Table 4 Summary of multiple logistic regression analyses and regression trees. PPE: probability of prediction error
From: Are there specific clinical characteristics associated with physician’s treatment choices in COPD?
Analysis | Multiple logistic regression analyses | Penalized logistic regression analyses | Regression trees | |||
|---|---|---|---|---|---|---|
Determinants | PPE | Determinants | PPE | Determinants | PPE | |
A vs BCDE (no maintenance vs at least one) | FEV1, FVC | 0.19 | FEV1, FVC, gender, age, CAT, DIRECT | 0.17 | FEV1, FVC | 0.19 |
B vs CDE (one vs > 1 maintenance treatment) | FEV1, FVC, gender, exacerbation history | 0.23 | FEV1, FVC, gender, DIRECT, exacerbation history | 0.22 | FEV1, age, CAT, HAD, exacerbation history | 0.20 |
BC vs DE (ICS vs no ICS in patients with a maintenance treatment) | FEV1, exacerbation history | 0.34 | FEV1, exacerbation history, age, gender, HAD, CAT, DIRECT | 0.34 | FEV1, FVC, age, HAD, exacerbation history | 0.28 |
E vs BCD (triple vs other maintenance treatment schemes) | FEV1, FVC, exacerbation history | 0.32 | FEV1, gender, HAD, DIRECT, exacerbation history | 0.32 | FEV1, DIRECT, exacerbation history | 0.29 |