Abstract
Background
A huge amount of literature suggests that adolescents’ health-related behaviors tend to occur in clusters, and the understanding of such behavioral clustering may have direct implications for the effective tailoring of health-promotion interventions. Despite the usefulness of analyzing clustering, Italian data on this topic are scant. This study aimed to evaluate the clustering patterns of health-related behaviors.
Methods
The present study is based on data from the Health Behaviors in School-aged Children (HBSC) study conducted in Tuscany in 2010, which involved 3291 11-, 13- and 15-year olds. To aggregate students’ data on 22 health-related behaviors, factor analysis and subsequent cluster analysis were performed.
Results
Factor analysis revealed eight factors, which were dubbed in accordance with their main traits: ‘Alcohol drinking’, ‘Smoking’, ‘Physical activity’, ‘Screen time’, ‘Signs & symptoms’, ‘Healthy eating’, ‘Violence’ and ‘Sweet tooth’. These factors explained 67% of variance and underwent cluster analysis. A six-cluster κ-means solution was established with a 93.8% level of classification validity. The between-cluster differences in both mean age and gender distribution were highly statistically significant.
Conclusions
Health-compromising behaviors are common among Tuscan teens and occur in distinct clusters. These results may be used by schools, health-promotion authorities and other stakeholders to design and implement tailored preventive interventions in Tuscany.
Keywords: methods, public health, young people
Background
Priority health-compromising behaviors, which usually begin in adolescence, shape future adult behavior and are associated with an increased risk of morbidity and premature mortality, and thus high societal costs.1,2 As in other industrialized societies, risk factors such as smoking, alcohol use, insufficient consumption of fruit and vegetables and excessive screen time are highly prevalent among Italian teenagers.3 These single unhealthy behaviors may be either interrelated or unrelated, and distinguishing between them is of practical significance. Indeed, public health interventions targeting single unhealthy behaviors could have an impact on unrelated behaviors, while the effect of similar interventions on interrelated behaviors will probably be far from optimal.4 Nevertheless, Italian research on adolescent health promotion has often focused only on some epidemiological features of single unhealthy behaviors.5–8
Today, there is considerable evidence that unhealthy behaviors, especially, the so-called ‘big four’: tobacco smoking, alcohol abuse, unhealthy diet and physical inactivity, may simultaneously involve the same subject.9–14 Indeed, a remarkable feature of health-related behaviors is their tendency to occur in clusters,15 and such clusters are not randomly distributed at the population level.16 Although studies on single health behaviors are able to highlight the relationship between a variable of interest which, however, requires prior classification and other health behaviors, they do not enable us to evaluate the role of single behaviors in determining clusters.4
Cluster and factor analyses are two distinct, powerful, multivariate tools and should be used complementarily rather than separately. Indeed, factor analysis enables us to find common dimensions within which a variable may be located, while clustering enables group membership to be established.17 For these reasons, factor and cluster analyses are among the most widely used techniques in the behavioral sciences18 and have been amply applied in order to explore health behaviors during adolescence (reviewed in Refs19,20). For instance, in their comprehensive review on obesogenic cluster patterns among children and adolescents, Leech et al.20 concluded that diet, physical activity and a sedentary lifestyle cluster together in very complex ways, and that cluster membership depends on age, sex and socio-economic status. A systematic review by Wiefferink et al.19 explored the clustering of four behaviors namely, smoking, alcohol abuse, safe sex and healthy nutrition and revealed that smoking clusters with alcohol abuse (correlations of 0.43–0.60), while sexual experience clusters with both smoking and alcohol abuse (correlations of 0.29–0.54). The large Healthy Lifestyle in Europe by Nutrition in Adolescence (HELENA) cross-sectional study21,22 identified five clusters among adolescents in European cities. In these studies, clear age and gender differences in clustering patterns were observed. In comparison with older adolescents, younger ones are more physically active and have a healthier diet21; males display a pattern of high levels of moderate/vigorous physical activity and low-quality diets, while females display an opposite pattern.22
Although research conducted in various geographical settings, including Europe,21,22 has identified similar clusters of health behaviors among teenagers, it is not known whether the results of these studies can be transferred to the Italian context. Indeed, unhealthy behaviors are subject to a certain level of cross-cultural variation.23 We previously investigated24 factors associated with health-related behaviors among school-aged children in Tuscany. These data, however, may not be enough to design and implement well-targeted health-promotion interventions; indeed, some groups of teens may gain no, or less, benefit from ‘universal’ health-promotion campaigns, thus increasing inequalities. Taking these considerations into account, we aimed to examine the clustering of health-related behaviors among Italian teenagers and thereby to identify patterns of behavior clusters.
