Time trends of socioeconomic inequalities in adolescent smoking in Okinawa, Japan, 2008–2016: a repeated cross-sectional study

Smoking is a leading preventable cause of early death and disability worldwide [1]. In Japan, smoking is also the top risk factor contributing to mortality from non-communicable diseases [2]. As smoking mostly begins during adolescence, smoking prevention among adolescents is critical to decreasing mortality and morbidity [3]. To develop effective smoking prevention programs, monitoring time trends in smoking and understanding its determinants are needed.

In recent decades, time trends in smoking prevalence among young people have shown a steady decline at the global level [4]. The cross-national studies, such as the European School Survey Project on Alcohol and Other Drugs and the Health Behaviour in School-aged Children study have shown that a general decreasing trend in cigarette smoking was observed in many European countries [5, 6]. The prevalence of smoking among Japanese adolescents was also trending downward over time [7, 8]. Nationwide surveys of high school students showed that current smoking prevalence among boys was 30.7% in 1996 and 3.5% in 2014, whereas among girls it was 12.6% and 1.4%, respectively [7].

Although adolescent smoking prevalence is declining, this progress is not even across population subgroups [4]. For example, those with low socioeconomic status (SES) were more likely to smoke than those with high SES [9], but previous research has shown socioeconomic inequalities in adolescent smoking trends [10‐14]. Socioeconomic inequalities are often assessed by both absolute measures, such as prevalence differences and slope index of inequality, and relative measures, such as prevalence ratios, odds ratios, and relative index of inequality [15]. A study conducted in seven European countries that used educational tracks as the SES indicator found that trends in absolute educational differences in adolescent smoking (i.e., prevalence differences between vocational students and academic students) varied across the countries while relative educational differences (i.e., odd ratios for vocational students compared with academic students) had a tendency to increase in all countries between 2002 and 2010 [12]. A Danish study which assessed parents’ occupational social class found that the absolute socioeconomic differences in daily smoking among 15-year-olds increased from 1991 to 2006 and declined from 2006 to 2014 and the relative socioeconomic differences increased over 1991–2014 [10]. A Finnish study created a cumulative socioeconomic adversity variable which combined parental education, parental unemployment, and family structure and showed that the overall prevalence of adolescent smoking decreased between 2000 and 2015. However, no similar changes were observed among adolescents with most socioeconomic adversities, which resulted in the increased relative socioeconomic differences over time [11]. On the other hand, a German study which assessed parental SES (family affluence) and student’s own SES (school type) found that both socioeconomic differences in adolescent smoking remained unchanged between 1994 and 2002 [13]. In a Slovak study which measured parental education, the socioeconomic differences in adolescent smoking decreased among boys and increased among girls between 1998 and 2006 [14].

In Japan, socioeconomic differences in adolescent smoking were also shown, indicating that those with low parental education, not living with both parents, or attending vocational high schools were more likely to smoke [16]. However, no study has reported time trends in socioeconomic differences in smoking among Japanese adolescents so far. This is surprising because income inequality in Japan has increased over the past several decades [17] and Japan’s child poverty rate was 13.9 % in 2015, which was higher than the average across the Organization for Economic Co-operation and Development countries [18]. A study on socioeconomic trends among Japanese adults showed that smoking inequalities in both absolute and relative measures widened between 2000 and 2010, especially among women [19]. Yet, it is not clear whether the findings of adult populations in Japan can be true of adolescent populations.

Measurements of SES in adolescents are categorized into two dimensions: familial SES and adolescent’s own SES [13, 20]. Parental education and family structure are familial SES indicators and school type reflects the adolescent’s own SES [13, 20]. Each SES indicator was associated with adolescent smoking in previous studies [9, 11‐14, 20]. The poverty rate in two-adult households with at least one child in Japan was 10.7%, while that in single-adult households with at least one child was 50.8% in 2015 [18, 21]. Therefore, family structure is a suitable indicator of SES. In Japan, after completing compulsory education, nearly all students go on to upper secondary schools known as senior high schools. Of them, roughly 70% attend general high schools and 30% attend vocational and specialized high schools [22]. As the advancement rates to tertiary education in general and vocational high schools in 2016 were 64.1% and 20.9%, respectively [22], school type can predict the future education level.

In sum, previous findings on socioeconomic trends in adolescent smoking were inconsistent and suggested that socioeconomic trends in adolescent smoking may depend upon study population, survey years, and the type of the SES indicators. In this study, we analyzed time trends in socioeconomic differences in smoking among Japanese high school students between 2008 and 2016. We included parental education, family structure, and school type as the SES indicators and assessed both absolute and relative differences of these indicators.

Table 1 shows the distribution of study participants by sociodemographic characteristics and survey years. Chi-square tests indicate significant relationships between survey year and region, parental education, and family structure (P < 0.05).

