Abstract
Context. Caring for a family member with advanced cancer strains family caregivers. Classification of family types has been shown to identify patients at risk of poor psychosocial function. However, little is known about how family relationships affect caregiver psychosocial function.
Objectives. To investigate family types identified by a cluster analysis and to examine the reproducibility of cluster analyses. We also sought to examine the relationship between family types and caregivers' psychosocial function.
Methods. Data from 622 caregivers of advanced cancer patients (part of the Coping with Cancer Study) were analyzed using Gaussian Mixture Modeling as the primary method to identify family types based on the Family Relationship Index (FRI) questionnaire. We then examined the relationship between family type and caregiver quality of life (Medical Outcome Survey Short Form), social support (Interpersonal Support Evaluation List), and perceived caregiver burden (Caregiving Burden Scale).
Results. Three family types emerged: Low-expressive, detached, and supportive. Analyses of variance (ANOVA) with post hoc comparisons showed that caregivers of detached and low-expressive family types experienced lower levels of quality of life and perceived social support in comparison to supportive family types.
Conclusion. The study identified supportive, low-expressive and detached family types among caregivers of advanced cancer patients. The supportive family type was associated with the best outcomes and detached with the worst. These findings indicate that family function is related to psychosocial function of caregivers of advanced cancer patients. Therefore, paying attention to family support, and family members' ability to share feelings and manage conflicts may serve as an important tool to improve psychosocial function in families affected by cancer.
Introduction
Family caregivers are important to advanced cancer patients' well-being throughout the course of their illness. Patients depend on caregivers for emotional, financial, and material support, transportation to and from medical appointments and treatment, and assistance with activities of daily living, including dressing and feeding (1). In the context of advanced illness, caregivers are often providing for patients' medical needs and making decisions about end-of-life care (2). As a result, caregivers are critically important to the care and well-being of cancer patients.
A substantial body of research has investigated the impact of caregiving on caregiver mental and physical health (3). Family caregivers of patients with advanced cancer are at increased risk of experiencing poor quality of life (4), depression (5-7), impaired sleep (8) and heightened social isolation (9). Prior research indicates that caregiving is a multifaceted concept in which objective factors such as the number of caregiving tasks, and subjective factors such as the motivations behind the care provision may affect caregivers' well-being (5). For instance, caregivers motivated by autonomous reasons for providing care, such as seeking comfort from being close to others, tend to have higher levels of life satisfaction than caregivers with extrinsic reasons for providing care, such as seeking to avoid disapproval from his or her social group (10). Family dynamics may influence the motivations behind, and mental health consequences of, caring for a dying family member.
In families of cancer patients, poor family functioning has been associated with increased risk of depression and anxiety (11, 12). In a cross-sectional study of 191 patient-caregiver dyads, lower levels of cohesiveness (family members' involvement in the family), expressiveness (family members' willingness to express their emotions openly) and conflict resolution (family's ability to solve disagreements) predicted depressive symptoms in both lung cancer patients and their caregivers (13). Previous studies investigating family functioning have shown that cohesiveness is the primary indicator of well-functioning families types (13-17). However, little is known about the relationship between family function and caregiver burden and well-being.
The Family Relationship Index (FRI) (18) assesses the cohesiveness, expressiveness, and conflict resolution skills of family units. In the field of health psychology, cluster analysis is being utilized to identify ‘at risk’ groups whose members might benefit from preventative interventions. In relation to the FRI, the cluster solution is generated based on the 3 FRI subscales (cohesiveness, expressiveness, conflict resolution). Cluster analysis of the FRI has been conducted in American, Japanese and Australian samples (14, 16, 17). The number of components generated from the cluster solutions varies from a 3 cluster solution (16) to a 5 cluster solution (12, 14, 17).
One of the main obstacles to consensus in studies employing cluster analyses is the variety of cluster methods being used (i.e. cluster selection method and software package) (19). Inconsistent cluster analysis methodology makes reproducing results difficult. For this reason, we employ the same cluster method as used by one the main studies published on this topic – an investigation by Schuler et al. (2014) (17). In addition to validating previous findings, this study will investigate differences in caregiver well-being and caregiving burden across family types (20-24).
Thus, the present study aims to identify family type clusters in an American sample of caregivers of terminally ill cancer patients based on the FRI and then to examine the relationship between these clusters and caregiver quality of life, social support and caregiving burden. Finally, this study aims to examine the reproducibility of cluster analyses. We hypothesize that the cluster model will generate five clusters as in Schuler et al. (2014) given the similarities in the samples and cluster method used (17). In line with previous research, we hypothesize that caregivers of family types with low levels of cohesiveness will have the most psychosocial dysfunction, whereas caregivers of family types with high levels of cohesiveness will have the least.
Methods
Detailed design and methods for this study have been described elsewhere and are summarized below (25).
