Developing a random forest algorithm to identify patent foramen ovale and atrial septal defects in Ontario administrative databases

Table 2 Description and performance of final random forest model to identify PFO patients

Model	Description		Accuracy		Sensitivity		Specificity
Model	Description		Test	Train	Test	Train	Test	Train
7	Demographics Age group Sex Comorbidity flags (< 5 years) AF CAD CHF COPD DM HTN Migraine Other CHD admissions Emb.*	Stroke/TIA Number of events < 5 years prior to closure Ischemic stroke Hemorrhagic stroke TIA Intervention codes Top 10 (yes/no) Charlson comorbidity index	0.946	0.756	0.908	0.657	0.978	0.848
7 (tuned)	Same variables as model 7 (above), but with hyperparameters tuned: mtry = 3 Classification threshold cut-off = 0.38,0.62		0.918	0.757	0.896	0.751	0.936	0.763

*Emb. peripheral arterial embolism, pulmonary embolism, or deep vein thrombosis

ISSN: 1472-6947