Hispanic/Latinos and non-Hispanic whites’ childhood cancer survivors and parents: a dyadic analysis of coping resources and mental health

Participant

Participants were from the Project Forward pilot study, a cross-sectional study that used population-based survey methods to recruit 160 parent–child dyad respondents [30]. CCS diagnosed at age < 18 years and were treated at either of two children hospitals: Children’s Hospital Los Angeles or Miller Children’s Hospital in Long Beach, were selected from the SEER cancer registry for Los Angeles County. Survivors met the inclusion criteria if they had (1) a diagnosis of any cancer except for Hodgkin’s lymphoma (these survivors were participating in a different study); (2) were diagnosed between 2000 and 2007; and (3) were able to read and write in English and/or Spanish to complete the survey.

Procedures

The standard recruitment procedure for CCS and parents included two methods (detailed previously) [31]. If CCS were < 18 years at the time of the survey, both the parent and survivor were invited to participate; otherwise, the CCS was directly invited and subsequently asked for parental contact information and permission. Informed consent was obtained from both CCS and parents. Upon completing the survey, both received a $20 gift card for participating. All study procedures were approved by the California Committee for the Protection of Human Subjects, California Cancer Registry, and by human subject research committees at the University of Southern California, Children’s Hospital of Los Angeles, and Miller Children’s Hospital.

Measures

A standard dyadic design was used for this study at the time of data collection in 2009. Only one parent and CCS participated in the study, and both members were measured on the same independent and dependent variables.

Independent variables (coping resources selected based on theoretical and existing literature)

Social support — single-item measure of social support (SIMSS), a validated predictor of morbidity strongly associated with the composite social support index, was used [32]. It asked participants, “how many people do you have that you can count on for help when you need them, such as, to give rides to the hospital or store, or to help if you are sick?” Response options range from 0 = none, 1 = 1–2, to 2 = 3 or more, with higher scores indicating greater social support.

Religiosity/spirituality — two questions were asked about aspects of religiosity and spirituality. The first question assessed religiosity by asking the “frequency of religious service attendance,” and response options range from 0 = never, 1 = every few years, 2 = several times a year, 3 = 2–3 times per month, to 4 = at least once per week, with higher scores indicating a greater frequency of religious service attendance. The second question assessed spirituality by asking participants about the “importance of religion or spirituality,” and response options ranged from 0 = not important, 1 = somewhat important, to 2 = very important, with higher scores indicating greater importance of religion or spirituality.

Dependent variables

Depressive symptoms — with the validated 20-item Center for Epidemiological Studies Depression Scale (CES-D) [33], participants are asked to report the frequency of depressive symptoms that occurred during the past week using a 4-point Likert scale ranging from (0) none of the time to (3) all of the time. The present analysis used the total sum score, with higher scores indicating greater depressive symptoms. Cronbach’s alpha in this dyad sample was 0.84 for parents and 0.92 for CCS.

Perceived Stress Scale — stress was assessed using the validated 4-item version of the Perceived Stress Scale (PSS) [34]. Participants were asked the extent to which they felt or thought about each statement in the past month, with responses ranging from 0 = never to 4 = very often. Items were summed to create a total perceived stress score, with higher scores indicating more stress. Cronbach’s alpha in this dyad sample was 0.71 for parents and 0.54 for CCS.

Covariates

Demographic and clinical information was obtained from self-report and cancer registry data. CCS’s current age at the survey was used as a continuous variable. Treatment intensity was calculated using the Intensity of Treatment Rating Scale 2.0 (ITR-2) [35], a 4-level validated scale ranging from 1 = least intensive treatment (surgery only) to 4 = most intensive, from cancer registry data and medical chart review. Parent Hispanic ethnicity was self-reported in the survey. Additional potential covariates were also assessed, including parents’ primary language, parental education, socioeconomic status, CCS time since diagnosis, and whether CCS lived with parents. Based on the model fit indices and the literature that previously found an association between these covariates and quality of life outcomes [11, 25], the primary analysis models included three selected covariates: age at survey, treatment intensity, and parent Hispanic ethnicity.

Statistical analysis

SAS version 9.4 was used for data cleaning and to conduct descriptive analysis [36]. Descriptive statistics were conducted to compare demographic characteristics between CCS and parents using t-tests for continuous variables and chi-square for categorical variables. Pearson’s correlation coefficient was used to evaluate non-independence between the dyadic member’s scores on the coping resources predictor (social support, religiosity, and spirituality), covariates (age of CCS, treatment intensity, and parent Hispanic ethnicity), and dependent outcomes (depressive symptoms and perceived stress). We used Cohen’s criterion, where 0.5 is a large correlation, 0.3 is a medium correlation, and 0.1 is a small correlation [37].

Path analyses of APIM in Mplus version 8 were used to model the reciprocity of parent-CCS dyadic relationships and examine our hypotheses [38]. The full information maximum likelihood estimation method was used, which includes all dyad data in the analyses as long as one individual responds to the respective outcome. Due to theoretical consideration, APIM for distinguishable dyads was used to implement all models, which means path analysis results would give us two equations—one for each person within the dyads (e.g., parents and CCS) [39]. Two separate actor-partner models were conducted based on our two dependent variables (depressive symptoms and perceived stress), and all three independent variables (social support, religiosity, and spirituality) were entered into each model. Supplemental Figs. 1 and 2 illustrate the specification of our two models.

Several models were estimated, and model fit was statistically evaluated using the criterion described below. First, we examined our two models without controlling for covariates and found that our model fit was poor. In these APIM models, both actor and partner effects are examined simultaneously while controlling for variance explained by the partner. Second, we tested whether adding all potential covariates (see “Measures”) to these two models demonstrated a better model fit. Next, we used the model modification indices results from these models to determine which additional covariate pathways were significantly associated with the key study variables and improved our model fit. Based on the procedure, covariates in the final model were selected.

In our final models, both actor and partner effects were examined simultaneously while controlling for variance explained by the partner. Additionally, we included our covariates and four correlational paths across covariates; these were added to improve the model fit indices. Finally, the moderating effects of Hispanic ethnicity were tested using the multigroup approach, allowing parameter estimates to vary among Hispanics or non-Hispanic dyads. We calculated standardized coefficients (β) and used a p-value < 0.05 (two-tailed) to determine statistical significance.

The overall model fit was assessed using chi‐square statistic (χ2), degree of freedom (df), a root mean square error of approximation (RMSEA), the standardized root mean residual (SRMR), Tucker‐Lewis index (TLI), and comparative fit index (CFI) [40]. Adequate fit for a specified model to the data requires a non-significant chi‐square statistic, RMSEA of less than or equal to 0.06, SRMR value of less than 0.08, TLI greater than 0.90, and CFI greater than 0.95.

留言 (0)

沒有登入
gif