Sample and setting
The current study uses data from the Coping with Cancer (CwC1) study (26). Participants in the CwC1 study were recruited between 2002 and 2008 from outpatient clinics in Connecticut, Massachusetts, New Hampshire, New York, and Texas. Patients who were diagnosed with distant metastasis and disease that was refractory to first line chemotherapy met the inclusion criteria. Family caregivers were identified by the patients as the person who provided most of their informal care. Caregivers were required to speak English or Spanish. Caregivers younger than 21 years old, or with significant cognitive impairment (Short Portable mental Status Questionnaire score ≥6) were excluded.
Procedure
This study was approved by the Institutional Review Board of each site, and all participants provided written informed consent. Patients and caregivers were interviewed separately by trained research staff. Patients were followed from baseline until death (median 4 months), and caregivers were interviewed at baseline and 6 months after the patients' death. The current study reports on data collected pre-loss at the baseline assessment.
Measures
Demographic data included the caregiver's self-reported age, gender, ethnicity, marital status, relationship to the patient, and the patient's self-reported stage of diagnosis, and presence of metastatic disease.
The Family Relationship Index (FRI) is a measure of caregiver perception of family functioning, and the scale derived from the short-form of the Family Environment Scale (15). The FRI consists of 12 true/false items divided into 3 subscales (4 items per subscale): Cohesion, Expressiveness, and Conflict Resolution. Each subscale score ranges from 0 to 4 with higher scores indicating better family functioning. The Cohesion subscale captures the sense of family togetherness, and the willingness to give time to the family (18). The Expressiveness subscale measures family members' capacity to express their feelings to each other. Finally, Conflict Resolution measures the family's ability to resolve disagreements. The FRI is well-validated in cancer populations (15, 27).
Quality of life was assessed using the Short Form Health Survey (SF-36). It consists of 36 items, which are evaluated on eight subscales. The six subscales used in this study where the scales that where most strongly correlated to physical health and mental health in a previous study (28): Physical Functioning (PF), Role Physical (RP), Bodily Pain (BP), Role Emotional (RE), Mental Health (MH), and Social Functioning (SF). The SF-36 is well-validated, and has been used in a number of chronic illness and cancer studies (4, 29). Higher scores on the subscales indicate better quality of life.
The Interpersonal Support Evaluation List (ISEL) was used to assess caregiver perceptions of social support. The ISEL (40 items) is made up of four subscales - Tangible Support, Belonging Support, Appraisal Support, and Total Support- assessing different types of support (30). The subscales Tangible Support, Belonging Support, and Appraisal Support have demonstrated adequate psychometric properties in previous investigations of the Coping with Cancer cohort (31-33), thus, these subscales will be used in the current study. Tangible Support refers to the amount of available instrumental support or material aid. Appraisal Support measures the perceived availability of someone to talk about personal problems (e.g. “when I need suggestions on how to deal with a personal problem, I know someone I can turn to”). The Belonging subscale assesses the availability of someone to engage in activity with.
The Caregiving Burden Scale (CBS) was included to measure strain due to caregiving tasks and comprises 2 subscales: the Demand subscale and the Difficulty subscale. The Demand subscale measures the amount of time spent by the caregiver on a variety of caregiving tasks. The Difficulty subscale measures caregivers' perception of the difficulty of each task. The items are answered on a 4-point Likert scale, where 1 is “little or no” and 4 is “a great deal”. The CBS is well-validated and has been used in previous cancer studies (1).
Statistical Analyses
Analyses to derive the family clusters subtypes proceeded in two stages. First, a dataset was constructed that included all participants for whom complete FRI data were available and these data were used in the cluster analysis. Second, a confirmatory dataset was constructed to test the validity of the cluster solution (19, 34). A number of factors influence cluster solutions, such as the software program used and cluster selection method etc., hence, comparisons across cluster solutions may be difficult (19). We decided to test for the optimal number of clusters, and to construct the cluster solution using the same procedure as described in Schuler et al. (2014) (17). Schuler et al. (2014) employed two steps in their cluster analysis. The first step consists of determining the optimal number of clusters using an agglomerative hierarchical clustering approach employing the Hartigan's Index score. NbClust was used as the cluster software program. The second step consists of constructing the clusters using a model-based approach. In order to perform an internal validity check (34) we conducted Analysis of variance (ANOVA) tests with post hoc comparison tests (Tukey) assessing differences in the FRI scores across groups. In order to perform an external validity check (34) of the cluster solution, Chi-square, Fisher's exact test or ANOVA analyses were performed to assess differences in the demographic variables, quality of life, social functioning, and caregiving burden across family types (19). Analyses were conducted using the statistical software R Core Team (35). The R package mclust version 4.4 (36, 37) was used to conduct the cluster analyses.
Results
The analyzed sample size was 622 caregivers. An examination of Hartigan's Index of the solutions with 2 to 15 clusters indicated that a solution with 4 clusters had the highest Hartigan's Index score (Hartigan's Index score=129.11). As shown in Figure 1 a rather small cluster (24 subjects) emerged, with nearly zero standard deviation on cohesion, expressiveness and conflict resolution and the mean parameter estimate at the upper end of the distribution. This cluster was essentially an artifact created by the upper limit (“ceiling”) of the scales. The method cannot distinguish between caregivers, who have a true score above the limits of the scale and accordingly these are all grouped in one cluster with a very low standard deviation. As this cluster is created by aspects of the data not related to family relations, we excluded this cluster from subsequent analyses. Thus, the final analyzed sample size was 598 caregivers, and a 3-cluster solution was accepted. We labeled the 3 family types: Low-expressive, detached, and supportive (Table 1).
