Predictors and covariates

The primary predictor was self-reported race/ethnicity, and was categorized as non-Hispanic White, non-Hispanic Black, Hispanic, and Other minority. This variable was created from two original questions, one asking the race of the individual with options of ‘White/Caucasian’, ‘Black/African American’, and ‘Other’; and a second question asking ethnicity with options of ‘Not Hispanic’ and ‘Hispanic’. Individuals responding Hispanic to the ethnicity question were categorized as Hispanic regardless of race selection. Individuals selecting White and Not Hispanic were categorized as non-Hispanic White, those selecting Black and Not Hispanic were categorized as non-Hispanic Black, and those selecting Other and Not Hispanic were categorized as Other minority.

Demographic and socioeconomic factors included gender (male/female), age (in years), education (less than high school, GED/high school graduate, some college, college and above), household income and assets which included shares of stock, jewelry, and real estate (grouped into quartiles). Sociodemographic factors were pulled from the RAND HRS data [28].

Behavioral factors included smoking status (grouped as non-smoker, current smoker, and former smoker). Comorbidities included a count of self-reported conditions: high blood pressure, cancer, lung disease, heart condition, stroke, emotional/psychiatric problems, and arthritis. Comorbidities were then grouped as low comorbidity (0–1), moderate comorbidity (2–3), and high comorbidity (4+). BMI was categorized as underweight, normal weight, overweight, and obese. Behavioral factors and comorbidities were pulled from the RAND HRS data [28].

Social factors included social support (positive social support score and negative social support score), loneliness (a score of the original 3-item loneliness index); social participation (a score of 18 items), personal sense of control (separated into perceived constraints score and perceived mastery score) [29]. Social support was based on 1–4 scale from four sets of 7 items, which examined the perceived support that respondents received from their spouse/partner, children, family, and friends. For each relationship category there are 3 positively worded items and 4 negatively worded items. Positive and negative social support scores are determined by averaging the reverse-coded scores within each dimension with higher scores for positive social support indicating higher levels of positive social support, and higher scores for negative social support indicating higher levels of negative social support. Loneliness was based on 1–3 scale, which asks questions such as “do you lack companionship?”, “feel left out?”, “feel isolated from others?”. Scores are determined by averaging the scores across all 3 reverse- coding items, with higher scores indicating higher levels of loneliness. Social participation was based on 0–18 scale from 18 items. For each item, an indicator was created, with 1 indicating yes for answers with daily, several times a week, once a week, several times a month, at least once a month; 0 for answers with not in the last month, never/not relevant. Scores were determined by summing the indicator of each item, with higher scores indicating higher levels of social participation. Personal sense of control was based on 1–6 scale from two sets of 5 items, 5 items for constraints and 5 items for mastery. The final scores were constructed by averaging the scores across 5 items, with higher scores indicating higher levels of personal sense of control. Social factors were pulled from the biennial core interview data and scoring was based on HRS documentation [29].