The memory performance task was included as part of a larger battery of cognitive tasks that was administered by computer-assisted telephone interview. Participants were presented with a word list (15 words, e.g., “drum”, “farmer”, “moon”), with a one-second interval between each word. Both immediate and delayed recall were measured by asking participants to repeat back as many words as they could remember. Immediate and delayed recall were combined into a single composite of memory performance and are presented in standard deviation (z-score) units. More details regarding this task can be found on the MIDUS Cognitive Project webpage (Ryff & Lachman, 2017) and in other publications (e.g., Stephan et al., 2014).
Resting Heart Rate Variability (HRV).
To assess cardiovascular health, we examined resting high-frequency heart rate variability (HRV), which was measured during the MIDUS Biomarker Project. HRV was measured in two 300-sec epochs. Due to the high correlation between HRV measurements https://brightwomen.net/blog/en-kort-historik-av-postordrebrud/ (r = 0.93), the two were averaged, and log transformed values are presented in this paper and used in the analyses. For details regarding HRV measurements and data collection, see details on the MIDUS Bio) and work by Dienberg Love et al. (2010). Respiration was also measured; however, results for HRV did not change appreciably when controlling for respiration.
Covariates.
For the purposes of this paper, selected covariates were those related to key demographic dimensions and those that were plausibly related to our outcomes. The majority of the covariates selected have been used in prior publications analyzing data from the Cognitive and Bio; D. Weiss & Weiss, 2016). Demographic variables were gender (?.5 = female, .5 = male) and education level; the latter was included given consistent associations between higher levels of education and better cognitive performance. Subjective health (how participants perceived their own health) was also included, given that it has been previously examined in research on subjective age bias (Stephan, Caudroit, & Chalabaev, 2011). Marriage length (in years) was also selected given that it was likely to be related to both relationship quality and age.
Variables measured during MIDUS Biomarker data collection that were related to physical health and cardiovascular status were also selected: diagnosis of heart disease (0 = no diagnosis, 1 = diagnosis), diagnosis of hypertension (0 = no diagnosis, 1 = diagnosis), smoker status (0 = not a current smoker, 1 = current smoker), and body mass index (BMI).
Description of Analytic Sample
As introduced previously, analyses drew on a subset of participants from the MIDUS project. In particular, the analytic sample was smaller than the full sample of MIDUS participants due to our inclusion of participants who were married or in a marriage-like relationship and our focus on HRV as an outcome, which was only measured among MIDUS Biomarker participants.
Selection Analyses
We first assessed whether participants included in our analytic sample differed significantly from those who were not on key dimensions: chronological age, subjective age bias (chronological age – subjective age, described in greater detail below), relationship quality, memory performance, and resting HRV. Summary statistics for these variables (means and standard deviations), test statistics, and effect sizes, are displayed in Table S1 of the Supplemental Materials.
Participants across sub-samples did not differ in chronological age or subjective age bias. Participants in our analytic sample, on average, had better memory performance, lower resting HRV, and higher relationship quality at Wave 1 compared to those not included in our analytic sample. However, in all cases, these difference were marginally significant and were below the conventional threshold for small effect sizes (Cohen’s ds 2 (1, N = 4557) = 3.07, p = 0.08. There was a more equal number of male and female participants in the analytic sample (male: n = 336; female: n = 342) compared to those not in the analytic sample (male: n = 1778; female: n = 2101).