Study design, setting, and sample
We used data from the National Health and Aging Trends Study (NHATS) wave 5 (2015) and wave 6 (2016). Since 2011, the NHATS has collected health information from a nationally representative sample of Medicare beneficiaries aged 65 and older in the U.S. [31], and the sample was replenished in 2015. Interviews were conducted annually to document the changes over time. In our study, a total of 6397 participants completed the interviews in both waves (time 1 [T1] and time 2 [T2]). We excluded 21 beneficiaries who had incomplete data on fear of falling or activity restrictions. Thus, a total of 6376 beneficiaries formed the final sample.
Measurements
Independent variables
Anxiety symptoms were measured by the 2-item Generalized Anxiety Disorder Scale (GAD-2) [31]. It screens anxiety symptoms by asking, “Over the last month, how often have you (a) felt nervous, anxious, or on edge, and (b) been unable to stop or control worrying?” Response options were “1=not at all”, “2 = several days”, “3=more than half the days”, and “4=nearly every day.” The total score ranges from two to eight, with the cut-off point of five or higher indicating the presence of anxiety symptoms [32].
Depressive symptoms were measured using the 2-item Patient Health Questionnaire (PHQ-2) [31]. PHQ-2 measures depressive symptoms by asking, “Over the last month, how often have you (a) had little interest or pleasure in doing things, and (b) felt down, depressed, or hopeless?” Responses were based on a four-point scale (1 = not at all, 2 = several days, 3 = more than half the days, 4 = nearly every day). The total score ranges from two to eight, and a total score of five or higher was used to indicate the presence of depressive symptoms [33].
Dependent variables
Fear of falling and activity restrictions were assessed by the two questions, “In the last month, did you worry about falling down?” If the answer was yes, then participants were asked, “In the last month, did this worry ever limit your activities?” Based on the participants’ responses, a three-category variable was created to indicate the fall worry levels: No fear of falling (coded 0), had fear of falling but not activity restrictions (coded 1) and had fear of falling-related activity restrictions (coded 2).
Covariates
We included covariates hypothesized to be associated with our outcomes of interest. These included demographic variables such as age, gender (female vs. male); race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, all other); education level (less than high school, high school graduates, some college or vocational school, bachelor’s degree or higher); and living arrangement (alone, with spouse/partner only, with others only, with spouse/partner and others).
Health related covariables included (a) the number of chronic illnesses (heart attack/heart disease, high blood pressure, arthritis, osteoporosis, diabetes, lung disease, stroke, and cancer); (b) whether the participants had dementia (yes/no); (c) the number of activities of daily living (ADL) impairment, ranging from 0 to 4 (feeding, bathing, toileting, dressing); (d) the number of instrumental activities of daily living (IADL) impairment, ranging from 0 to 7 (bed transfer, moving inside the house, doing laundry, shopping, preparing meals, taking medication, and managing money); (e) whether the participants were bothered by pain in the last month (yes/no); (f) body mass index (BMI) of the participants (normal/obesity [≥30 kg/m2]); (g) whether the participants were hospitalized over the past year (yes/no); and (h) whether the participants had problems with balance or coordination in the last month (yes/no).
Statistical analyses
Continuous variables were presented as mean ± standard deviations and categorical variables as frequencies and percentages. We used two-sample t-tests to estimate the distribution of age, ADLs, and IADLs across the fall worry levels. Chi-square tests were used to test the differences among groups of fall worry levels for categorical variables.
To determine whether anxiety and/or depressive symptoms at T1 could predict the fall worry levels at T2 independently, we performed three sets of multinominal logistic regression models in three steps. First, we modeled the effects of anxiety symptoms at T1 on fall worry levels at T2, in which fall worry levels at wave 6 were treated as the outcome while fall worry levels at T1 were controlled. Next, we conducted similar models replacing anxiety symptoms with depressive symptoms as the main predictor to examine the effects of depressive symptoms on fall worry levels. Finally, we examined the independent effects of anxiety and depressive symptoms on fall worry levels by including them as main predictors simultaneously in the models. In each model set, we first estimated the crude effects (Model 1), followed by the effects adjusted for demographic variables (Model 2), and finally the effects adjusted for demographic and health-related covariates (Model 3).
To improve the robustness of the results, we performed a sensitivity analysis excluding samples who were interviewed by proxy (n = 439). Due to the small proportion of missing data and the large sample size, we did not use any techniques to handle the missing data. For all models, relative risk ratios (RRR) and 95% confidence intervals were reported. P values less than 0.05 indicated statistical significance. All analyses were conducted using Stata/SE 15.0 (Stata Corp., College Station, TX).
