Time series analysis may have been more informative
Dear Editor,
We are writing in response to Cavelaars et al paper concerning the
prevalence of smoking among high and low educated groups in twelve
European countries.
The authors have identified several limitations regarding the
comparability of the international data used. One limitation that they
have overlooked is that the data were collected over eight years, whereas
the application of these data to the smoking epidemic model implies the
study of a single point in time ("around 1990"). Adjustments could have
been made for the year of collection, sampling source and interview method
to improve the validity of the results. In particular, the Swiss data were
collected by telephone which could create significant bias against the
inclusion of the lower educated group.
The smoking epidemic model is a linear one, describing the trends of
smoking prevalence among a population over time. The use of cross-
sectional data is, therefore, not an ideal way of demonstrating it. We
suggest that it would be more useful to analyse time series data from each
country. This could then be used to chart the smoking prevalence in the
population and the ratio of smoking rates between the two educational
groups over time. In countries found not to correlate with the four stage
model it would be interesting to look at factors which may account for
this, such as cigarette pricing, advertising restrictions, health
education and awareness of health risks. This may give some insight into
prevention strategies, especially in low socioeconomic groups and women.
Rapid Response:
Time series analysis may have been more informative
Dear Editor,
We are writing in response to Cavelaars et al paper concerning the
prevalence of smoking among high and low educated groups in twelve
European countries.
The authors have identified several limitations regarding the
comparability of the international data used. One limitation that they
have overlooked is that the data were collected over eight years, whereas
the application of these data to the smoking epidemic model implies the
study of a single point in time ("around 1990"). Adjustments could have
been made for the year of collection, sampling source and interview method
to improve the validity of the results. In particular, the Swiss data were
collected by telephone which could create significant bias against the
inclusion of the lower educated group.
The smoking epidemic model is a linear one, describing the trends of
smoking prevalence among a population over time. The use of cross-
sectional data is, therefore, not an ideal way of demonstrating it. We
suggest that it would be more useful to analyse time series data from each
country. This could then be used to chart the smoking prevalence in the
population and the ratio of smoking rates between the two educational
groups over time. In countries found not to correlate with the four stage
model it would be interesting to look at factors which may account for
this, such as cigarette pricing, advertising restrictions, health
education and awareness of health risks. This may give some insight into
prevention strategies, especially in low socioeconomic groups and women.
Yours sincerely,
Claire Dobson, Fiona Craig, Jenny Brown, Susannah McLean
Claire.Dobson@ncl.ac.uk
Stage 3 Medical students
Department of Epidemiology and Public Health,
The Medical School,
University of Newcastle upon Tyne
Competing interests: No competing interests