Methods
Source of data
The data used in the present study are from a WHO-promoted cross-national Health Behaviors in School-aged Children (HBSC) study carried out in Tuscany in 2010. The detailed methodology and survey tools used to collect these data are reported elsewhere.25 Briefly, data were gathered on a 4-year basis from adolescents aged 11–15 years (specifically, 11-, 13- and 15-year olds); the cluster sampling strategy was applied to obtain a random sample of school classes. The questionnaire used was self-administered (in the presence of trained personnel) in schools, in accordance with international standards.26 Of note, the questionnaire is subject to continuous development and validation by all participating countries. Parental approval of children's participation in the study was mandatory. The study protocol was approved by the Ethics Committee of the National Institute of Health; a parental opt-out consent form was used. All data were gathered anonymously, so that single participants could not be identified.
Health-related behaviors of interest
In this study, multiple health-related (both health-compromising and health-protective) behaviors were analyzed. Items on physical activity included both frequency (N of times for at least 60 min) and duration (N of hours). These items were accompanied by a brief student-friendly definition of physical activity. Dietary behavior was quantified by asking students how often they consumed fruit, vegetables, sweets and sweetened beverages. The recall period for physical activity and dietary patterns was set to 7 days. Sedentary behavior patterns were assessed by means of items on the frequency of prolonged (>2 consecutive hours) screen time (watching television/DVD, using the computer to chat, surf the web etc., playing video/computer games).
Questions on smoking included both past smoking experience and current smoking (frequency and number of cigarettes smoked in the past 30 days). Alcohol-related behavioral patterns were assessed both by type of alcoholic beverages (beer, wine, spirits/liquors, alcopops and other drinks) and by the frequency of their use.
Violent behavior was evaluated through two items on the frequency of participation in fights in the last 12 months and engaging in bullying in the last 2 months. The frequency of somatic and psychological symptoms (headache, stomach ache, backache, feeling sad, being irritable/in a bad mood/nervous, difficulty in falling asleep, dizziness) was also investigated (recall period of 6 months).
In sum, data on 22 health-related behaviors (5 protective and 17 health-compromising) were collected; the number of response options (levels) varied by item, as shown in Table 1.
Table 1.
Distribution (%) of health-related behaviors among Tuscan adolescents
Behavioral item | N | 0/we | <1/we | 1/we | 2–4/we | 5–6/we | 1/d | >1/d |
---|---|---|---|---|---|---|---|---|
Eating fruit, times/period | 3283 | 4.6 | 8.2 | 9.6 | 24.4 | 12.4 | 21.2 | 19.6 |
Eating vegetables, times/period | 3274 | 7.0 | 7.0 | 12.7 | 27.4 | 16.9 | 16.6 | 12.4 |
Eating sweets, times/period | 3259 | 3.3 | 13.1 | 17.7 | 25.6 | 14.3 | 13.3 | 12.7 |
Drinking sweetened beverages, times/period | 3208 | 44.7 | 20.9 | 13.5 | 10.7 | 4.4 | 3.3 | 2.5 |
0 | 1 | 2 | 3 | 4 | 5 | 6–7 | ||
Physical activity in the last we, d | 3257 | 6.0 | 9.4 | 19.7 | 19.4 | 17.4 | 12.3 | 15.7 |
0/mo | <1/mo | 1/mo | 1/we | 2–3/we | 4–6/we | 1/d | ||
Physical activity outside school, times/period | 3273 | 10.1 | 3.7 | 4.5 | 13.0 | 40.2 | 21.2 | 7.3 |
0 | 0.5 | 1 | 2–3 | 4–6 | ≥7 | |||
Physical activity outside school, h | 3275 | 13.3 | 10.0 | 20.2 | 29.5 | 16.4 | 10.7 | |
0 | 0.5 | 1 | 2 | 3 | 4 | 5–7 | ||
Use of pc, h/d | 3278 | 19.7 | 17.9 | 23.6 | 17.1 | 10.0 | 5.6 | 6.1 |
Watching TV, h/d | 3274 | 5.5 | 14.8 | 26.1 | 26.1 | 14.5 | 6.8 | 6.2 |
Video/pc gaming, h/d | 3269 | 25.3 | 21.4 | 24.0 | 16.2 | 6.2 | 3.1 | 3.7 |
