Random subsamples from the Centers for Medicare & Medicaid Services Medicare enrollment database provided the sampling frame for the NHATS (Montaquila, Freedman, Edwards, & Kasper, 2012). To ensure a nationally representative sample of Medicare beneficiaries age 65 and older, the NHATS study design included baseline stratification of 5-year age groups (i.e., 65 to 89, ≥90). Two subgroups often underrepresented in research were oversampled—individuals who were Black and individuals age 90 and older. Selection into the sample was designed to generate equal probability samples, including targeted sample sizes by age group and race/ethnicity. NHATS protocols were approved by the Johns Hopkins University review board, and all study participants provided written informed consent. Face-to-face annual interviews were conducted by trained personnel with the same participants each year and provided the main source of data for the NHATS. Participants who resided in the community and completed the first (N = 7,609) and second (N = 6,056) year participant interviews were included in the current study. No exclusion criteria were used.
Starting in Year 2, some of the first-year community-dwelling participants transitioned into nursing homes and completed the participant interview in their new setting. Others did not complete their Year 2 interview, were lost to follow-up, or had died, such that Year 2 N = 6,056 (including 64 [1.06%] nursing home residents). Falls are a risk factor for nursing home placement (Lach, 2010), so participants who were residing in a nursing home in Year 2 and had participated in the interview in Year 1 were not removed from the analytic sample for the current analysis.
The NHATS is a publicly available data set accessed by registering online (access http://www.nhats.org) and downloading the data files. Additional demographic data that were viewed as sensitive, such as age, were available through a simple application process. The University of Washington Human Subjects Division established that the current study did not meet the definition of research concerning human participants because the data were de-identified.
A data subset was created for the current study with NHATS variables that are known to be associated with or predictive of falls (Berry, Lee, Cai, & Dore, 2013; Stevens & Phelan, 2013; Tinetti, Gordon, Sogolow, Lapin, & Bradley, 2006) and social isolation (Coyle & Dugan, 2012; DiNapoli, Wu, & Scogin, 2014; Iliffe et al., 2007; McCrae et al., 2005; Pantell et al., 2013; Shankar et al., 2011; Shub et al., 2011). The variables in the data subset included living-alone status, sleep medicine used, current smoker or non-smoker, hearing device used, visual correction used, ADL, instrumental activities of daily living (IADL), worry about falling, assistive mobility device, taking walks, vigorous activity, Short Physical Performance Battery (SPPB), overall health, heart disease, arthritis, osteoporosis, diabetes, stroke, Alzheimer's disease or other dementia, and depression risk.
Outcome Variable. Several questions regarding falling were asked, prefaced by the definition of falling as “any fall, slip, or trip in which you lose your balance and land on the floor or ground or at a lower level” (Gadkaree, Sun, Huang, Varadhan, & Agrawal, 2015, p. 3). The outcome variable for the current study was assessed by asking participants if they had experienced a fall in the past 12 months. Participants were also asked if they worried about falling in the past 1 month, with a possible response of yes or no (Gell, Wallace, LaCroix, Mroz, & Patel, 2015).
Independent Variable. For the current study, social isolation was defined as a social circumstance in which an older adult had a deficiency of network contacts and integrating relationships with contacts (Nicholson, 2010; Pohl, Cochrane, Schepp, & Woods, 2017). Social isolation was operationalized as a comprehensive and domain-inclusive measure based on Berkman and Syme's (1979) Social Network Index (SNI) using comparable items (i.e., indicators) available in the NHATS data set. Berkman and Syme (1979) suggested the SNI is a strong predictor of health and mortality outcomes. The four SNI conceptual domains include marriage, family and friend contact, church, and club participation, and have been operationalized within many research studies to create a constructed measure of social network and integration using data items (indicators) that varied from study to study. The continued use of the SNI domains over time support its validity for identifying individuals who are socially isolated (Nicholson, 2010; Pantell et al., 2013; Shankar et al., 2011).
