field, preferably using a broad-spectrum and standardized measure, to address the multidimensional qualities of religiosity and religious coping, as well as to take into consideration the religious pluralism in Malaysia.
U n i v e r s i t y o f M a l a y a

LIMITATIONS AND STRENGTHS OF THE STUDY
There are several limitations present in this study. First, it was a cross-sectional study, which did not describe the temporal relationship between the caregiver burden and the various factors examined in the study. Thus, causal inferences could not be made between these variables. A longitudinal prospective study may be a better design for this purpose.
This can be done by recruiting a study cohort consists of the caregivers of cancer patients at the time when the patients receive a diagnosis of cancer. The caregivers are then followed up at specific interval, for example 12-weekly, up to 52 weeks or until patient's death. Data collection can be done either through face-to-face clinic appointments or through telephone interviews at the time of recruitment and subsequently during each follow-up. Sociodemographic and caregiving-related information is collected and measures of patient symptoms, caregiver burden, depressive symptoms, religiosity level and religious coping patterns are obtained. With this design, life course of an individual caregiver can be recorded and multiple observations or measurements can be made over time.
Second, convenience or non-random sampling techniques used in this study would generate sampling bias, especially when the study was conducted in a single setting based in government hospital. Therefore, the sample chosen might not represent the general population of the cancer caregivers, which also includes caregivers in the community setting as well as those in the private medical centres. Nonetheless, convenience sampling was chosen as the author faced difficulties in terms of costs, manpower, and time spent on performing the study with random samples.

U n i v e r s i t y o f M a l a y a
Third, a relatively smaller sample size could have contributed to the study having small effect size and wide confidence intervals, which further restricted the generalization of its findings, and the statistical power necessary to detect the complex relationships between the study variables. A study with a larger sample size of family caregivers is needed to obtain a more robust finding. Fourth, the study also could not run away from information or recall bias, as the information obtained was based solely from the caregivers' account for most part of the data collection.
In addition, the study did not differentiate between the physical and the psychological components of caregiver burden, which may have resulted in different patterns of burden reported among the caregivers (Hsu et al., 2014). Last but not least, potential important information about the cancer caregivers and patients that might influence the caregiver burden was not included in the study questionnaires. Thus, some potential confounders for caregiver burden were not measured in this study. Examples include the presence of medical comorbidities, the exact caregiving tasks performed with stricter criteria on caregiving hours, as well as whether the caregivers are staying together or separately from the patients, etc.
As for the validation process of the Malay version of the Zarit Burden Interview (MZBI), one limitation is that the construct validity of the scale was not assessed using the factor analysis as was being done in previous validation studies by Lu et al. (2009) and Chattat et al. (2011). It will be worthwhile to carry out both exploratory and confirmatory factor analyses in the future to enhance the overall validity of the MZBI scale in the context of cancer caregiver burden.
The use of MCES-D scale to determine the rate of caregiver depression in this study may also lead to overestimation of the findings (Vilagut et al., 2016). Also, as U n i v e r s i t y o f M a l a y a MCES-D scale was already used to determine the cut-off score for MZBI in Phase I of the study with the assumption that burden is well correlated with depression, the use of the same scales in Phase II to study the association between caregiver burden and depression would not be appropriate, despite the study subjects did not overlap between the two phases of the study.
It is recommended that a different scale for measurement of depression should be used in Phase II to explore the association between caregiver burden and depression in order to avoid bias. Future studies on this aspect should utilize valid diagnostic tool such as the Mini International Neuropsychiatric Interview (MINI) for a more accurate estimation of the prevalence of depression among the caregivers. Addition of distress scale e.g. Distress Thermometer, General Health Questionnaire or Brief Symptom Inventory can also be done to measure the distress level of the caregivers in order to ascertain whether they are truly depressed or just feeling distressed.
On the other hand, the current study demonstrated a few notable strengths as well. Despite being a single-centred study, it was conducted in a tertiary referral centre for oncology services with cancer patients and caregivers originated from various parts of the Sarawak state. This, together with the multiracial background of the study subjects, made up for the diversity of the samples. The cross-sectional design has its fair share of advantages as well, in which it allows the study to measure the prevalence of caregiver burden and depression, and to explore multiple factors and outcomes at the same time. This helps to bring into perspective the resulting effects of cancer caregiving and the caregiving problems relevant to the local Malaysia population in a short period of time.

