All analytical procedures were conducted using Statistical Analysis Software (SAS Version 9.2; Cary, NC). A complete-case framework was used.
We used multivariate logistic regression for survey data (PROC SURVEYLOGISTIC; SAS Institute, Cary, NC) to examine the association between aspirin use and sociodemographic variables. Two separate analyses were conducted, one of participants for whom aspirin therapy was indicated and the other for participants for whom it was not. The outcomes were modeled dichotomously, as regular vs nonregular aspirin users, and a collinearity check was done. 21
Initially, we created univariate models to gauge the crude relationship between each variable and aspirin use. Any variable with P<.20 in its univariate association with regular aspirin use was considered for inclusion in the final multivariate regression model. In the multivariate analyses, we sequentially eliminated variables with the weakest association with aspirin use until only significant (P<.10) independent predictors remained. Appropriate weighting was applied based on survey strata and cluster structure.19
RESULTS
Of the 831 participants who met the eligibility criteria for our analysis, 268 (32%) had an aspirin indication. TABLE 1 shows the key characteristics of the analytical sample, stratified by those for whom aspirin was indicated and those for whom it was not. The sample was primarily middle-aged (mean age 52.4±0.36) and non-Hispanic white (93%). Compared with those for whom aspirin therapy was not indicated, the group with an aspirin indication was significantly older (56.9 vs 50.3) and had a significantly higher proportion of males (97% vs 19%). As expected, those for whom aspirin was indicated were also at higher risk for CHD and stroke, most notably as a result of significantly higher systolic BP (131.9 vs 121.5 mm Hg) and lower HDL cholesterol (42.5 vs 52.6 mg/dL) compared with participants without an aspirin indication.
TABLE 1
Study sample, by sociodemographic variable and aspirin indication
| Variable | Full sample (N=831) | Aspirin indicated (n=268) | Aspirin not indicated (n=563) |
| Mean age, y | 52.4 | 56.9 | 50.3 |
| Sex, n Male Female | 367 464 | 259 9 | 108 455 |
| Race/ethnicity, n White, non-Hispanic Nonwhite/Hispanic | 776 55 | 252 16 | 524 39 |
| Marital status, n Married/partnered Not married or partnered | 637 194 | 215 53 | 422 141 |
| Health insurance, n Uninsured Insured | 76 755 | 26 242 | 50 513 |
| Education, n ≤High school Associate’s degree ≥Bachelor’s degree | 217 312 302 | 77 107 84 | 140 205 218 |
| Employment, n Unemployed Student/retiree/home Employed | 98 147 586 | 33 52 183 | 65 95 403 |
When aspirin was indicated, use was linked to age and sex
In the group with an aspirin indication (n=268), 83 (31%) reported taking aspirin most days of the week. The initial examination of sociodemographic variables showed that age, sex, and employment status demonstrated significant univariate associations with regular aspirin use (TABLE 2). In the multivariate model, however, the odds of regular aspirin use were significantly greater among participants who were older (odds ratio [OR], 1.07; P<.001) or female (OR, 3.49; P=.021) compared with participants who were younger or male, respectively.
TABLE 2
Participants who have an aspirin indication: Association between sociodemographic variables and regular aspirin use
| Variable | Regular aspirin use, OR (95% CI) | P value* |
| Age Older vs younger | 1.07 (1.04-1.11) | .001 |
| Sex Female vs male | 3.89 (1.42-10.67)† | .008 |
| Race/ethnicity Nonwhite/Hispanic vs white non-Hispanic | 0.55 (0.09-3.47) | .526 |
| Marital status Not married/partnered vs married/partner | 0.83 (0.36-1.95) | .678 |
| Health insurance Uninsured vs insured | 0.86 (0.50-1.47) | .579 |
| Education ≥Bachelor’s degree vs ≤high school Associate’s degree/some college vs ≤high school | 1.58 (0.75-3.34) 1.36 (0.74-2.49) | .234 .325 |
| Employment Student or retired vs employed Unemployed vs employed | 2.96 (1.74-5.03) 0.62 (0.25-1.56) | .001 .314 |
| *Significance was defined as P<.10. †Multivariate adjusted model: 3.49 (95% CI, 1.21-10.07; P=.021). CI, confidence interval; OR, odds ratio. | ||
When aspirin was not indicated, age and sex still affected use
Among the 563 participants for whom aspirin therapy was not indicated, 102 (18%) reported taking aspirin regularly. Age, sex, race/ethnicity, health insurance, and employment (TABLE 3), as well as region of residence and study enrollment year, had significant univariate associations with regular aspirin use; these variables were tested for potential inclusion in the multivariate model. In the final multivariate regression model, the odds of regular aspirin use were significantly greater among participants who were older (OR, 1.07; P<.001) and significantly lower among participants who were Hispanic or nonwhite (OR, 0.32; P=.063).
TABLE 3
Participants who do not have an aspirin indication: Association between sociodemographic variables and regular aspirin use
| Variable | Regular aspirin use, OR (95% CI) | P value* |
| Age Older vs younger | 1.07 (1.04-1.10) | .001 |
| Sex Female vs male | 1.60 (0.84-3.04) | .152 |
| Race/ethnicity Nonwhite or Hispanic vs white non-Hispanic | 0.23 (0.07- 0.73)† | .013 |
| Marital status Not married/partnered vs married/partnered | 1.00 (0.63-1.59) | .992 |
| Health insurance Uninsured vs insured | 0.36 (0.11- 1.15) | .086 |
| Education Bachelor’s or higher vs high school or less Associate’s/some college vs high school or less | 0.74 (0.35-1.57) 0.67 (0.38-1.17) | .431 .158 |
| Employment Student/retired vs employed Unemployed vs employed | 2.35 (1.32-4.20) 0.78 (0.26- 2.34) | .004 .652 |
| *Significance was defined as P<.10. †Multivariate adjusted model: 0.32 (95% CI, 0.10-1.06; P=.063). CI, confidence interval; OR, odds ratio. | ||