No | Yes | |||||||
Smoking experience | 3271 | 78.0 | 22.0 | |||||
0/we | <1/we | ≥ 1/we <1/d | ≥1/d | |||||
Frequency of current smoking, times/period | 3278 | 89.0 | 3.1 | 3.0 | 4.8 | |||
0 | < 1/we | < 1/d | 1−5/d | 6–10/d | 11−20/d | > 20/d | ||
Number of cigarettes smoked in the last 30 d | 3267 | 88.3 | 3.6 | 1.6 | 3.3 | 1.7 | 0.9 | 0.5 |
Never | Rarely | Every mo | Every we | Every d | ||||
Drinking beer | 3234 | 67.3 | 22.1 | 3.1 | 6.6 | 0.9 | ||
Drinking wine | 3230 | 67.3 | 22.1 | 3.4 | 4.8 | 2.4 | ||
Drinking spirits/liquors | 3201 | 79.8 | 13.9 | 2.8 | 3.2 | 0.3 | ||
Drinking alcopops | 3209 | 65.4 | 22.6 | 5.7 | 5.3 | 1.0 | ||
Drinking other drinks | 3216 | 70.6 | 17.8 | 5.1 | 5.9 | 0.7 | ||
0 | 1 | 2 | 3 | ≥4 | ||||
Participation in a fight in the last 12 mo, times | 3272 | 67.1 | 16.7 | 6.2 | 3.3 | 6.7 | ||
0 | 1–2/mo | 2–3/mo | 1/we | >1/we | ||||
Being a bully in the last 2 mo, times/period | 3267 | 85.2 | 10.7 | 2.0 | 0.6 | 1.4 | ||
Never/rarely | 1/mo | 1/we | >1/we | >1/d | ||||
Somatic symptoms (≥1), times/period | 3279 | 12.3 | 22.8 | 20.2 | 25.6 | 19.1 | ||
Psychological symptoms, (≥1), times/period | 3275 | 8.5 | 20.4 | 24.8 | 29.6 | 16.7 |
d, day; h, hour; mo, month; we, week.
Statistical analysis
For descriptive purposes, all frequency data were reported as percentages. Factor and subsequent cluster analyses were performed as described by Yen et al.27 Since the original scales of measurement (response options) varied by item, the variables were first standardized. Factor analysis was made as follows. Preliminarily, sampling adequacy was verified by means of Bartlett's test of sphericity and overall Kaiser–Meyer–Olkin (KMO), while the presence of multicollinearity was checked by the determinant of correlation matrix. KMO must not be <0.5, while values of 0.8 and higher are considered optimal.28 To identify factors, determine their scores and uncover the latent structure, principal-component analysis with varimax orthogonal rotation was performed. The number of factors was determined by applying Kaiser's eigenvalue cutoff of 1. Factor loading plots were also examined. The factors identified were named according to their core components.
Cluster analysis was then performed in order to classify subjects according to their health-related behaviors. First, hierarchical clustering was analyzed by means of Ward's method, which uses squared Euclidean distance as a similarity measure. Subsequently, in order to maximize within-group similarities and find clusters, the κ-means algorithm for non-hierarchical clustering was computed to establish the final set of clusters. Scheffé’s test was used to make multiple comparisons.
Each cluster was characterized in terms of dimension, age and sex distributions; the between-cluster differences in the age and sex distributions were formally tested by means of one-way analysis of variance and χ2 test, respectively.
Statistical significance was set to two-sided α < 0.05. All analyses were carried out by means of SPSS statistical software, version 20.0 (SPSS Inc., Chicago, IL).
Results
Characteristics of study subjects
The socio-demographic characteristics of the participants have been reported elsewhere.24
The distributions of students’ health-related behaviors are reported in Table 1. Approximately, a quarter of adolescents ate fruit [22.4% (95% confidence interval, CI: 21.0–23.9%)] and vegetables [26.7% (95% CI: 25.2–28.2%)] only once a week or less, while 40.3% (95% CI: 38.6–42.0%) ate sweets at least 5 times a week. Similarly, most students did not exercise at all, or else much less than recommended (≥1 hour/day). With regard to screen-time entertainment, 21.7% (95% CI: 20.3–23.1%), 27.5% (95% CI: 26.0–29.0%) and 13.0% (95% CI: 11.9–14.2%) spent >2 hours per day using the computer, watching television and playing video games, respectively.
