Table 2.
Logistic regression for receipt of screening
| Univariable | Final Multivariable Modela | ||
|---|---|---|---|
| OR (95 % CI) | OR (95 % CI) | ||
| Age | ≥50 versus <50 | 1.69 (1.31-2.18) | 1.68 (1.29-2.18) |
| Sex | Female vs Male | 0.94 (0.68-1.33) | |
| Race | Caucasian | Ref | |
| Black or African American | 1.31 (1.01-1.69) | ||
| Asian | 0.69 (0.29-1.62) | ||
| Other or unknown | 0.74 (0.45-1.21) | ||
| Hypertension | 2.65 (1.72-4.10) | 2.12 (1.35-3.32) | |
| Hyperlipidemia Diagnosis | 1.61 (0.58-4.46) | ||
| Diabetes mellitus | 2.22 (1.61-3.07) | 2.06 (1.48-2.87) | |
| Obesity | 2.62 (1.61-4.27) | 2.52 (1.51-4.19) | |
| BMIa | 1.04 (1.01-1.07) | ||
| BMI Categorya | Normal (18.5-24.9) | REF | |
| Overweight (25–29.9) | 1.66 (0.93-2.96) | ||
| Obese (≥30) | 2.17 (1.24-3.78) | ||
| Underweight (<18.5) | 0.40 (0.10-1.62) | ||
| Peripheral Arterial Disease | 1.92 (0.17-21.3) | ||
| Tobacco use | Current smoker vs non-smoker or past-smoker (n = 348) | 0.70 (0.41-1.18) | |
The c-statistic (equivalent to area under the curve) for the model was 0.63 for the association between the predicted probabilities and observed responses for the final multivariable model
aGiven the large amount of missing data for BMI and the risk for selection bias in using a complete case analysis, we have instead used a binary variable for obesity identified using ICD9 codes. The OR of 1.04 is for each unit increase in BMI