1. Although in principle the presence of multiple observations on some subjects may introduce statistical dependencies that undermine the results from this logistic regression, a direct test for the presence of such dependencies using an exact probability method revealed no evidence for these. Similar results were obtained when the logistic regression was limited to the first observation in each case.
2. References, author details, and dates of receipt and acceptance are given at the end of the following companion paper.
3. Three with a chronic episode had a positive event and recovered soon afterwards, just outside the 12 months. These have been classed as without a positive event as the key interest here is to explain the persistence of depression.
4. While the life-table approach is preferable, the majority of studies reviewed here do not employ this technique. Nevertheless, in the light of our own findings, their results on one- and two-year outcome might well not differ significantly.