More than a fifth [22.0% (95% CI: 20.6–23.5%)] had experience of smoking, while 10.9% (95% CI: 9.9–12.0%) were defined as current smokers. About two-thirds of teens had never drunk beer [67.3 (95% CI: 65.7–68.9%)], wine [67.3 (95% CI:65.7–68.9%)] or alcopops [65.4 (95% CI: 63.8–7.0%)]. In the previous year, a third [32.9% (95% CI: 31.3–34.5%)] of adolescents had participated in a fight at least once. Engaging in bullying at least once in the previous 2 months was reported by 14.8% (95% CI: 13.6–16.1%) of students.
Somatic and psychological symptoms were also very prevalent: 87.7% (95% CI: 86.5–88.8%) and 91.5% (95% CI: 90.5–92.4%) of teens had at least one somatic or psychological symptom, respectively, at least once a month.
Factor analysis
As shown by the KMO coefficient of 0.80, the highly statistically significant Bartlett's test ( = 20 178.51, P < 0.001) and determinant of correlation matrix (0.001), the factor analysis was suitable. A total of 8 factors with eigenvalues >1 were identified and named according to their main traits. Specifically, Factors 1 (‘Alcohol drinking’) and 2 (‘Smoking’) included items on the frequency of consumption of various alcoholic beverages, previous smoking experience, frequency of current smoking and number of cigarettes smoked. These two factors explained 21.1 and 11.0% of variance, respectively. Factor 3 had loadings on three items on physical activity, including both frequency and duration (R2 = 0.08), while Factor 4 (‘Screen time’) had loadings on frequency items forming sedentary behavior patterns, and displayed an explained variance of 5.9%. The remaining four factors all included two loadings [Factor 5 ‘Signs & symptoms’ (frequency of somatic and psychological symptoms), Factor 6 ‘Healthy eating’ (frequency of consumption of fruit and vegetables), Factor 7 ‘Violence’ (frequency of participation in fights and active bullying) and Factor 8 ‘Sweet tooth’ (frequency of consumption of sweets and soft drinks)]. Cumulatively, the eight factors explained 67% of variance. Each of the 22 behaviors displayed some degree of common variance with other behaviors, since all communalities exceeded 0.5 (Table 2).
Table 2.
Factor analysis of 22 health-related behaviors among Tuscan adolescents
Behavior | Factor names | Factors | Communalities | |||||||
---|---|---|---|---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | |||
Frequency of drinking beer | Alcohol drinking | 0.757 | 0.208 | −0.020 | −0.035 | −0.004 | 0.023 | −0.113 | 0.044 | 0.63 |
Frequency of drinking of wine | 0.761 | −0.062 | −0.026 | −0.027 | −0.036 | −0.033 | −0.148 | 0.047 | 0.61 | |
Frequency of drinking spirits/liquors | 0.677 | 0.334 | −0.012 | −0.065 | 0.059 | 0.019 | −0.034 | −0.060 | 0.58 | |
Frequency of drinking alcopops | 0.697 | 0.169 | −0.035 | −0.108 | 0.097 | 0.014 | −0.035 | −0.127 | 0.55 | |
Frequency of drinking other alcoholic drinks | 0.728 | 0.371 | −0.009 | −0.088 | 0.078 | 0.040 | 0.003 | −0.071 | 0.69 | |
Smoking experience | Smoking | 0.378 | 0.674 | 0.008 | −0.114 | 0.099 | 0.046 | −0.003 | −0.014 | 0.62 |
Current smoking frequency | 0.233 | 0.914 | 0.056 | −0.020 | 0.036 | 0.024 | −0.091 | −0.015 | 0.90 | |
Cigarettes smoked in the last 30 days | −0.239 | −0.903 | −0.054 | 0.044 | −0.029 | −0.046 | 0.121 | 0.026 | 0.89 | |
Frequency of physical activity in the last week | Physical activity | −0.027 | 0.048 | 0.800 | −0.016 | 0.015 | 0.044 | 0.034 | 0.011 | 0.65 |
Frequency of physical activity outside school | −0.006 | −0.039 | −0.888 | 0.056 | −0.039 | −0.075 | −0.066 | −0.028 | 0.80 | |
Hours of physical activity outside school | −0.058 | 0.002 | 0.865 | −0.012 | 0.075 | 0.037 | −0.004 | 0.002 | 0.76 | |
Daily hours of TV watching | Screen time | −0.028 | 0.004 | −0.057 | 0.704 | 0.002 | −0.006 | 0.073 | 0.125 | 0.52 |
