Data and participants
The National Health and Aging Trends Study (NHATS) is an ongoing, longitudinal study that surveys a nationally representative sample of Medicare beneficiaries ages 65 years and older in the United States [18]. All participants gave informed consent and ethical approval was obtained. The National Study of Caregiving (NSOC) studies family and other unpaid caregivers to older persons living with limitations in daily activities, which had been conducted three times in conjunction with the NHATS [19].
We used data from the 2011, 2015, and 2017 waves of the NHATS and NSOC. We did not include data from other waves because the NSOC surveys were only distributed to caregivers in 2011, 2015, and 2017. A total of 12,427 care recipients participated in the 2011, 2015, 2017 waves of the NHATS and 3,778 caregivers were surveyed in NSOC. Our analytic sample comprised of 1,067 community-dwelling older adults with probable dementia and their caregivers. A care recipient is counted as probable dementia if they 1) had self- or proxy-reported diagnosis; 2) met the AD8 diagnosis criteria for dementia based on their orientation, executive function, and memory assessments; or 3) scored more than 1.5 standard deviations below the mean in two or more cognitive testing domains. For care recipients with multiple caregivers participating in the NSOC, we selected their primary caregiver who had the most caregiving hours in the last month prior to their interview [20]. For caregivers who were surveyed in multiple waves of NSOC, we used data from the first wave.
Comorbid chronic disease burden
Nine comorbid chronic conditions were considered: heart disease (heart attacks, myocardial infarction, angina or congestive heart failure, and other heart diseases), hypertension, arthritis, osteoporosis, diabetes, lung disease, stroke, cancer, and depressive symptoms. Having depressive symptoms was determined based on the Patient Health Questionnaire-2 (PHQ-2) screening test, which inquires about the frequency of depressed mood over the past two weeks. Each participant was asked, “Over the last month, how often have you a) had little interest or pleasure in doing things; b) felt down, depressed, or hopeless?” A response of “Not at all”, “Several days”, “More than half the days”, and “Nearly every day” was coded 0, 1, 2, and 3, respectively. The PHQ-2 score ranges from 0–6. A person with a score of 3 or greater was considered having major depressive disorder. All other chronic conditions were self- or proxy-reported.
Caregiving burden and gain
We considered four aspects of caregiving burden: physical, psychological, social, and financial burden. Each aspect was assessed by a composite score calculated based on multiple self-reported items capturing caregiving experience. Physical burden includes five items: whether the caregiver is exhausted when go to bed at night and whether their activities are limited by pain, breathing problems, low strength, or low energy (1 = very much, somewhat). Psychological burden includes eighteen items: whether the caregiver has anxiety or depression, which are determined based on Generalized Anxiety Disorder 2-item (GAD-2) and PHQ-2 criteria (1 = Yes); has more things to do than they can handle, doesn’t have time for themselves, or needs change as soon as they get a routine going (1 = very much, somewhat); has a life with meaning and purpose, feels confident and good about themselves, likes their living situation very much, has an easy time adjusting to changes, or gets over illness and hardship quickly (1 = disagree somewhat, disagree strongly); gives up trying to improve their life a long time ago, or feels lonely because they have few close friends (1 = agree somewhat, agree strongly); feels bored, lonely, upset (1 = some days, most days, everyday); or feels cheerful, calm and peaceful, or full of life (1 = rarely, never). Social burden includes five items: whether caregiving keeps them from visiting families or friends, attending religious services, going out for enjoyment, volunteering, or caring for someone else (1 = Yes). Financial burden includes two items: whether caregiving keeps them from working for pay or made it harder for them to get work done in the past month (1 = Yes). Composite scores were calculated for each of the four domains, resulting in a physical burden composite score ranges from 0 to 5, psychological burden composite score ranges from 0 to 18, social burden composite score ranges from 0 to 5, and financial burden composite score ranges from 0 to 2.
We also considered caregiving gain, which includes four items: whether caregiving makes them more confident in abilities, teaches them to deal with difficulties, brings them closer to the care recipient, or gives them the satisfaction that the care recipient is well cared for (1 = very much, somewhat). A composite score for caregiving gain was calculated and ranges from 0 to 4.
Covariates
Care recipients’ background characteristics include their gender (1 = female), age (65–74 years old, 75–84 years old, 85 + years old), and race (white non-Hispanic, Black non-Hispanic, Other non-Hispanic, and Hispanic). Caregivers’ background characteristics include gender (1 = female), age (1 = over 65 years), race (white non-Hispanic, Black non-Hispanic, Other non-Hispanic, and Hispanic), relationship to care recipient (1 = spouse, 0 = non-spouse), education level (1 = college or above), total caregiving hours in the past month, and whether the caregiver is involved in long-term caregiving for over five years (1 = Yes). We also adjusted for the type of caregiving assistance that the caregiver provides for their care recipients, including assistance with activities of daily living and instrumental activities of daily living (ADL-IADL), health managements, and medical tasks. The formal and informal caregiving support that caregiver received are also included as covariates.
Statistical analysis
We described the sociodemographic characteristics of care recipients, caregivers, and caregiving burden by the year the care recipient was surveyed (2011, 2015, 2017) and the number of comorbid chronic conditions the care recipients have (0–1, 2, 3, 4, 5 or more). We used means and standard deviations for continuous variables and counts and percentages for categorical variables.
Care recipients’ comorbidity burden was measured in two ways: (1) count of chronic conditions and (2) pattern of comorbidity identified by the latent class analysis (LCA). For count of chronic conditions, we calculated the total number of comorbid chronic conditions the care recipient has in addition to dementia out of the nine conditions we considered. We modeled the count of comorbidities both continuously and in categories (0–1, 2, 3, 4, and 5 or more). For the LCA analysis, we identified patterns from the nine comorbid chronic conditions by examining two to five comorbidity burden classes and selected the optimal number of latent classes based on statistical criteria including: the Bayesian Information Criterion (BIC), adjusted BIC (aBIC), Akaike Information Criteria (AIC), Lo-Mendell-Rubin adjusted likelihood ratio test (LMR), and entropy. Models with smaller BIC, aBIC, and AIC values, LMR adjusted test p-value greater than 0.05, and greater entropy value indicate better model fit. After we determined the final LCA model, we assigned each participant to a comorbidity burden class based on the highest estimated probability of class membership.
We used a series of linear regressions to identify the unadjusted and adjusted association between the count of comorbidities among PWD and their caregivers’ physical, psychological, social, and financial burden separately, as well as caregiving gain. We repeated the linear regression models using the latent class membership of each care recipient. Care recipient’s gender, age, and race, as well as caregiver’s gender, age, race, relationship to care recipient, education level, caregiving hours, long-term caregiving status, caregiving activities (helping care recipient with ADL-IADL, health management, medical tasks), and formal and informal caregiving support were included as covariates in the multivariable adjusted models.
Descriptive analysis and linear regression models were carried out using STATA/SE, version 16.1 [21], and LCA was carried out using Mplus, version 8.7 [22].