Research inhabitants
Details had been from the NHANES, a big, nationally representative cross-sectional study and physical examination conducted in 2-yr cycles, assessing the health and dietary position of the civilian, noninstitutionalized United States population. Information of the study and laboratory strategies are released somewhere else37. The complete cognitive evaluation was readily available among the 2937 adults ≥ 60 years old from two continual NHANES facts releases: 1364 from the 2011–2012 cycle and 1573 from the 2013–2014 cycle. Our samples excluded 1624 members without having details on lifestyle aspects, SES elements, and covariates. Our closing analytic sample provided 1313 participants from the NHANES 2011–2014 (Figure S1).
Common protocol approvals, registrations, and affected person consents
The NHANES protocol followed the moral tips of the 1975 Declaration of Helsinki and was authorised by the National Heart for Wellbeing Data Analysis Ethics Assessment Board, and composed educated consent was attained from all individuals or lawfully authorized reps. The Nationwide Center for Overall health Data Investigation Ethnics Overview Board permitted all review protocols, and composed educated consent was acquired from all members. All strategies were done in accordance with appropriate pointers and regulations and adopted the Strengthening the Reporting of Observational Scientific tests for Epidemiology (STROBE) rules38.
Lifestyle component evaluation
We calculated a composite healthful way of living rating consisting of multiple domains, together with self-documented cigarette smoking cigarettes, alcoholic beverages use, bodily activity, and diet, which align with past NHANES scientific studies and tips set forth by the Environment Overall health Business39,40,41. Facts on all life-style factors had been collected employing structured questionnaires or 24-h nutritional remember. Cigarette smoking was categorized into in no way cigarette smoking (smoked less than 100 cigarettes in lifetime) or using tobacco background, with by no means using tobacco viewed as a balanced level. Liquor use was defined by every day intake of liquor, and a wholesome stage was outlined as day-to-day intake of no a lot more than one particular drink for women of all ages and no far more than two beverages for adult men42. Bodily exercise was assessed weekly by metabolic equivalent hrs of leisure time (Satisfied). Achieved was categorized into tertiles, and we viewed as members in the prime tertile to show evidence of suitable bodily action (healthful level).
Diet regime was characterised by the wholesome eating index-2015 (HEI-2015)43. The HEI-2015 is a diet plan quality index to assess the adherence to 2015–2020 Nutritional Pointers for Us residents43. Ingestion of every HEI-2015 ingredient was scored proportionately in between the minimal and maximum standards. The facts of HEI-2015 parts and scoring specifications are shown in Desk S1. HEI-2015 scores variety from to 100. A healthier diet plan was outlined as HEI-2015 scores in the top rated 40%, though those people in the base 60% confirmed evidence of unhealthy ingesting39,40.
We made a composite nutritious life-style score. For each way of living aspect, we assigned 1 stage for a balanced amount and stage for an unhealthy degree. The healthful lifestyle score was calculated by summing each individual person life style variable rating, and feasible scores ranged from to 4. Bigger balanced life-style scores indicated better adherence to balanced behaviors. Healthier way of living scores have been classified into approximate tertiles, comprising 0–1, 2, and 3–4 wholesome behaviors.
SES evaluation
The current analyze centered on unique-stage SES, characterized by instruction, relatives cash flow, profession, wellness insurance policies, and meals stability44. Education concentrations had been categorized into fewer than superior university, substantial school or equivalent, and college or over. Family income degrees have been defined as the ratios of loved ones income to the poverty thresholds particular to the study 12 months. Household money stages had been categorised into lower earnings (spouse and children income to poverty ratio ≤ 1), center earnings (> 1 and < 4), and high income (≥ 4)45. Occupation was defined by each participant’s longest job, and categorized as blue collar (service, farming/forestry/fishing, precision production/craft/repair, and operator/fabricator/laborer), white collar (managerial/professional specialty, and technical/sales/administrative support), unemployed, or not in labor force (retired persons, homemakers, and students)46. Health insurance was classified into government insurance (Medicare, Medi-Gap, Medicaid, State Children’s Health Insurance Program, military health plan, Indian Health Service, state-sponsored health plan, or other government insurance), private insurance, both government and private insurance, or uninsured. Food insecurity was measured using 18-item Food Security Survey Module, and dichotomized into full food security and food insecurity47.
For composite SES, LCA was used to identify the presence of underlying constructs (or classes) in the observed SES components (education, occupation, income-to-poverty ratio, health insurance, and food security)48. LCA estimates conditional class membership probability and classifies individuals who are homogeneous in terms of particular criteria. We compared the performance of LCA with two, three, and four classes, and the optimal number of latent classes was determined based on the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC)48. We identified three latent classes—high, medium, and low SES groups.