Figure 1. Classification plot by mclust.
Pairs plot of the dataset showing classification into family types. All cross plots of the measures (expressiveness, cohesion and conflict resolution) are given. Each measure is in separate plots presented both on the x-axis and on the y-axis. Colors and symbols indicate family type (i.e. latent class membership) and the four ellipsoids in each plot indicate the two-dimensional normal distribution for each of the four latent classes.
Table 1.
Examining the Validity of the 3-cluster model based on Comparison Across FRI-Subscales (N=598).
| CLUSTER 1 “Low-expressive” |
CLUSTER 2 “Detached” |
CLUSTER 3 “Supportive” |
ANOVA PROBABILITY | ||
|---|---|---|---|---|---|
|
| |||||
| N=231(39%) | N=144(24%) | N=223(37%) | |||
|
| |||||
| Mean (SD) | Mean (SD) | Mean (SD) | F ratio | P-value | |
| FRI subscale | |||||
| Cohesiveness | 3.30(0.28) | 2.84(0.55) | 3.84(0.15) | 398.8 | P<0.0011 |
| Expressiveness | 2.75(0.54) | 2.71(0.58) | 3.13(0.45) | 40.79 | P<0.0012 |
| Conflict resolution | 3.35(0.37) | 2.42(0.64) | 3.33(0.47) | 195.13 | P<0.0013 |
Post hoc, Tukey. All clusters are significantly different.
Post hoc, Tukey. The following clusters are significantly different 3-1, 2-3.
Post hoc, Tukey. The following clusters are significantly different 1-2 , 2-3.
Internal validation
Differences in FRI scores across family types
Low-expressive
Low-expressive family types accounted for 39% of the sample and were characterized by significantly higher cohesion (mean=3.30, SD=0.28) and conflict resolution scores (mean=3.35, SD=0.37) than that of the detached family types. Caregivers of the low-expressive families showed significantly lower scores on expressiveness (mean=2.75, SD=0.54), and cohesion, than those of the supportive family types.
Detached
The detached family types constituted 24% of the sample. The detached family types had the significantly lowest cohesion scores of all types (mean=2.84, SD=0.55). The caregivers of the detached family types also demonstrated the significantly lowest score on conflict resolution (mean=2.42, SD=0.64). Caregivers of detached families showed a significantly lower score on expressiveness (mean=2.71, SD=0.58) than that of the supportive family types.
Supportive
The supportive family type accounted for 37% of the sample. This type was characterized by the significantly highest scores on the cohesion (mean=3.84, SD=0.15), and expressiveness (mean=3.13, SD=0.45) subscales in comparison to the remaining family types. Caregivers of supportive families showed a mean conflict resolution score (mean =3.33, SD=0.47) similar to that of the low-expressive families, yet better than that of the detached family types.
External validation
Descriptive statistics across the different family types are presented in Supplementary Table 1, whereas descriptive data on the entire sample are illustrated in a Supplementary Table 2. In terms of socio-demographic or medical variables (caregiver's age, gender, ethnicity, marital status, relationship to the patient, patient's stage of diagnosis, and patient's metastases) caregivers of the detached family types had a significantly lower mean age score (mean=50.27, SD=14.28) than that of low-expressive families. No significant differences emerged on the remaining socio-demographic variables.
Low-expressive vs. detached
No significant differences between detached types vs. low-expressive types were found for the caregiving burden demand subscale. Yet, the detached types experienced more difficulties in carrying out caregiving tasks compared to the low-expressive types. Caregivers of low-expressive and detached families differed on 3 out of 6 quality of life subscales. Specifically, detached family types showed significantly lower levels of physical functioning, role emotional, and mental health scores than the low-expressive. The detached types experienced lower levels of social support on the belonging and tangible subscales score compared to the low-expressive family types.
Low-expressive vs. supportive
No significant differences between the low-expressive and supportive family types were found for caregiving burden and physical health (Table 2). However, in terms of mental health the low-expressive types experienced poorer functioning in role emotional, mental health, and social functioning than the supportive family types. Moreover, the low-expressive family types showed lower levels on all support subscale than the supportive family types.
Table 2.