Table 2 shows time trends in smoking prevalence by each SES indicator stratified by sex. Overall, smoking prevalence among boys and girls significantly declined from 11.5% and 6.2% in 2008 to 4.7% and 1.9% in 2016, respectively. The declining trends in smoking were observed for most of SES groups. However, there were no significant trends among boys with the highest parental education (P for trend = 0.183) and girls belonging to the “others” group for family structure (P for trend = 0.111). The prevalence in these groups remained constant over time. Table 2 also shows the results of chi-square tests of homogeneity. There were significant differences in smoking prevalence among most of the SES categories, indicating that smoking was more prevalent among low SES categories than high SES categories. However, parental education categories among boys in 2016 and family structure categories among boys in 2012 and both sexes in 2016 did not show any significant differences in smoking prevalence.

Table 3 shows time trends of the PDs and PRs in smoking for the SES indicators. The PDs for parental education in boys decreased from 8.5% (95% CI 4.8–12.3) in 2008 to 1.6% (95% CI −1.1 to 4.3) in 2016 (P for trend = 0.018). The PDs for family structure in girls also decreased from 6.2% (95% CI 2.5–9.9) in 2008 to 1.4% (95% CI −0.4 to 3.2) in 2016 (P for trend = 0.025). In addition, there were no significant absolute differences for parental education among boys in 2012 and 2016 and for family structure among both sexes in 2016. On the other hand, the PRs for school type in boys increased from 1.51 (95% CI 1.11–2.06) in 2008 to 4.18 (95% CI 2.50–7.00) in 2016 (P for trend < 0.001). Similarly, the PRs for school type in girls increased from 1.62 (95% CI 1.06–2.49) in 2008 to 3.50 (95% CI 1.62–7.59), but this trend was not statistically significant (P for trend = 0.053). For other SES indicators, the PDs and PRs in smoking among both sexes persisted over time. The adjusted models showed almost the same results as the crude models.

Table 4 shows time trends in the SII and RII for each SES indicator. We observed the same trends as the PDs and PRs. The SII for parental education in boys and family structure in girls decreased over time (P for trend = 0.005 and 0.006). The SII for family structure in boys also decreased, although it was not statistically significant (P for trend = 0.066). On the other hand, the RIIs for school type in boys and girls increased over time (P for trend < 0.001 and 0.054). The adjusted models showed the same results as the crude models.

This study showed that time trends in socioeconomic inequalities in smoking among Japanese adolescents varied by SES indicators as well as between absolute and relative measures. The absolute inequalities in smoking for parental education in boys decreased over time. This trend was due to a decreasing prevalence in the low parental education group while remaining unchanged at low levels in the high parental education group. The absolute inequalities for family structure in girls also decreased over time. In addition, the absolute differences for family structure in 2016 among both sexes disappeared, indicating no difference in smoking prevalence between students in intact families and students in non-intact families in the latest study period. Accordingly, the absolute inequalities in smoking for the familial SES indicators, such as parental education and family structure, seem to decrease from 2008 to 2016. Conversely, the relative inequalities in smoking for school type increased during the study period. As both of the simple and summary measures of inequality showed the same results, there was robust evidence for these trends in socioeconomic inequalities in smoking among Japanese adolescents.

The relative measure of socioeconomic inequality is likely to be unstable by a mathematical consequence when smoking prevalence is very low [12]. In fact, Holstein et al. [10] argued that “a high relative social inequality in smoking may not be so important if the smoking prevalence is low and/or declining rapidly.” Meanwhile, although the absolute measure may decrease when smoking prevalence decreases [30], it represents the excess number of adolescent smokers from lower SES groups [10], and thus is appropriate to evaluate the effect of public health policies to reduce socioeconomic inequalities in smoking [30]. As this study showed a low and decreasing prevalence of smoking, we gave more emphasis to the absolute measure when interpreting time trends in socioeconomic inequalities in smoking. The present finding about the narrowing in absolute inequalities in smoking for familial SES indicators among Japanese adolescents is a desirable result from the viewpoint of closing the gap. Our findings were partly consistent with a Slovak study [14] that found that the socioeconomic differences in smoking for parental education decreased among boys and increased among girls over time. However, some previous studies showed that time trends in smoking differences between familial SES groups have persisted or increased over time [10, 11, 13, 20].

