Further analysis of the data collected at age 42 in 2000 and age 46 in 2004 is shown
in Tables 7a and 7b. It can be seen that, for both men and women, the mean units of
alcohol drunk each week by those who report drinking on most days appears to have
declined dramatically. This is further evidence that the question in the telephone
interview has failed to measure accurately the amount of alcohol that individuals are
consuming.
Table 7a The mean number of alcohol units consumed by the reported
frequency of drinking: NCDS 2000 and 2004 surveys
|
Men_______________________ | ||||||
|
_______2000 (NCDS cohort)_______ |
2004 (NCDS cohort) | |||||
|
Frequency of drinking alcohol_______ |
Mean |
N_____ |
Std. Deviation |
Mean |
N___ |
Std. Deviation |
|
Most days_______________________ |
44.4 |
1380 |
_________38.8 |
24.3 |
1316 |
18.7 |
|
1,2,or 3 times/week_________________ |
20.6 |
3003 |
__________20.7 |
11.1 |
2412 |
11.6 |
|
2 or 3 times/month_________________ |
7.3 |
492 |
__________13.0 |
^B |
^B |
^B |
Table 7b The mean number of alcohol units consumed by the reported
frequency of drinking : NCDS 2000 and 2004 surveys
|
Women__________________ | ||||||
|
________2000 (NCDS cohort)________ |
2004 (NCDS cohort) | |||||
|
Frequency of drinking alcohol_______ |
Mean |
N____ |
Std. Deviation |
Mean |
N___ |
Std. Deviation |
|
Most days_______________________ |
19.8 |
860 |
____________17.9 |
15.2 |
914 |
____________10.7 |
|
1,2,or 3 times/week_________________ |
8.4 |
2810 |
_______________9.0 |
6.2 |
2306 |
______________4.8 |
|
2,or 3 times/month_________________ |
2.9 |
719 |
______________3.9 |
- |
- |
- |
Note: Compared with tables 3a and 3b, the coding differed in two respects: for the 2000
survey, the drinking frequency ‘1,2 or 3 times/month’ was split into 2 categories, and the
amount consumed was only asked if the frequency was 2 or 3 times/month. At the 2004
survey, the amount drunk was only asked if the frequency was ‘1,2 or 3 times a week’ or ’most
days’.
Impact of missing data on reported alcohol consumption
As was indicated in the background information about the 1958 and 1970 cohort
studies above, there has been some loss to the cohort samples over time. In part this
is due to the death or emigration of cohort members but there is also attrition due to
loss of contact and refusals. As Plewis et al (2004) have highlighted, the cohort
samples do not decline monotonically over time but rather some cohort members
rejoin the sample having not been included in a previous sweep. The following tables
therefore provide a summary of the mean weekly alcohol consumption, reported by
cohort members, disaggregated by sex and by whether the cohort member was
successfully interviewed in the next sweep of the study. This provides an initial
indication of the extent to which estimates of mean alcohol consumption may be
biased due to any association between levels of reported alcohol consumption and
the probability of not being successfully interviewed. It can be seen, for example, that
20