Further sub-categorise abstainers into lifetime abstainers and quitters

[ld-archer]

Past studies have found that those who abstain from alcohol consumption actually have a slightly higher risk of being diagnosed with certain chronic diseases than do moderate drinkers (cardiovascular disease - 1, cardiovascular disease - 2, dementia, and asthma).

HOWEVER, these results are somewhat contentious, with some arguing that the studies have studied abstainers as a whole group, where there should be a distinction between those who are lifetime abstainers and those who have quit for any reason. Frisher et al. (2015) found (using ELSA!) that quitters were much more likely to report being in poor self-rated health than lifetime abstainers (although both types of abstainers were more likely to report poor self-rated health than any other drinking profile, from infrequent moderates to frequent high-risk). Also, Knott et al. (2015) found that there is no link between low-moderate alcohol consumption and reduced risk of mortality when compared with lifetime abstainers. They conclude that previous studies that have shown a reduced risk in this group most likely used an inappropriate reference group (all abstainers) or did not properly account for confounding variables.

Therefore, it's probably a good idea to separate our abstainers into two groups: lifetime abstainers and quitters. Based on the fact that Frisher et al. managed to categorise abstainers into these two categories it should be possible (however they only used data from waves 0 (HSE) and 5, and also never fully explain how they put people into these categories). My initial thoughts are that we can't specify truly lifetime abstainers (don't have the data), but we can split people by whether they have never drank whilst in the survey. A quitter will also have to refer to someone who was a drinker in their first wave in the survey, but at a point stops drinking and never restarts.

There is also an argument to consider the occasional drinker and categorise these people separately from other drinkers. This can be step 2 of this issue.

Step 1

• [ ] Categorise abstainers into survey abstainers and quitters
• [ ] See whether these two groups have a difference in terms of chronic disease outcomes in ELSA
• [ ] Check coefficients of these vars in transition models for some chronic diseases

Step 2

• [ ] Look at the occasional drinker as a separate group, see if they have similar coefficients to the quitters and consider merging if they do

[ld-archer]

Unfortunately deciding to follow a different route, and not going to retry the consumption based variables for alcohol as the underlying data on consumption has a lot of problems. Replacing alcohol consumption information with a frequency variable (scako).