Table 3.

Proportional confusion matrix information for all four models after 5-fold cross-validation repeated 10 times. Accuracy is provided at the default classification cutoff point of 0.50, rounded to the nearest hundredth. Matrix information is provided in terms of average proportions of cell counts across resamples. For example, a true positive value of .50 would suggest that on average across resamples 50% of the observations were classified as positive cases when in fact they were positive cases (i.e., they were classified correctly). Positive classification in this case refers to a classification of “Good Outcome,” while negative refers to “Poor Outcome.” The summation of true positives and true negatives equals the accuracy, which is presented at the top alongside the corresponding 95% confidence interval.

Model Type:	14-Day Decision Tree	14-Day Random Forest	3-Month Decision Tree	3-Month Random Forest
Accuracy	0.745	0.787	0.720	0.730
95% CI:	[0.728, 0.761]	[0.771, 0.802]	[0.699, 0.739]	[0.710, 0.750]
True Positive Proportion:	.138	.108	.350	.371
True Negative Proportion:	.607	.679	.369	.359
False Positive Proportion:	.127	.054	.143	.153
False Negative Proportion:	.128	.158	.137	.117