Gender and age differences among current smokers in a general population survey

We used the "Mikrozensus", a cluster household sample representative for residents in Germany across the whole age range [12]. Data were collected in April 1999. At that time Germany was a country with very few activities in smoking prevention. Accordingly, there was an underdeveloped anti-smoking climate in the nation [13] and a lower intention to quit smoking compared to countries with more activities in smoking prevention [14]. Every member of each selected household was included into the study as a participant. Participation was mandatory by law for a core part of the assessment, mainly including questions about demographic data, housing and employment. Of the eligible individuals, 97.4 % participated in the core part of the study (N = 724,139). An additional, voluntary part included questions about tobacco smoking in a 45 % random subsample. Available for scientific use were the data of a 70 % random subsample of the 724,139 participants from the German Federal Statistical Office (N = 506,897) [12]. Among this subsample, 232,397 (45.8 %) had participated in the voluntary part of the Mikrozensus. We excluded children aged less than 10 years (n = 22,156). Among the remaining individuals, there were 30,769 with missing data for smoking status. They were excluded from the data analysis as well. Subjects with and without information about the smoking status did not differ in gender, age, school education and having been ill during the last four weeks prior to the interview, when effect sizes are considered. Among those without information about the smoking status, 52.7 % were women, and among those with information about smoking status 52.1 % were women (Likelihood chi² 3.0; not significant; Cohen's w .004 [15]). Among women without information about smoking the mean age was 45.5 (Standard deviation, Std, 21.9), and among women with information about smoking status it was 46.7 (Std 20.7; t-test; p < .001; d 0.06 [15]). Among women without information about the smoking status, there were 17.4 % aged 70 or older, and among women with information about the smoking status this figure was 16.6 (Likelihood chi² 632.3; p < .001; Cohen's w .08 [15]). Among men without information about smoking the mean age was 41.0 (Std 18.9), and among men with information about the smoking status it was 43.6 (Std 19.1; t-test, p < .001; d 0.13 [15]). Among men without information according to smoking status, there were 8.1 % at the age of 70 or older, among men with information about smoking status 9.9 % (Likelihood chi² 420.0; p < .001; Cohen's w .06, [15]). For school education Cohen's w was .06 for women and .05 for men, for having had a disease during the last four weeks prior to the interview Cohen's w was .02 for women and .03 for men indicating no effect. The final sample consisted of 179,472 residents, among them 78,959 ever smokers (44.0 %). The data of these female and male ever smokers were used in our analysis.

The final sample is representative for the general population of Germany aged 10 or older with respect to gender and age [16]: In the 10 year age groups, the maximum deviance of the proportion of women in the final sample from the proportion of women in the same 10 year age groups of the general population was 2.4 percentage points (mean deviance: 1.0 percentage points). The maximum deviance in the distribution of gender over the 10 year age groups was 1.3 percentage points (mean deviance: 0.6 percentage points) among women, among men the maximum deviance was 1.7 percentage points (mean deviance: 0.7 percentage points).

Every person of the household able to understand and to answer the interview questions and present in the household when the interviewer showed up responded in the face-to-face interview. The persons gave information about those household members that were not present. If nobody could be contacted personally the interviewer left a questionnaire to be filled in for all household members. The interview and the questionnaire included the same questions. According to smoking, the individuals were asked: "Are you currently a smoker?" (Yes, regularly/Yes, occasionally/No), if not: "Did you smoke in the past?" (Yes, regularly/Yes, occasionally/No). If the respondent answered "Yes" to one of the two questions s/he was asked: "How old have you been when you started smoking?" (age in years). "What do/did you smoke predominantly?" (cigarettes, cigars, pipe tobacco). "How many cigarettes do you/did you smoke per day?" (less than 5, 5 – 20, 21 – 40, more than 40). The data included the age of the respondent in years, and those aged 95 or older as one group. The individuals were asked "Have you been ill during the last 4 weeks (including today) or have you been injured by accident?" (Yes, ill/Yes, injured by accident/No), if yes: "How long did your illness or your injury last? (7 categories that were recoded to: 1 year or less, more than 1 year). School education was assessed by the respondent's school graduation (not including university degrees; degrees 1–5 for lowest to highest education). Income per individual was assessed by 18 categories.

For the bivariate data analysis, means with standard deviations, proportions and Likelihood chi² tests were used. For proportions, the effect size estimate Cohen's w was included, and values .10 or higher were interpreted as indicating an effect [15]. For means, the effect size measure d was used with values 0.20 to 0.49 interpreted as indicating a small and values 0.50 to 0.79 indicating a medium effect [15]. For the multivariate data analysis, logistic regression analysis was performed. SPSS 13.0 was used for all analyses.

According to the bivariate data analysis, among the 179,472 residents there were 78,959 ever smokers (44.0 %). Among the 93,588 women, there were 33.4 %, among the 85,884 men 55.6 % ever smokers, more among men than among women (Chi² 9031.6; df 1; p < .001; w .23). The female ever smokers included 64.0 %, the male ever smokers 58.7 % current smokers (Chi² 224.3; df 1; p < .001; w .05). For age groups below 40, the data revealed lower proportions of current smokers among the female ever smokers (Table 1). For age groups 40 or older, there were higher proportions of current smokers among female than among male ever smokers with effects in the age groups between 60 and 89 years. No effect existed among the age groups 10 – 19 and 30 – 59.

The mean age at onset of smoking among female ever smokers aged 40 to 59 was higher than among male ever smokers aged 40 to 59 with a small and among those aged 60 or older with a medium effect (Table 2). At the age of 50 or older women disclosed a higher proportion of subjects who smoked 5 – 20 cpd than men with an effect.

The proportion of current smokers among the ever smokers decreased by age (women: Chi² 2463.6; p < .001; w .28; men: Chi² 9103.7; p < .001; w .42). A lower age of onset of smoking was associated with a lower proportion of current smokers among ever smokers (women: Chi² 297.2; p < .001; w .10; men: Chi² 379.5; p < .001; w .09). Smokers who had been ill during the last four weeks prior to the interview showed a lower proportion of current smokers than those who said that they had not been ill. Those with a disease of 1 year or longer showed a lower proportion of current smokers than those who had been ill for a shorter time (women: Chi² 212.4; p < .001; w .08; men: Chi² 561.1; p < .001; w .11).

According to the multivariate data analysis, there was an interaction effect of gender and age for the odds of being a current smoker (Table 3). Women who were 80 to 89 years old had an odds ratio of 4.3 (CI 2.4 – 7.6) to be current smokers compared to men at this age. No interaction was found for the age of onset of smoking, for cigarettes per day or disease with gender.