Study population
The Consortium for the Longitudinal Evaluation of African Americans with Early Rheumatoid Arthritis Registry (CLEAR) was established by the National Institute of Arthritis and Musculoskeletal and Skin Disease (NIAMS) in order to provide the research community with thorough data on traditionally under-represented African Americans. Additional information about the CLEAR consortium can be found in the literature [22] and online [23].
This study focuses on the cross-sectional arm of the registry, CLEAR II, which collected parental socioeconomic data. Self-identified African-Americans over the age of 18 qualified for enrollment if they provided informed consent, had RA following the American College of Rheumatology definition, and no concurrent rheumatic disease diagnostic. Recruitment took place between 2006 and 2011 at the University of Alabama at Birmingham (Birmingham, Alabama), Emory University (Atlanta, Georgia), the Medical University of South Carolina (Charleston, South Carolina), the University of North Carolina at Chapel Hill (Chapel Hill, North Carolina), and Washington University (St. Louis, Missouri). The present study includes all CLEAR II participants with complete data on sociodemographic information (n = 516, Fig. 1). CLEAR II was approved by the institutional review boards of each of the five participating institution listed above. All materials and methods for this study were approved by University of North Carolina at Chapel Hill biomedical institutional review board.
Measures
Socioeconomic status
Current and childhood SES were assessed for participants and for their parents at the time of participant childhood, using educational and homeownership data collected in CLEAR II questionnaires.
Participant education was collected by asking “What is the highest degree or level of school you have completed?” and was dichotomized as greater than high-school (some college, but no degree; associate’s degree; bachelor’s degree; graduate school or degree; postgraduate school or degree) or high school and lower (8th grade or less; some high school; no diploma; high school graduate or equivalent;). Current homeownership, collected by asking “Do you own your own home?”, was classified as homeowner (Yes), or non-homeowners (No).
CLEAR II questionnaires collected parental education for both parents, asking participants to recall the highest grade of schooling their mother and father had completed when they were born. Data on the father was missing for one in four respondents, and we used maternal education to represent this dimension of childhood SES, when available. Where data was missing on the mother, we used father’s education (N = 10), or education of designated primary caretaker absent of data on either parent (N = 7). Parental education was dichotomized as high-school or greater (high school graduate or equivalent; some college, but no degree; associate’s degree; bachelor’s degree; graduate school or degree; postgraduate school or degree) or lower than high school (8th grade or less, Some high school, no diploma). The different cutpoints for participant and parental education were used to maintain comparable cell sizes of approximately 50 % in high and low education categories, and reflect the rapid rise in high-school graduation rates throughout the first half of the twentieth century [24].
Parental homeownership was collected by asking “When you were a child, did your parents, or the persons who raised you, own or were they buying their home, paying rent, or did they have some other living arrangement, such as living with relatives?” and it was classified, as for participant homeownership, as homeowner (Own or buying), or non-homeowner (Paying rent, Other living arrangement).
Self-reported health status
Self-reported health outcomes included in this study were the Health Assessment Questionnaire (HAQ), the helplessness subscale of the Rheumatology Attitudes Index (RAI), and Visual Analog Scale ratings for pain and fatigue (VAS).
The HAQ is a validated disability index widely-used in the monitoring of rheumatic diseases [25]. It consists of 20 items assessing a patient’s functional ability in eight domains of daily living: dressing and grooming (2 items), arising (2 items), eating (3 items), walking (2 items), hygiene (3 items), reach (2 items), grip (3 items), and usual activities (3 items). Questions are answered on a four-level scale (0: without any difficulty, 3: unable to do), and the highest ratings for each domain are averaged into a final score out of three, where higher scores indicate greater disability.
The RAI helplessness subscale has participants rate the following five statements on a 5-point integer scale (1: strongly disagree, 5: strongly agree):”My condition is controlling my life”, “I would feel helpless if I couldn’t rely on other people for help with my condition”, “No matter what I do, or how hard I try, I just can’t seem to get relief from my pain”, “I am not coping effectively with my condition”, and “It seems as though fate and other factors beyond my control affect my condition”. The five items are averaged into a final score, where higher values indicate greater levels of helplessness. The RAI is a validated indicator of perceived control over rheumatic conditions and is acceptably reliable for research and population screening purposes [26, 27].
Pain and fatigue VAS scores are produced by having participants rate their symptoms, respectively of pain or fatigue, along a continuous 10 cm line, with 0 cm corresponding to no pain/fatigue, and 10 cm meaning pain/fatigue as bad as could be. VAS measures of pain and fatigue have good test-retest reliability and internal validity [28, 29].
Covariates
The following variables were considered as control variables: age (continuous years); sex, body mass index (BMI, kg/m2); disease duration (continuous years); smoking status (never, former or current); current methotrexate/leflunomide use (yes/no) and current biologic agent use (yes/no).
Statistical models
We used linear regression models to examine the relationships between SES and each of the four outcome measures and included random effects to account for possible intra-site correlations. Our first model concurrently included participant education with parental education. The second model included participant homeownership and parental homeownership. The final model included all four socioeconomic factors. Backward elimination of covariates in multivariable models was carried out and change in estimate and likelihood ratio tests (LRT) were used to select covariates for the final model.
We report parameter estimates (β) and their 95 % confidence intervals (CI) comparing low-SES to high-SES (reference) for each outcome variable, as well as p-values from the LRT comparing the first two models to that including all four SES measures. Tests of statistical significance were two-sided and are considered significant at the p = 0.05 level. All statistical analyses were completed using the statistical software package SAS 9.4 (SAS Institute Inc. Cary, NC).