Table 3 Pooled AUCs of readmission prediction algorithms
Model | Under 1-year readmission | 1-year and more readmission | Not specified timing | Overall | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
N | Pooled AUC | Confidence interval | N | Pooled AUC | Confidence interval | N | Pooled AUC | Confidence interval | N | Pooled AUC | Confidence interval | |||
Random forest | 8 | 0.682 | 0.591–0.774 | 4 | 0.721 | 0.614–0.828 | 10 | 0.698 | 0.612–0.784 | 20 | 0.688 | 0.630–0.746 | ||
Logistic regression | 7 | 0.622 | 0.557–0.686 | 3 | 0.649 | 0.535–0.763 | 9 | 0.682 | 0.642–0.722 | 19 | 0.652 | 0.611–0.692 | ||
Support vector machine | 3 | 0.764 | 0.617–0.910 | 1 | NA | NA | 6 | 0.740 | 0.610–0.870 | 10 | 0.733 | 0.647–0.756 | ||
Gradient boosting | 7 | 0.645 | 0.583–0.706 | 2 | 0.673 | 0.518–0.829 | 9 | 0.767 | 0.697–0.836 | 18 | 0.702 | 0.649–0.758 | ||
Neural Network | 3 | 0.580 | 0.558–0.602 | 2 | 0.660 | 0.520–0.799 | 4 | 0.671 | 0.512–0.830 | 7 | 0.647 | 0.567–0.727 | ||
K-nearest neighbors | 3 | 0.716 | 0.544–0.887 | 1 | NA | NA | 0 | NA | NA | 3 | 0.706 | 0.527–0.885 | ||
Lasso regression | 0 | NA | NA | 0 | NA | NA | 4 | 0.715 | 0.626–0.804 | 5 | 0.716 | 0.633–0.798 | ||
Decision tree | 3 | 0.618 | 0.558–0.679 | 0 | NA | NA | 1 | NA | NA | 3 | 0.618 | 0.558–0.679 | ||
naïve bayes | 2 | 0.656 | 0.591–0.720 | 0 | NA | NA | 1 | NA | NA | 3 | 0.618 | 0.558–0.679 | ||