Consideration of confounding was suboptimal in the reporting of observational studies in psychiatry: a meta-epidemiological study.