Examining the External Validity of the 3-cluster model based on Comparison of Family Caregivers' scores on Caregiving Burden Scale (CBS), Medical Outcome Survey (SF-36), and Interpersonal Support Evaluation List (ISEL) across Clusters.
| CLUSTER 1 “Low-expressive” |
CLUSTER 2 “Detached” |
CLUSTER 3 “Supportive” |
ANOVA | PROBABILITY | ||||
|---|---|---|---|---|---|---|---|---|
|
| ||||||||
| N I (%)II |
Mean (SD) | N II (%) II |
Mean (SD) | N II (%) II |
Mean (SD) | F ratio | p-values | |
| CBS scales | ||||||||
| Demand | 223 (97%) |
2.33(0.63) | 142 (99%) |
2.49 (0.70) | 216 (97%) |
2.41 (0.69) | 2.20 | 0.11 |
| Difficulty | 181 (78%) |
1.36 (0.63) | 110 (76%) |
1.51 (0.70) | 180 (81%) |
1.29 (0.69) | 7.54 | P<0.0014 |
| Physical health | ||||||||
| Physical Functioning | 226 (97%) |
87.23 (18.97) | 143 (99%) |
81.57 (25.26) | 220 (99%) |
88.89 (18.69) | 5.72 | P<0.0015 |
| Role physical | 225 (97%) |
82.44 (39.38) | 143 (99%) |
76.40 (33.58) | 221 (99%) |
88.69 (29.20) | 6.09 | P<0.0016 |
| Bodily Pain | 227 (98%) |
80.24 (22.51) | 144 (100%) |
75.23 (25.98) | 221 (99%) |
84.38 (20.26) | 7.18 | P<0.0017 |
| Mental health | ||||||||
| Role Emotional | 227 (98%) |
70.04 (41.05) | 144 (100%) |
56.71 (42.65) | 223 (100%) |
79.97 (35.05) | 15.34 | P<0.0018 |
| Mental Health | 225 (97%) |
69.19 (18.50) | 144 (100%) |
62.31(21.74) | 220 (99%) |
74.36 (18.11) | 17.19 | P<0.0019 |
| Social Functioning | 227 (98%) |
75.99 (25.98) | 144 (100%) |
71.09 (27.24) | 220 (99%) |
82.61 (24.43) | 9.17 | P<0.00110 |
| ISEL Scales | ||||||||
| Belonging Support | 226 (97%) |
13.62 (2.38) | 142 (99%) |
13.06 (2.74) | 220 (99%) |
14.94 (1.71) | 34.84 | P<0.00111 |
| Appraisal Support | 226 (97%) |
12.08 (1.42) | 144 (100%) |
11.83 (1.61) | 220 (99%) |
12.54 (1.13) | 13.00 | P<0.00112 |
| Tangible Support | 226 (97%) |
13.78 (2.46) | 142 (99%) |
12.96 (2.67) | 216 (97%) |
14.72 (1.70) | 26.46 | P<0.00113 |
N=number of participants in the cluster,
%= percentage of participants from the complete case dataset included in the analyses
Post hoc, Tukey. The following clusters are significantly different 1-2, 2-3.
Post hoc, Tukey. The following clusters are significantly different 1-2, 2-3.
Post hoc, Tukey. The following clusters are significantly different 2-3.
Post hoc, Tukey. The following clusters are significantly different 2-3.
Post hoc, Tukey. The following clusters are significantly different 1-2, 1-3, 2-3.
Post hoc, Tukey. The following clusters are significantly different 1-2, 1-3, 2-3.
Post hoc, Tukey. The following clusters are significantly different 1-3, 2-3.
Post hoc, Tukey. The following clusters are significantly different 1-2, 1-3, 2-3.
Post hoc, Tukey. The following clusters are significantly different 1-3, 2-3.
Post hoc, Tukey. The following clusters are significantly different 1-2, 1-3, 2-3.
Post hoc, Tukey. The following clusters are significantly different 1-2, 1-3, 2-3.
Detached vs. supportive
In terms of difficulties in carrying out caregiving tasks, detached types experienced more difficulties compared to the supportive types. The detached family types reported significantly lower levels of quality of life scores (all mental health subscales and physical health subscales) compared to supportive family types. Finally, the detached types experienced lower levels of social support on all subscales (belonging, appraisal, and tangible) compared to the supportive family types.
Discussion
This study identified 3 family types: low-expressive, detached, and supportive. The low-expressive cluster characterized the largest proportion of caregivers, followed by the supportive and detached clusters. Caregivers of detached families reported the lowest levels of mental and physical health and the greatest difficulty performing caregiving tasks. Caregivers of supportive families reported the highest levels of mental and physical health relative to the other clusters. Low-expressive families reported intermediate levels of physical and mental health. These findings indicate that, based on the perspective of the caregiver, families of advanced cancer patients vary their functioning and that the way in which a family functions has implications for caregivers' mental and physical health and caregiving burden.