A possible explanation for our findings is that recent changes in tobacco price policies in Japan might affect smoking inequalities among adolescents. A systematic review that assessed the equity impact of interventions/policies in youth smoking has concluded that increased tobacco price was the most effective in reducing socioeconomic inequalities in youth smoking, indicating that low SES youth were more responsive to tobacco price increase than high SES youth [32]. Likewise, in this study, it seems that the tobacco price increase might contribute to reduced smoking among Japanese adolescents from low familial SES groups, which resulted in the reduction of socioeconomic inequalities in smoking over time. During the study period, the tobacco price in Japan increased in 2010, 2014, and 2016. The price of the most popular cigarette brand in Japan, Mevius (the former name was Mild Seven before 2013) [33], increased from 300 yen in 2008 to 440 yen in 2016 (a 47% increase). A study among Japanese adults showed that the tobacco price increase in 2010 had a significant impact on smoking cessation across all socioeconomic subgroups and that this might be due to the affordable tobacco price for adults even after the price increase [33]. Yet, the increased tobacco price might have a greater impact on restricting tobacco purchase among Japanese adolescents who have less spending money, specifically the low familial SES groups.

Other tobacco control interventions/policies implemented during the study period were the installation of age verification vending machines using IC cards in 2008 [34] and the application of revised reference materials for school-based smoking, drinking, and drug abuse prevention instructions from 2010 to 2012 [35‐37]. A Japanese study showed that age verification cards could not completely prevent adolescents from accessing tobacco products because adolescent smokers borrowed the IC cards from someone to buy cigarettes through vending machines [34]. Although school-based smoking prevention programs contribute to reduce smoking initiation among adolescents [38], a review showed that school-based interventions had mixed equity results [32]. In this study, it is not clear whether these two interventions contribute to reduce socioeconomic inequalities in smoking among Japanese adolescents.

In contrast to the familial SES indicators, time trends in absolute inequalities for school type persisted among boys and girls while relative inequalities increased over time. Although smoking prevalence decreased in both general and vocational schools, the absolute differences between both school types remained unchanged throughout the study period. It is noted that if the prevalence of the health outcome declines, relative differences may increase while absolute differences remain constant [39]. This pattern was demonstrated for school type in this study. The present findings were partly consistent with findings from previous studies that found no changes or increases in relative smoking inequalities for school type over time [12, 13, 20]. This may suggest that the different trends in smoking between the familial SES indicators and school type could be attributable to differential effects of different dimensions of SES; that is, the familial SES indicators reflect the social class of origin and school type represents the student’s individual social position [13, 20]. To reduce smoking inequalities for school type, smoking prevention policies and interventions should focus on vocational high school students.

To our knowledge, this study is the first to examine time trends in socioeconomic inequalities in smoking among Japanese adolescents. Nonetheless, several limitations should be noted. First, we assessed adolescent smoking using a self-reported single question. Although the question’s reliability has been confirmed, it is uncertain whether the question most successfully measured current smoking. This point might bring misclassification or underestimation of adolescent smoking in this study. In addition, we did not ask if students had used new tobacco-related products, such as electronic cigarettes (e-cigarettes) and heat-not-burn tobacco products. Although cigarette smoking among adolescents has declined, e-cigarette use among adolescents has been increasing [40]. A recent study in Japan also pointed out that e-cigarette use is gaining popularity among Japanese adolescents and many adolescents use new tobacco-related products only [41]. Thus, the overall declining trends in smoking may suggest that new tobacco-related products might be replacing cigarette smoking [42]. Second, we could not assess household income as an SES indicator. Thus, time trends in smoking inequalities for direct monetary aspects of SES are unknown. Third, we could not examine which father’s or mother’s education level was more important to adolescent smoking. Previous studies suggested that father’s education, but not mother’s education was related to adolescent smoking [43] and father’s education might be a stronger familial SES indicator than mother’s education was [44]. The present study showed that more than 70% of parental education included father’s education and the proportions were stable over time (data not shown). Thus, our indicator of parental education may partially reflect the role of the familial SES in adolescent smoking. Finally, since the study participants were exclusively from public high schools in Okinawa Prefecture, which is one of the poorest prefectures in Japan, the generalizability of the present findings to adolescents in Japan as a whole may be limited. As a cross-national study reported that smoking inequalities were much larger in the lowest income countries [45], smoking inequalities in Okinawa, which presents the worst SES indices, such as prefectural income per capita, unemployment rate, and the proportion of senior high graduates going to further education [46], may be prominently manifested. Meanwhile, it is unknown how the extent of poverty may have affected the trends in smoking inequalities. At any rate, our findings obtained from the poorest area may be unique and valuable for understanding and tackling socioeconomic inequalities in adolescent smoking.

Smoking prevalence among Japanese adolescents has been declining and the absolute socioeconomic inequalities in smoking for parental education and family structure decreased between 2008 and 2016. In contrast, the relative socioeconomic inequalities for school type increased over time. Our findings suggest that time trends in socioeconomic inequalities in smoking varied by absolute and relative measures and that further policies and/or interventions to reduce smoking inequalities should focus on the context of schools, especially in vocational high schools.