Cognitive function assessment
Comprehensive cognitive testing was completed by participants aged ≥ 60 years during NHANES 2011–2012 and 2013–2014 cycles. Four tests were administered, including the CERAD Word Learning subtests to evaluate immediate and delayed learning ability49, the Animal Fluency test to assess categorical verbal fluency (component of executive function)50, and the DSST from the Wechsler Adult Intelligence Scale to assess processing speed, sustained attention, and working memory51. Individual results of the four tests (CERAD immediate and delayed learning, Animal Fluency, DSST) converted to z-scores using age appropriate normal means. Individual test z scores are averaged to form the cognitive composite z-score, similar to previous studies52.
Covariates
Covariates were selected based on previous research and included age, sex, race/ethnicity, marital status, birthplace, and BMI53. Sex included male and female. Participants were assigned to one of these categories by NHANES based on their questionnaire responses. Race/ethnicity categories provided by NHANES included non-Hispanic White, non-Hispanic Black, non-Hispanic Asian, Hispanic, or other races including multiracial. Marital status included never married, married, or not married (including widowed, divorced, separated, or living with partner). Birthplace was categorized as born in or outside of the United States. BMI (kg/ m2) was computed as the ratio of weight (kg) and height squared (m2).
Statistical analyses
Mean and standard error (SE) were computed for continuous variables, and percentage (%) was calculated for categorical variables. Participant characteristics, composite cognition z-scores, and scores of the four individual cognitive tests were compared by composite healthy lifestyle and composite SES groups. The sampling weights and design variables were used for all analyses.
In the primary analysis, we utilized survey-weighted linear regression models to examine the associations of healthy lifestyle score and composite SES derived from LCA with composite cognition z-score, after adjusting for age, sex, race/ethnicity, marital status, and birthplace. To assess the joint associations of lifestyles and SES, we classified participants into nine groups based on three SES classes (low, medium, high) and three healthy lifestyle groups (0–1, 2, 3–4). Differences in cognition z-score and 95% confidence intervals were calculated with participants with low SES and 0–1 health lifestyle as the reference group. Likelihood-ratio test comparing models with and without interaction terms between SES and healthy lifestyle groups was used to test the interaction between healthy lifestyle score and composite SES.
In the secondary analysis of individual SES and lifestyle factors, we performed linear regression models to estimate differences in composite cognitive z-score and 95% confidence intervals. Model 1 included age, sex, race/ethnicity, marital status, birthplace, and individual lifestyle factors (smoking status, alcohol consumption, physical activity, and HEI-2015). Model 2 included age, sex, race/ethnicity, marital status, birthplace, and individual SES factors (education, family income, occupation, health insurance, and food security). Model 3 included age, sex, race/ethnicity, marital status, birthplace, as well as both individual lifestyle and SES components.
To test the robustness of our findings, we conducted several sensitivity analyses. First, to assess the different cognitive domains, we determined the relationships between lifestyle and SES factors with z-scores of each of the four cognitive test scores. Second, to assess potential differential influences between healthy lifestyle factors on cognition, we computed a weighted healthy lifestyle score. The weighted healthy lifestyle score was constructed as the sum of lifestyle factor scores where weights are beta coefficients of each individual lifestyle factor derived from Model 1 in the analysis of individual lifestyle factors. We then used this weighted healthy lifestyle score in a multivariable regression analysis. Furthermore, BMI could be an intermediate factor linking exposures and cognition, so we did not adjust for BMI in our primary analyses. However, in sensitivity analyses, we additionally controlled for BMI. Moreover, we repeated our analyses in a subpopulation excluding participants with stroke or cardiovascular disease, which might impair cognitive functions54. We did not excluding these participants in the primary analysis in consideration of potential overadjustment bias because stroke and cardiovascular disease may serve as risk factors for cognitive decline and could be also influenced by SES and healthy lifestyle factors. Finally, to explore the impact of missing values on the observed results, we conducted multiple imputations with chain equation to impute missing values55. This procedure used PROC MI in SAS to create 20 datasets for missing values and computed pooled effect estimates using PROC MIANALYZE. We repeated Models 1–3 on each of the imputed datasets and compared the effect estimates to the primary findings using measured data. All statistical analyses were performed using SAS (version 9.4, SAS Institute Inc.). Statistical significance was set at a two-sided P < 0.05.