Reproducibility of the cluster analyses methods
To the best of our knowledge, this is the first study to replicate the cluster analyses method used in Schuler et al. (2014) (17). Despite the methodological similarities between the cluster analyses found in Schuler et al. (2014) (17), and those employed in the current study, Schuler et al. (2014) identified 5 clusters, and we identified 3 (plus one statistical artifact cluster).There are several possible explanations for differences across the studies. First, our study sampled from multiple sites across the United States while the study by Schuler et al. (2014) recruited only from Memorial Sloan-Kettering Cancer Center in New York City. Second, Schuler et al. (2014) (17) investigated patients whereas we sampled caregivers. Third, perhaps because it was a patient sample, the Schuler sample had poorer function in terms of cohesion, expressiveness, and conflict resolution than our sample. Moreover, none of the clusters in the current sample had mean scores on cohesiveness, expressiveness and conflict resolution below 2, which is in contrast to the clusters generated in Schuler et al. (2014) (17). The difference in the number of clusters identified and their properties also reflects an inherent shortcoming of cluster analyses namely that they are sensitive to both precise method choice and data foundation. In the present analysis we made every effort to use exactly the same methods and control parameters (i.e.. type of cluster analysis) as in Schuler et al. (2014) (17). However, differences in the cluster solution may also simply be a consequence of the inherent sensitivity of cluster analyses methods.
Characterization and needs assessment of the different family types
Low-expressive or “threshold” families
Edwards and Clark (2005) conducted a sensitivity-specificity analysis on the FRI, and in this analysis families are “deemed at risk if one or more member records a total FRI score equal to, or less than 9” ((11); p.549). According to Kissane et al. (1994/2003) (14, 15) and Edwards and Clark (2005) (11) caregivers of low-expressive families with a total FRI score above 9 or a cohesiveness score above 3 should not be deemed at risk of a maladaptive outcome. Low-expressive families fell above these thresholds and they reported better quality of life, less caregiving burden, and higher levels of social support than detached families. However, they exhibited lower levels of social support and mental health than supportive family types. In sum, low-expressive families may be characterized as a “threshold” cluster, because they do not function extraordinary well, but neither do they function very poorly. Likewise, previous studies have identified these “threshold” clusters; the low-communicating families in Schuler et al. (2014) (17), the intermediate in Kissane et al. (2003) (15), and the intermediate families in Ozono et al. (2005) (16). These families might be overlooked by clinicians due to their “threshold” status. However, they do experience psychosocial distress and may therefore benefit from additional support and psychosocial care. Because the low-expressive families had low score on expressiveness clinicians may need to make more strenuous efforts at communicating with such families. Interventions directed to low-expressive families might aim at improving their ability to share feelings and concerns.
Detached families or “at risk” families
The detached families had a cohesiveness score below 3 and, therefore, are at risk for psychosocial problems. The low score on cohesiveness suggests that these families refrain from prioritizing spending time within the family. Previous studies have identified clusters similar to detached families, such as the less-involved families in Schuler et al. (2014) (17), and the intermediate in Ozono et al. (2005) (16). The low cohesiveness of detached families may be conceptually similar to attachment insecurity, explaining the difficulties in caregiving (24). Previous studies have identified a significant association between higher levels of caregiving difficulties and higher levels of attachment insecurity (24). If the caregiver feels insecure, then care for the other will be impaired. This study suggests that health care providers should aim at building trust in the relationship with the caregiver, and to improve caregiver skills and well-being.
Supportive families or the well-functioning families
The supportive families fell above the threshold, and they may be characterized as the best adjusted families. Similarly, previous studies have identified a family type as most adaptive; Schuler et al. (2014) identified the mutually supportive (17) (FRI total above 10), Ozono et al. (2005) (16) identified the supportive (FRI total above 10), and Kissane et al. (1994) also identified a family type called the supportive (FRI total above 10) (14). Previous studies have indicated that high levels of cohesiveness are predictive of families' adaptive coping strategies (14). Consistent with these findings, the supportive families in the current study reported the highest levels of psychosocial well-being.
Clinical implications and methodological reflections
The present study suggests that assessing family function may assist in the identification of caregivers at risk for psychosocial problems. The FRI may be a possible screening tool for the detection of family functioning styles associated with caregiver distress. However, additional research on the sensitivity and specificity of the FRI for detecting caregiver distress is necessary before it can be used in clinical settings to identify at-risk caregivers.
Cluster analyses of the FRI generated a cluster with a zero standard deviation and mean parameter at the maximum of the scales (Table 3). Similarly, the supportive families in Kissane et al. (1994) (14) and Ozono et al. (2005) (16) both showed zero standard deviation and mean parameter at the maximum of the scales on cohesiveness (both Ozono et al. (2005) and Kissane et al. (1994)) and conflict (only Kissane et al. (1994)) similar to our cluster, which we interpreted as a statistical artifact. This study suggests that the zero-standard deviation clusters should be better interpreted as statistical artifact from a ceiling effect caused by the methodology employed.
Table 3.
