| Equal opportunity . | ||||
|---|---|---|---|---|
| Medium-risk threshold (≥15) . | High-risk threshold (≥45) . | |||
| Characteristic . | Difference (95% CI) . | Ratio (95% CI) . | Difference (95% CI) . | Ratio (95% CI) . |
| Age | 0.33 (0.30-0.35) | 0.57 (0.54-0.60) | 0.12 (0.09-0.15) | 0.59 (0.52-0.67) |
| Sex | 0.06 (0.04-0.09) | 0.90 (0.86-0.94) | 0.06 (0.04-0.08) | 0.76 (0.70-0.84) |
| Race | 0.07 (0.04-0.13) | 0.88 (0.80-0.93) | 0.09 (0.04-0.14) | 0.69 (0.53-0.86) |
| Ethnicity | 0.06 (<0.01-0.18) | 0.90 (0.75-0.99) | 0.04 (<0.01-0.14) | 0.82 (0.41-0.99) |
| Primary care provider | 0.13 (0.06-0.25) | 0.79 (0.59-0.90) | 0.15 (0.07-0.22) | 0.40 (0.13-0.73) |
| Past outpatient encounters | 0.22 (0.20-0.25) | 0.65 (0.61-0.69) | 0.17 (0.15-0.19) | 0.41 (0.36-0.47) |
| Equal opportunity . | ||||
|---|---|---|---|---|
| Medium-risk threshold (≥15) . | High-risk threshold (≥45) . | |||
| Characteristic . | Difference (95% CI) . | Ratio (95% CI) . | Difference (95% CI) . | Ratio (95% CI) . |
| Age | 0.33 (0.30-0.35) | 0.57 (0.54-0.60) | 0.12 (0.09-0.15) | 0.59 (0.52-0.67) |
| Sex | 0.06 (0.04-0.09) | 0.90 (0.86-0.94) | 0.06 (0.04-0.08) | 0.76 (0.70-0.84) |
| Race | 0.07 (0.04-0.13) | 0.88 (0.80-0.93) | 0.09 (0.04-0.14) | 0.69 (0.53-0.86) |
| Ethnicity | 0.06 (<0.01-0.18) | 0.90 (0.75-0.99) | 0.04 (<0.01-0.14) | 0.82 (0.41-0.99) |
| Primary care provider | 0.13 (0.06-0.25) | 0.79 (0.59-0.90) | 0.15 (0.07-0.22) | 0.40 (0.13-0.73) |
| Past outpatient encounters | 0.22 (0.20-0.25) | 0.65 (0.61-0.69) | 0.17 (0.15-0.19) | 0.41 (0.36-0.47) |
Abbreviations: CI = Confidence Interval based on bootstrap re-sampling. Equal Opportunity defined as the maximum absolute difference (or minimum ratio) comparing the maximum and minimum true positive rate (sensitivity) across categories within each characteristic subgroup. Estimates for equalized odds, which also consider specificity, are not shown as differences in sensitivity in the performance of the EOL-CI dominate all subgroup comparisons.
| Equal opportunity . | ||||
|---|---|---|---|---|
| Medium-risk threshold (≥15) . | High-risk threshold (≥45) . | |||
| Characteristic . | Difference (95% CI) . | Ratio (95% CI) . | Difference (95% CI) . | Ratio (95% CI) . |
| Age | 0.33 (0.30-0.35) | 0.57 (0.54-0.60) | 0.12 (0.09-0.15) | 0.59 (0.52-0.67) |
| Sex | 0.06 (0.04-0.09) | 0.90 (0.86-0.94) | 0.06 (0.04-0.08) | 0.76 (0.70-0.84) |
| Race | 0.07 (0.04-0.13) | 0.88 (0.80-0.93) | 0.09 (0.04-0.14) | 0.69 (0.53-0.86) |
| Ethnicity | 0.06 (<0.01-0.18) | 0.90 (0.75-0.99) | 0.04 (<0.01-0.14) | 0.82 (0.41-0.99) |
| Primary care provider | 0.13 (0.06-0.25) | 0.79 (0.59-0.90) | 0.15 (0.07-0.22) | 0.40 (0.13-0.73) |
| Past outpatient encounters | 0.22 (0.20-0.25) | 0.65 (0.61-0.69) | 0.17 (0.15-0.19) | 0.41 (0.36-0.47) |
| Equal opportunity . | ||||
|---|---|---|---|---|
| Medium-risk threshold (≥15) . | High-risk threshold (≥45) . | |||
| Characteristic . | Difference (95% CI) . | Ratio (95% CI) . | Difference (95% CI) . | Ratio (95% CI) . |
| Age | 0.33 (0.30-0.35) | 0.57 (0.54-0.60) | 0.12 (0.09-0.15) | 0.59 (0.52-0.67) |
| Sex | 0.06 (0.04-0.09) | 0.90 (0.86-0.94) | 0.06 (0.04-0.08) | 0.76 (0.70-0.84) |
| Race | 0.07 (0.04-0.13) | 0.88 (0.80-0.93) | 0.09 (0.04-0.14) | 0.69 (0.53-0.86) |
| Ethnicity | 0.06 (<0.01-0.18) | 0.90 (0.75-0.99) | 0.04 (<0.01-0.14) | 0.82 (0.41-0.99) |
| Primary care provider | 0.13 (0.06-0.25) | 0.79 (0.59-0.90) | 0.15 (0.07-0.22) | 0.40 (0.13-0.73) |
| Past outpatient encounters | 0.22 (0.20-0.25) | 0.65 (0.61-0.69) | 0.17 (0.15-0.19) | 0.41 (0.36-0.47) |
Abbreviations: CI = Confidence Interval based on bootstrap re-sampling. Equal Opportunity defined as the maximum absolute difference (or minimum ratio) comparing the maximum and minimum true positive rate (sensitivity) across categories within each characteristic subgroup. Estimates for equalized odds, which also consider specificity, are not shown as differences in sensitivity in the performance of the EOL-CI dominate all subgroup comparisons.
This PDF is available to Subscribers Only
View Article Abstract & Purchase OptionsFor full access to this pdf, sign in to an existing account, or purchase an annual subscription.
