Data Source
Data for this study came from the Behavioral Risk Factor Surveillance System (BRFSS) conducted by the Centers for Disease Control and Prevention (CDC), which is a national system of telephone surveys that collect data from U.S. residents regarding their chronic health conditions, health-related behaviors, and use of healthcare services [17]. Since 2009, the CDC has included an optional Caregiving Module as part of this survey, screening whether a respondent has served as a caregiver in the past 30 days or expects to in the next 2 years. From respondents who respond in the affirmative, the module assesses the care recipient’s major health problem, the care recipient’s relationship to the caregiver, duration and hours per week spent on caregiving, caregiving tasks, any unmet support service needs. The de-identified BRFSS data is publicly available. Therefore, this study was deemed exempted from institutional review board approval.
Study population
The sample included respondents who answered "Yes" to the question “During the past 30 days, did you provide regular care or assistance to a friend or family member who has a health problem or disability?” and selected cancer as the major health problem of the care recipient. To increase the statistical power for this study, we merged the most recent BRFSS data from four years of annual survey waves, 2015 to 2018. The study sample consists of 5,814 self-identified cancer caregivers across 45 states, which represents an estimated 4,792,344 individuals in these states. Detailed information of participating states is listed in the Appendix.
Measures
Mentally Unhealthy Days and Frequent Mental Distress
The dependent variable, cancer caregivers' mental health, was measured using a question asking about mentally unhealthy days (MUDs): "Thinking about your mental health, which includes stress, depression, and problems with emotions, for how many days during the past 30 days was your mental health not good?" The responses range from 0 to 30 days. In addition, MUDs were dichotomized and labeled in accordance with other studies as Frequent Mental Distress [18], such that a response of 14 or more mentally unhealthy days was coded as 1 ("distressed") and below 14 was coded as 0 ("not distressed"). MUD was developed by CDC in the 1990s and has been used in national surveys.20 Both MUD and FMD have been used in other empirical studies [19].
Unmet needs for support services
Unmet needs for support services was accessed by asking the question “Of the following support services, which one do you most need, that you are not currently getting?” Responses included: “Classes about giving care, such as giving medications,” “Help in getting access to services,” “Support groups,” “Individual counseling to help cope with giving care,” “Respite care,” and “You don’t need any of these support services.” A binary variable was created for reporting any unmet needs or no unmet needs.
Caregiving characteristics
Caregiving characteristics in this study included care recipient-caregiver relationship, caregiving intensity, and caregiving tasks that a caregiver performed. Relationship categories included spouse or partner, parents or parents-in-law, other relatives (child, siblings, grandparents, grandchild, and others), and family friends. Caregiving intensity was constructed from two dimensions including hours per week caregiving and duration in months as operationalized in literature6. Specifically, average weekly time caregiving was dichotomized as more or less than 20 hours, and caregiving duration was dichotomized at more or less than 2 years. The caregiving intensity variable had four categories, including high hours/long duration, high hours/short duration, low hours/long duration, and low hours/short duration. Caregiving tasks categories include no personal care or household tasks, personal care only, household task only, and personal care & household tasks.
Socio-demographic variables
Caregivers’ sociodemographic information included their age, gender, race/ethnicity, education, marital status, household income, employment, and region of the country. Age at the survey included three categories: 18 to 34, 35 to 54, and above 55 years old. Gender was a binary variable with male as 1 and female as 0. Race included three categories: non-Hispanic White, non-Hispanic Black, and other races, including Hispanics, Asian, American Indian or Native American, Native Hawaiian or other Pacific Islander, and multiracial. Education included three categories: graduated from high school or less, attended some college or technical school, and graduated from college or technical school. Marital status was a binary variable as 1 denotes “currently married or partnered” and 0 denotes “unmarried.” Household income categories included less than $25K, $25K-50K, and above $50K. Employment status is a binary variable with employed as 1 and unemployed as 0. Region include Northeast (Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, Vermont, New Jersey, New York, Pennsylvania), Midwest (Illinois, Indiana, Michigan, Ohio, Wisconsin, Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, South Dakota), South (Delaware, Florida, Georgia, Maryland, North Carolina, South Carolina, Virginia, District of Columbia, West Virginia, Alabama, Kentucky, Mississippi, Tennessee, Arkansas, Louisiana, Oklahoma, Texas), and West (Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, Wyoming, Alaska, California, Hawaii, Oregon, Washington).
Statistical analyses
Descriptive statistics included unweighted and weighted means and standard deviations for continuous variables and frequencies and percentages for categorical variables. Chi-square tests and two-sample t-tests were conducted to compare the differences between women caregivers and men caregivers. To examine the moderation hypothesis of gender on the associations between unmet needs and mental health outcomes, we conducted (1) logistic regression models to examine the association between unmet needs and FMD (the main effect model) and the moderation effect of gender on this association (model including an interaction term between gender and unmet need); and (2) zero-inflated negative binomial regression models (ZINB) as suggested by Zhou et al. [20] and the distribution of MUDs in our sample (Figure 1) to examine the association between unmet needs and MUDs (the main effect model) and the moderation effect of gender (model including an interaction term between gender and unmet need). In all models, we adjusted for all socio-demographics and caregiving characteristics listed above. All statistical analyses were conducted addressing the complex survey design of BRFSS by incorporating sampling weights in R Studio 1.1.383 [21] and Stata 16 [22]. Results were considered as significant for a two-sided p-value<0.05.