Has sampling bias been ruled out? With people more open to discussing public mental health in recent years, certainly more people have been seeking treatment and diagnoses have probably improved. I would assume historical data strongly under sampled people with mental health issues due to the great stigma and misunderstandings about these diseases.