Comparison of Family Relationship Index scores across studies investigating cluster solutions.
| FRI subscale | ||||||
|---|---|---|---|---|---|---|
| Source | “Name” | Proportion (%) | Cohesiveness Mean (SD) | Expressiveness Mean (SD) | Conflict* Resolution | Total FRI |
| CLUSTER | ||||||
| Nissen et al. | Supportive | 37 | 3.84(0.15) | 3.13(0.45) | 3.33(0.47) | 10.3 |
| Ozono et al. 2005 | Supportive | 34 | 4.00(0.00) | 3.27(0.90) | 0.82(0.74) | 10.45 |
| Kissane et al. 2003 | Intermediate | 50 | 3.66(0.54) | 2.45(1.03) | 0.70(0.89) | 9.41 |
| Schuler et al.2014 | Mutually Supportive | 50 | 3.79(0.41) | 3.52(0.50) | 3.43(0.69) | 10.74 |
| Kissane et al.1994 | Supportive | - | 4.00(0.00) | 2.64(1.06) | 0.00(0.00) | 10.64 |
| CLUSTER | ||||||
| Nissen et al. | Low-expressive | 39 | 3.30(0.28) | 2.75(0.54) | 3.35(0.37) | 9.4 |
| Ozono et al. 2005 | Intermediate | 32 | 2.46(0.79) | 2.27(1.34) | 0.65(0.61) | 8.08 |
| Kissane et al. 2003 | Hostile | 23 | 2.71(1.24) | 1.35(1.16) | 1.65(1.33) | 6.41 |
| Schuler et al.2014 | Low-communicating | 21 | 3.69(0.46) | 1.57(0.62) | 3.63(0.48) | 8.89 |
| Kissane et al.1994 | Ordinary | - | 3.00(0.00) | 1.90(1.07) | 0.96(1.13) | 7.94 |
| CLUSTER | ||||||
| Nissen et al. | Detached | 24 | 2.84(0.55) | 2.71(0.58) | 2.42(0.64) | 7.97 |
| Ozono et al. 2005 | Conflictive | 34 | 2.82(1.06) | 2.76(1.08) | 3.12(0.67) | 6.54 |
| Kissane et al. 2003 | Sullen | 29 | 3.22(0.80) | 1.93(1.14) | 1.39(1.12) | 7.76 |
| Schuler et al.2014 | Conflict resolving | 23 | 3.59(0.49) | 2.16(0.95) | 1.65(0.55) | 7.4 |
| Kissane et al.1994 | Resolvers | - | 4.00(0.00) | 2.27(1.11) | 1.87(0.83) | 8.4 |
| CLUSTER | ||||||
| Nissen et al. | - | |||||
| Ozono et al. 2005 | - | |||||
| Kissane et al. 2003 | ||||||
| Schuler et al.2014 | Less involved | 5 | 1.91(0.29) | 2.07(1.11) | 3.03(0.82) | 7.01 |
| Kissane et al.1994 | Sullen | - | 1.97(0.19) | 1.74(1.13) | 1.34(1.26) | 6.37 |
| CLUSTER | ||||||
| Nissen et al. | - | |||||
| Ozono et al. 2005 | - | |||||
| Kissane et al. 2003 | - | |||||
| Schuler et al.2014 | Hostile | 5 | 1.30(0.29) | 1.18(0.97) | 1.18(0.75) | 3.66 |
| Kissane et al.1994 | Hostile | - | 0.74(0.56) | 0.61(0.61) | 2.65(1.57) | 2.7 |
Ozono et al. 2005, Kissane 2003, Kissane 1994 examines conflict instead of conflict resolution
Limitations
There are a few important limitations to this study that warrant discussion. A study of Australian caregivers found that family clusters differed between patients and caregivers (14). In the current study, the FRI was not administered to patients, in order to minimize patient burden. As a result, we are unable to compare family clusters across patients and caregivers. Another limitation of the study is that we did not have repeated assessments to determine the stability of family typologies over time or causality in their predictive validity.
Conclusion
The FRI is an instrument which may help to identify caregivers at risk for poor psychosocial outcomes. In this study, family caregivers were grouped into 3 clusters using a Gaussian Mixture method. Family types differed in caregiver quality of life, social support and perceived caregiver burden. Detached families were at greatest risk for poor outcomes. Goal-oriented interventions for caregivers of at-risk families that target building basic trust, caregiving skills, expressing emotions, and conflict management may improve caregiver psychosocial function.
Supplementary Material
Supplementary Table 1. Sample Description of the 3-Clusters, and Comparison Across Socio-demographic Variables for the 3-Cluster Model.
Supplementary Table 2. Caregiver Sociodemographic , and Patient Disease/Prognostic Variables at Assessment (N=598).
Acknowledgments
This research was supported in part by the following grants: MH63892 (HGP) from the National Institute of Mental Health and CA106370 (HGP) and CA197730 (HGP) from the National Cancer Institute, MD007652 (PKM, HGP) from the National Institute of Minority Heath and Health Disparities. Dr. Trevino's time was supported by a grant from the National Institute on Aging and the American Federation for Aging Research: K23 AG048632.
Contributor Information
Kathrine G. Nissen, Department of Psychology, Copenhagen University, DK
Kelly Trevino, Division of Geriatrics and Palliative Medicine, Department of Medicine, Weill Cornell Medicine, NYC, NY and the Center for Research on End of Life Care, Cornell University, Ithaca, NY, USA.
Theis Lange, Section of Biostatistics, University of Copenhagen, DK & Center for Statistical Science, Peking University, China.
