Here's a thoughtful piece on self rated health, commenting on three papers in the current IJE and trying to reconcile their differing findings.
The basic truth is that people who rate their health as poor die sooner - self rated health predicts mortality. But why? Mostly, people go with their current physical functioning when rating their health. There is some - but not much - place for intellectually useful info such as whether they smoke too much / have high blood pressure etc. It's another example of how lay views of health diverge from professional views, and how complex any self-report data are.
Worth a slow read.