TABLE 3.
Binary Logistic Regressions Predicting Low Pain Medication Stress*
| Variable | b | SE | Significance | Odds ratio (95% CI) |
|---|---|---|---|---|
| Male | .206 | .472 | .662 | 1.229 (.487, 3.100) |
| African American | .773 | .568 | .174 | 2.167 (.711, 6.599) |
| Age | −.020 | .021 | .342 | .980(.940, 1.022) |
| Religious affiliation | −.383 | .505 | .449 | .682 (.253, 1.835) |
| Catholic | .148 | .741 | .841 | 1.160 (.272, 4.951) |
| Medicaid only | .399 | .754 | .597 | 1.490 (.340, 6.533) |
| Medicare only | 1.392 | .796 | .080 | 4.022 (.846, 19.133) |
| Uninsured | 2.176 | .813 | .007 | 8.815 (1.791, 43.381) |
| Medicaid and Medicare | −.372 | .793 | .639 | .689 (.146, 3.264) |
| DNR | −.462 | .899 | .607 | .630 (.108, 3.671) |
| Cancer | .102 | .477 | .831 | 1.107 (.435, 2.819) |
| Patient unaware of prognosis | ||||
| Unknown: | 2.617 | .875 | .003 | 13.696 (2.463, 76.155) |
| Yes: | .303 | .698 | .665 | 1.354 (.344, 5.320) |
| Length of stay (>3 months) | −1.593 | .527 | .002 | .203 (.072, .571)† |
| 1. Model χ2 (df) | 63.154 | (14) | <.0001 | |
| 2. Nagelkerke’sR-Square | .471 |
Initially, Pain Medication Stress with four ordinal categories (i.e., no pain medication, pain medication-no difficulty, pain medication-some difficulty, pain medication-a lot of difficulty) was predicted; however, the assumption of parallel slopes was violated (i.e., the predictors cannot be assumed to have the same impact across all ordinal thresholds). Therefore, Pain Medication Stress was re-specified by collapsing the first two and then the last two categories, and di-chotomous logistic regression was used. The prediction of low pain medication stress is reported because it yields positive b parameters that result in odds ratios >1.
The odds ratio and 95% CI become 4.919 (1.752, 13.814) when the dependent dummy variable was recoded to predict high Pain Medication Stress. When Length of Stay (>3 months) was dropped, the remaining parameters did not change appreciably.