Can machine learning be useful as a screening tool for depression in primary care?