Indicators of the SNI domains of social isolation in the current study used NHATS data items on marriage/partner status; talked with family; talked with friends; visited in family's, friends', or own homes; attended religious services; and participated in club/community activities. In the current study, the domain items were recoded to ensure a higher score indicated greater isolation. One point was recorded for each negative (isolated) response to an item, and the total score (range = 0 to 6) was based on the sum of the negative responses. Adapted from Nicholson (2010), the Figure characterizes the relationships among social isolation, SNI network contacts and integrating relationship domains, and the indicators (items) in the NHATS interview. A complete description of this measure development was published previously (Pohl et al., 2017).
Conceptualization of the social isolation measure, the Social Network Index network contact, and integrating relationship domains, and the indicators in the National Health and Aging Trends Study interview. Adapted from Nicholson (2010). Reprinted with permission from Pohl, J.S., Cochrane, B.B., Schepp, K.G., & Woods, N.F. (2017). Measuring social isolation in the National Health and Aging Trends Study. Research in Gerontological Nursing, 10, 277–287. Copyright © SLACK Incorporated
Demographic Variables. Age, sex, self-identified race/ethnicity, and education, as reported in the baseline interview, were included. Education was included as an indicator of socioeconomic status (Winkleby, Jatulis, Frank, & Fortmann, 1992). Race/ethnicity was categorized into White non-Hispanic, Black non-Hispanic, or Hispanic. Additional non-Hispanic groups including American Indian, Asian, and Native Hawaiian/Pacific Islander were reported as other. Nine education levels were collapsed into three: no schooling to 12th grade schooling but no diploma, high school graduate to high school graduate with some college, and associate to graduate degree.
Behavioral and Sensory Variables. Living-alone status was measured as a dichotomous variable. Living alone is the most frequently reported measure of social isolation related to falls (Elliott et al., 2009; Flabeau et al., 2013; Kharicha et al., 2007). The use of sleep medicines, current cigarette smoking, and sensory impairment (i.e., hearing device used in the past 1 month, glasses or contacts needed to see things at a distance) were also measured as dichotomous variables.
Health Variables. Participants rated their general health as excellent, very good, good, fair, or poor. Participants were asked if they had been told by their health care professional that they had a broken bone, broken hip, heart attack, heart disease, arthritis, osteoporosis, diabetes, stroke, cancer, or dementia, with possible answers of yes or no. Depression risk was identified with the Patient Health Questionnaire-2 (PHQ-2), a brief screening instrument (Löwe et al., 2010). The items began with the question, “Over the last month, how often have you…” and included: (a) had little interest or pleasure in doing things, and (b) felt down, depressed, or hopeless. The ordinal response options were rated from not at all to nearly every day, and the score was the sum of both items, with a higher score indicating the need for further depression assessment (Löwe et al., 2010). The reliability and validity of this measure has been established in the general population (Kroenke, Spitzer, & Williams, 2003).
Physical Function Variables. The SPPB was used to measure physical performance (Kasper, Freedman, & Niefeld, 2012; Sun, Huang, Varadhan, & Agrawal, 2016). This assessment consists of balance stands, gait speed, and chair stands. Scores for each measure ranged from 0 to 4, with 0 indicating an inability to complete the test and 4 indicating highest performance. The total SPPB score was the sum of the three scores. The internal consistency reliability of the SPPB has been previously reported with a 0.76 Cronbach's alpha in older adults (Guralnik et al., 1994). The use of an assistive mobility device was assessed by asking if in the past 1 month, a device (e.g., cane, scooter, walker, wheel-chair) was used. Physical activity was assessed with two questions that asked participants if they ever go for walks or spend time doing a vigorous activity that increases their breathing and heart rate. Response options for these measures were yes/no. ADL scores were assessed with questions regarding help needed for dressing, eating, bathing, and toileting. IADL scores were assessed with questions regarding help needed to do laundry, shop, cook hot meals, handle bills and banking, keep track of medications, and drive. Higher scores indicated more help was needed. The validity of the ADL and IADL measures and a high degree of internal consistency reliability have been documented in previous research with older adults (Katz, 1983).