Holly G. Prigerson, Division of Geriatrics and Palliative Medicine, Department of Medicine, Weill Cornell Medicine, NYC, NY and the Center for Research on End of Life Care, Cornell University, Ithaca, NY, USA
References
- 1.Braun M, Hales S, Gilad L, et al. Caregiving styles and attachment orientations in couples facing advanced cancer. Psycho-Oncology. 2012;21:935–43. doi: 10.1002/pon.1988. [DOI] [PubMed] [Google Scholar]
- 2.Turan B, Goldstein Mary KK, Garber Alan MM, Carstensen Laura LL. Health psychology : official journal of the Division of Health Psychology. Vol. 30. American Psychological Association; Knowing loved ones' end-of-life health care wishes: attachment security predicts caregivers' accuracy; pp. 814–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Stenberg U, Ruland CM, Miaskowski C. Review of the literature on the effects of caring for a patient with cancer. Psycho-Oncology. 2010;19:1013–1025. doi: 10.1002/pon.1670. [DOI] [PubMed] [Google Scholar]
- 4.Weitzner MA, McMillan SC, Jacobsen PB. Family caregiver quality of life: Differences between curative and palliative cancer treatment settings. Journal of Pain and Symptom Management. 1999;17:418–428. doi: 10.1016/s0885-3924(99)00014-7. [DOI] [PubMed] [Google Scholar]
- 5.Braun M, Mikulincer M, Rydall A, Walsh A, Rodin G. Hidden morbidity in cancer: spouse caregivers. J Clin Oncol. 2007;25:4829–34. doi: 10.1200/JCO.2006.10.0909. [DOI] [PubMed] [Google Scholar]
- 6.Tang ST, Chang WC, Chen JS, et al. Course and predictors of depressive symptoms among family caregivers of terminally ill cancer patients until their death. Psycho-Oncology. 2013;22:1312–1318. doi: 10.1002/pon.3141. [DOI] [PubMed] [Google Scholar]
- 7.Tang ST, Huang GH, Wei YC, et al. Trajectories of caregiver depressive symptoms while providing end-of-life care. Psycho-Oncology. 2013;22:2702–2710. doi: 10.1002/pon.3334. [DOI] [PubMed] [Google Scholar]
- 8.Brummett BH, Babyak MA, Siegler IC, et al. Associations among perceptions of social support, negative affect, and quality of sleep in caregivers and noncaregivers. Health Psychology. 2006;25:220–225. doi: 10.1037/0278-6133.25.2.220. [DOI] [PubMed] [Google Scholar]
- 9.Robinson-Whelen S, Tada Y, MacCallum RC, McGuire L, Kiecolt-Glaser JK. Long-term caregiving: What happens when it ends? Journal of Abnormal Psychology. 2001;110:573–584. doi: 10.1037//0021-843x.110.4.573. [DOI] [PubMed] [Google Scholar]
- 10.Kim Y, Schulz R, Carver CS. Benefit-finding in the cancer caregiving experience. Psychosom Med. 2007b;69:283–91. doi: 10.1097/PSY.0b013e3180417cf4. [DOI] [PubMed] [Google Scholar]
- 11.Edwards B, Clarke V. The validity of the family relationships index as a screening tool for psychological risk in families of cancer patients. Psycho-Oncology. 2005;14:546–554. doi: 10.1002/pon.876. [DOI] [PubMed] [Google Scholar]
- 12.Kissane DW, Bloch S, Dowe DL, et al. The Melbourne family grief study, I: Perceptions of family functioning in bereavement. American Journal of Psychiatry. 1996;153:650–658. doi: 10.1176/ajp.153.5.650. [DOI] [PubMed] [Google Scholar]
- 13.Siminoff LA, Wilson-Genderson M, Baker S., Jr Depressive symptoms in lung cancer patients and their family caregivers and the influence of family environment. Psycho-Oncology. 2010;19:1285–1293. doi: 10.1002/pon.1696. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Kissane DW, Bloch S, Burns WI, et al. Perceptions of family functioning and cancer. Psycho-Oncology. 1994;3:259–269. [Google Scholar]
- 15.Kissane DW, McKenzie M, McKenzie DP, et al. Psychosocial morbidity associated with patterns of family functioning in palliative care: Baseline data from the Family Focused Grief Therapy controlled trial. Palliative Medicine. 2003;17:527–537. doi: 10.1191/0269216303pm808oa. [DOI] [PubMed] [Google Scholar]
- 16.Ozono S, Saeki T, Inoue S, et al. Family functioning and psychological distress among Japanese breast cancer patients and families. Support Care Cancer. 2005;13:1044–50. doi: 10.1007/s00520-005-0816-5. [DOI] [PubMed] [Google Scholar]
- 17.Schuler TA, Zaider TI, Li Y, et al. Typology of perceived family functioning in an American sample of patients with advanced cancer. Journal of Pain and Symptom Management. 2014;48:281–288. doi: 10.1016/j.jpainsymman.2013.09.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Moos RH, Moos BS. Family environment scale : manual. Palo Alto, Calif: Consulting Psychologists Press; 1981. [Google Scholar]
