From: Early prediction of the risk of scoring lower than 500 on the COMLEX 1
Model | Training data (N = 728) | Testing data (N = 157) | ||
---|---|---|---|---|
Sensitivity | Specificity | Sensitivity | Specificity | |
Cutoff value of 0.5 | ||||
Multiple logistic regression | 82/170 (48.2%) | 528/558 (94.6%) | 17/38 (44.7%) | 110/119 (92.4%) |
Cutoff value of 0.25 | ||||
Multiple logistic regression | 129/170 (75.9%) | 475/558 (85.1%) | 26/38 (68.4%) | 105/119 (88.2%) |
Backward logistic regression | 135/170 (79.4%) | 447/558 (80.1%) | 25/38 (65.8%) | 105/119 (88.2%) |
Logistic regression of average scores | 134/170 (78.8%) | 430/558 (77.0%) | 27/38 (71.0%) | 99/119 (83.2%) |