Daily hours of playing video games | −0.059 | −0.001 | −0.006 | 0.788 | 0.022 | −0.116 | 0.193 | 0.023 | 0.68 | |
Daily hours of pc use | −0.169 | −0.170 | −0.014 | 0.723 | −0.156 | −0.038 | −0.082 | 0.016 | 0.61 | |
Frequency of somatic symptoms | Signs & symptoms | 0.052 | 0.061 | 0.012 | −0.014 | 0.834 | 0.042 | −0.018 | −0.018 | 0.70 |
Frequency of psychological symptoms | 0.062 | 0.053 | 0.110 | −0.086 | 0.813 | 0.015 | −0.118 | −0.022 | 0.70 | |
Frequency of fruit consumption | Healthy eating | 0.040 | 0.064 | 0.087 | −0.064 | 0.046 | 0.815 | −0.033 | 0.056 | 0.69 |
Frequency of vegetable consumption | −0.009 | 0.018 | 0.057 | −0.067 | 0.010 | 0.811 | −0.056 | −0.119 | 0.68 | |
Frequency of participation in fights | Violence | −0.132 | 0.002 | 0.069 | 0.081 | −0.131 | −0.087 | 0.744 | 0.050 | 0.61 |
Frequency of being a bully | −0.108 | −0.166 | 0.021 | 0.089 | −0.006 | −0.004 | 0.750 | 0.033 | 0.61 | |
Frequency of sweet consumption | Sweet tooth | −0.043 | −0.025 | 0.034 | 0.131 | −0.068 | −0.229 | −0.014 | 0.685 | 0.55 |
Frequency of soft drink consumption | −0.050 | −0.020 | 0.005 | 0.037 | 0.025 | 0.138 | 0.095 | 0.804 | 0.68 | |
Eigenvalue | — | 4.66 | 2.41 | 1.83 | 1.30 | 1.28 | 1.16 | 1.06 | 1.02 | — |
R2 | — | 0.212 | 0.110 | 0.083 | 0.059 | 0.058 | 0.053 | 0.048 | 0.046 | — |
Cumulative R2 | — | 0.212 | 0.322 | 0.405 | 0.464 | 0.522 | 0.575 | 0.623 | 0.669 | — |
Loadings > |0.5| are evidenced in bold.
Cluster analysis
The optimal number of clusters was identified as follows. An increase in agglomerative coefficients (Δ of 17.8%) was observed when the eight clusters were reduced to seven. Further reduction to six groups produced an even higher increase (Δ of 19.7%) in agglomerative coefficients. The six-cluster κ-means solution was therefore established; the discriminant analysis confirmed a high (93.8%) level of validity of the classification. As shown by Sheffé’s test statistics, there were several significant comparisons among the eight behavioral factors; only scores for the ‘Sweet tooth’ factor were not significantly related to the other clusters.
As reported in Table 3, Cluster 1 had the highest positive score (2.26) for the factor ‘Violence’ and was named ‘Violent group’. Cluster 2 ‘Non-drinking smokers group’ showed a high z-score for the factor ‘Smoking’ and a very low score for ‘Alcohol drinking’, while Cluster 3 was dubbed ‘Smoking drinkers group’ owing to the high scores for both factors. Cluster 4 was judged to be the healthiest, as its scores on health-compromising factors were either negative or close to zero, while those on both protective behaviors were negative; it was therefore named ‘Quasi-healthy group’. Clusters 5 and 6 displayed their highest z-scores on the factors ‘Signs & symptoms’ (0.97) and ‘Screen time’ (1.88) and were therefore called ‘Symptomatic’ and ‘Screen passion’ groups, respectively.
Table 3.
Comparison of z factor scores for the clusters identified
Behavioral factors | Cluster | F | Sheffé’s test | |||||
---|---|---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | 5 | 6 | |||
Alcohol drinking | 0.20 | −2.00 | 0.37 | −0.57 | 0.34 | 0.22 | 699.03* | 3 > (1,2,4,6) |
Smoking | 0.22 | 0.61 | 0.20 | −2.92 | 0.17 | 0.25 | 1635.79* | 2 > (1,3,4,5,6) |
Physical activity | 0.14 | 0.16 | 0.01 | −0.22 | 0.01 | −0.16 | 6.50* | (1,2) > (4,6) |
Screen time | −0.08 | −0.08 | −0.41 | 0.03 | −0.22 | 1.88 | 482.71* | 6 > (1,2,3,4,5) |
Signs & symptoms | −0.22 | −0.09 | −0.79 | −0.09 | 0.97 | −0.29 | 571.72* | 5 > (1,2,3,4,6) |
Healthy eating | −0.14 | −0.02 | 0.09 | −0.13 | 0.02 | −0.06 | 3.66* | 3 > 1 |
Violence | 2.26 | −0.23 | −0.27 | −0.03 | −0.25 | −0.27 | 685.23* | 1 > (2,3,4,5,6) |
Sweet tooth | 0.07 | 00.05 | −0.03 | 00.05 | −0.01 | −0.06 | 0.919** | — |
*P < 0.001; **P > 0.05.