- 19.Clatworthy J, Buick D, Hankins M, Weinman J, Horne R. The use and reporting of cluster analysis in health psychology: A review. British Journal of Health Psychology. 2005;10:329–358. doi: 10.1348/135910705X25697. [DOI] [PubMed] [Google Scholar]
- 20.Kim Y, Carver CS, Deci EL, Kasser T. Adult attachment and psychological well-being in cancer caregivers: the mediational role of spouses' motives for caregiving. Health Psychol. 2008;27:S144–54. doi: 10.1037/0278-6133.27.2(Suppl.).S144. [DOI] [PubMed] [Google Scholar]
- 21.Kuscu MK, Dural U, Onen P, et al. The association between individual attachment patterns, the perceived social support, and the psychological well-being of Turkish informal caregivers. Psycho- Oncology. 2009;18:927–35. doi: 10.1002/pon.1441. [DOI] [PubMed] [Google Scholar]
- 22.Lo C, Hales S, Braun M, et al. Couples facing advanced cancer: examination of an interdependent relational system. Psycho-Oncology. 2013;22:2283–2290. doi: 10.1002/pon.3289. [DOI] [PubMed] [Google Scholar]
- 23.Forbat L, McManus E, Haraldsdottir E. Clinical Implications for Supporting Caregivers at the End-of-Life: Findings and from a Qualitative Study. Contemporary Family Therapy. 2012;34:282–292. [Google Scholar]
- 24.Kim Y, Carver CS. Frequency and difficulty in caregiving among spouses of individuals with cancer: Effects of adult attachment and gender. Psycho-Oncology. 2007a;16:714–723. doi: 10.1002/pon.1110. [DOI] [PubMed] [Google Scholar]
- 25.Kadan-Lottick NS, Vanderwerker LC, Block SD, Zhang B, Prigerson HG. Psychiatric disorders and mental health service use in patients with advanced cancer: a report from the coping with cancer study. Cancer. 2005;104:2872–81. doi: 10.1002/cncr.21532. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Garrido MM, Prigerson HG. The end-of-life experience: modifiable predictors of caregivers' bereavement adjustment. Cancer. 2014;120:918–25. doi: 10.1002/cncr.28495. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Edwards B, Clarke V. The validity of the family relationships index as a screening tool for psychological risk in families of cancer patients. Psycho-Oncology. 2005;14:546–54. doi: 10.1002/pon.876. [DOI] [PubMed] [Google Scholar]
- 28.Mishra G, Schofield MJ. Norms for the physical and mental health component summary scores of the SF-36 for young, middle-aged and older Australian women. Quality of Life Research. 1998;7:215–220. doi: 10.1023/a:1024917510063. [DOI] [PubMed] [Google Scholar]
- 29.Boele FW, Douw L, Reijneveld JC, et al. Health-related quality of life in stable, long-term survivors of low-grade glioma. Journal of Clinical Oncology. 2015;33:1023–1029. doi: 10.1200/JCO.2014.56.9079. [DOI] [PubMed] [Google Scholar]
- 30.Cohen S, Mermelstein R, Kamarck T, Hoberman H. Measuring the functional components of social support. The Hague: Martinus Nijhoff 1985; 1985. [Google Scholar]
- 31.Tomarken A, Holland J, Schachter S, et al. Factors of complicated grief pre-death in caregivers of cancer patients. Psycho-Oncology. 2008;17:105–111. doi: 10.1002/pon.1188. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Trevino KM, Fasciano K, Prigerson HG. Patient-oncologist alliance, psychosocial well-being, and treatment adherence among young adults with advanced cancer. J Clin Oncol. 2013;31:1683–1689. doi: 10.1200/JCO.2012.46.7993. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Trevino KM, Fasciano K, Block S, Prigerson HG. Correlates of social support in young adults with advanced cancer. Support Care Cancer Supportive Care in Cancer. 2013;21:421–429. doi: 10.1007/s00520-012-1536-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Rokach L, Maimon O. Clustering Methods. In: Maimon O, Rokach L, editors. Data Mining and Knowledge Discovery Handbook. Springer; US: 2005. pp. 321–352. [Google Scholar]
- 35.Team RC. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2014. [Google Scholar]
- 36.Fraley C, Raftery A. Model-based Methods of Classification: Using the mclust Software in Chemometrics. Journal of Statistical Software. 2007;18 [Google Scholar]
- 37.Fraley C, Raftery A, Murphy T, Scrucca L. mclust Version 4 for R: Normal Mixture Modeling for Model-Based Clustering, Classification, and Density Estimation. Seattle, USA: University of Washington; 2012. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Table 1. Sample Description of the 3-Clusters, and Comparison Across Socio-demographic Variables for the 3-Cluster Model.
Supplementary Table 2. Caregiver Sociodemographic , and Patient Disease/Prognostic Variables at Assessment (N=598).