U n i v e r s i t y o f M a l a y a
Based on the available local data to date, this study also prides itself on being the first local study to validate the Malay version of the Zarit Burden Interview for use on the local population. On top of that, the study was the first to examine the context of caregiver burden among the local family caregivers of multiracial background and of various types of cancer patients, not limiting to just a single-race caregivers as studied by Cheong and Putit (2011), or a single type of cancer patients as studied by Mahadevan et al. (2013). The findings from the present study would add to the growing literature on cancer caregiving and caregiver burden in the local setting. In a general context, family caregivers often received little attention from the healthcare providers compared to the cancer patients that they are caring for (Payne et al., 1999). As a result, many of the caregivers have unmet needs, both physical as well as psychosocial needs, which manifested as caregiver burden and/or depression following the untoward effects of the cancer caregiving process (Romito et al., 2013).
Therefore, the findings from this study should alert the local healthcare personnel about the importance of addressing the needs and concerns of the caregivers.
Studies found that majority of the family caregivers were not experienced and had no formal training in cancer caregiving (Northouse et al., 2012;Lukhmana et al., 2015;Chua et al., 2016). According to the National Alliance for Caregiving (2016), as many as 84% of the cancer caregivers in the United States have indicated that they need more help and information on at least one caregiving-related subject. The caregivers need to be educated on important topics pertaining to patient's illness features, treatment options, specific care needs, managing own emotional stress, and making endof-life decisions, etc.
Training on specific caregiving tasks such as transferring patients, feeding with nasogastric tubes, dealing with incontinence, and cleaning of bodily secretions will help to guide the inexperienced caregivers, and also to reduce both the caregiver burden and U n i v e r s i t y o f M a l a y a depression (Butler et al., 2005). In addition, exposure to soft skills training such as effective communication skills, problem-solving skills, self-motivation, and financial management may be beneficial as well (Northouse et al., 2012). Caregivers should also be taught on how to seek help and engage other relatives or outside resources to relieve their burden. In this regard, participation in the caregiver support groups would be very useful and should be encouraged.
The high rate of caregiver depression and its significance in the prediction of caregiver burden in the study point to a need to actively screen and identify those distressed caregivers at risk of developing depression. Unrecognised depression will have a devastating effect on the physical and psychological well-beings of the caregivers . It is not uncommon to find high rates of suicidal ideation and suicide attempts among depressed caregivers of cancer patients (Park et al., 2013). Hence, it is important that psychiatric consultations and mental health services should be made available and easily accessible to the caregivers at any time by the policymakers. Counseling and psychoeducation about caregiver distress, depression, and treatment strategies will help caregivers to better cope with these conditions. If needed, antidepressant medications and/or psychotherapeutic interventions may be provided based on the recommendations by the psychiatrists.
Caregiver's relationship of being the offspring of cancer patients was found to be significantly associated with caregiver burden in this study, with an interesting finding that children or grandchildren caregivers experienced fewer burdens than nonchildren or grandchildren caregivers. This finding from the present study would highlight the need for the policymakers, researchers, and healthcare providers to devote more attention to the non-offspring caregivers on top of the general focus given to the cancer caregivers as a whole.

U n i v e r s i t y o f M a l a y a
Specific caregiver interventions and social policies should be implemented accordingly to address the needs of the non-offspring caregivers and to improve their caregiving experiences without neglecting the well-beings of the offspring caregivers.
In particular, more educational programmes that provide information on direct care should be made available to these caregivers with relevance to the finding in the research report by the National Alliance for Caregiving (2016). Healthcare providers should also go beyond their clinical duties by linking the caregivers with outside community services and resources for sustained support (Hoffman & Zucker, 2016).
Another way to help relieve caregiver burden is by providing respite care services to these caregivers, especially to those full-time caregivers who stay together with the patients (Mahadevan et al., 2013; National Alliance for Caregiving, 2016).
These services should be made affordable with assistance and subsidies from both the government and non-governmental organizations taking into account the socioeconomic status of the local caregivers. Various respite services or approaches suitable in the cancer context could be explored. Apart from short-term nursing home placement, brief daycare programme that provides temporary relief for the caregivers is another promising option.
Similar to many Asian countries with deep-rooted traditional and religious value systems, adult children in Malaysia often have strong cultural obligations to look after their sick parents as a form of filial duty (Chan et al., 2012). The use of culturally validated tool such as the MZBI would lead to a more appropriate assessment of their caregiving difficulties in future studies. Following that, the relevant policy planners should establish services that provide culturally relevant support systems to the family caregivers (Khalaila & Litwin, 2011). Example is the provision of psychoeducation and counselling in a relevant language catered to the language and cultural background of U n i v e r s i t y o f M a l a y a the caregivers (Lee & Yim, 2013). The cultural influence of filial piety, which possibly explain the fewer burden experienced by the offspring caregivers in the present study, should be encouraged and promoted through the media and academic platforms.
Despite the roles of religion on cancer caregiving are not very clear from this study, the presence of high religiosity and religious coping use among the cancer caregivers in this study should be an encouraging finding that religious faith may still be a valued resource with immense potential waiting to be explored. Future studies to validate the findings from the present study may be helpful and should be designed with an aim to develop more evidence-based strategies to empower the family caregivers and to alleviate their burdens.

SUMMARY
In conclusion, this study confirmed that cancer caregiving is burdensome. A relatively high rate of caregiver burden and depression was observed among the local family caregivers of cancer patients. Caregivers who were the children or grandchildren of cancer patients were experiencing significantly fewer burdens than the non-offspring caregivers. On the other hand, caregivers who had probable depression were more likely to experience significant caregiver burden. The Malay version of the Zarit Burden Interview (MZBI) is a reliable and valid tool for the assessment of caregiver burden among this group of caregivers. The significant risk factors identified in this study can serve as a stimulus for future intervention strategies to be planned and tailored accordingly so as to target the multiple dimensions of the caregiver burden.