In follow-up, those who have fared worst will be dead, unavailable, or uncooperative. Ellason and Ross obtained follow-up information on only 40% of their cohort. The high degree of selection in their study radically undermines their findings. We get no idea how many patients have deteriorated to suicide or died of self-neglect, have become incarcerated or hospitalized elsewhere, are homeless or cannot afford a telephone, or have suffered other causes of unavailability. Such patients would be unlikely to raise scores for improvement.