There was a 4-fold difference in the dimensions of clusters: ‘Symptomatic’ and ‘Smoking drinker’ groups (31.3 and 29.7%, respectively) were the largest, while ‘Quasi-healthy’ and ‘Violent’ groups (7.8 and 9.5%, respectively) were substantially less copious (Table 4). Male students were more prevalent in ‘Violent’, ‘Non-drinking smoker’ and ‘Symptomatic’ groups. The ‘Quasi-healthy’ group comprised older students (mean age 15.2 years), while the ‘Violent’ cluster was composed of younger adolescents (mean age 12.8 years). The between-group differences in both mean age and gender distribution were highly statistically significant (P < 0.001) (Table 4).
Table 4.
Dimensions, age and sex distributions of the six clusters
Cluster | Parameter | ||
---|---|---|---|
Dimensions, % (95% CI) | % boys (95% CI) | Mean age (SD), years | |
1: Violent group | 9.5 (8.5–10.6) | 75.7 (70.3–80.4) | 12.8 (1.5) |
2: Non-drinking smoker group | 10.8 (9.7–12.0) | 68.1 (62.7–73.0) | 14.4 (1.4) |
3: Smoking drinker group | 29.7 (28.1–31.4) | 35.0 (31.9–38.3) | 13.0 (1.6) |
4: Quasi-healthy group | 7.8 (6.8–8.8) | 40.7 (34.5–47.2) | 15.2 (0.9) |
5: Symptomatic group | 31.3 (29.6–33.0) | 56.6 (53.3–59.8) | 12.9 (1.5) |
6: Screen passion group | 10.9 (9.8–12.1) | 46.4 (41.0–51.9) | 13.7 (1.5) |
Discussion
Main finding of this study
We found that early and mid-adolescents in Tuscany formed six clusters: namely, ‘Violent’, ‘Non-drinking smoker’, ‘Smoking drinker’, ‘Quasi-healthy’, ‘Symptomatic’ and ‘Screen passion’ groups; each of these showed a unique behavioral pattern. It is alarming that most Clusters (4 of 6) had higher than average scores of both ‘Smoking’ and ‘Alcohol drinking’ behaviors; this finding confirms a previously reported observation19 that these two modifiable risk factors are among the commonest among adolescents and usually cluster together. This result is also of a great practical significance, since health-promotion interventions targeting both risk behaviors simultaneously would be beneficial to most students.
What is already known on this topic
Two clusters of alcohol consumers were established; these had very similar scores on all behaviors, except for diametrically opposite scores on the ‘Signs & Symptoms’ factor. Italy, being situated in the south of Europe, is a typical ‘wine-drinking’ country30 and wine consumption is fairly common, especially during main meals.31 Indeed, the European Comparative Alcohol Study (ECAS) revealed that Italy was the top country in Europe in terms of daily drinking (as opposed to binge drinking, which is more common in northern and central Europe)30; moreover, an increasing proportion of female and young daily drinkers in Italy has been also established.32 Indeed, as the two above-mentioned clusters comprised 53.9% of girls, this finding confirms the increasing trend in alcohol consumption among female teens. On the other hand, it is interesting that the proportion of ‘wine ever’ drinkers (those who stated having tried wine at least once in their lives) was almost equal than that of ‘beer and alcopops ever’ drinkers. This observation is consistent with previously described33 patterns on the internationalization of drinking habits (i.e. with a shift from the more traditional wine to alcopops or beer). Since teens may regard low-alcohol drinks as a ‘healthier’ alternative, tailor-made interventions should also highlight the potential dangers of such drinks. Taking into consideration that an average European adolescent begins to drink alcoholic beverages at the age of 12.5 years and gets drunk for the first time at 14 years,30 it is essential to deliver age-tailored preventive interventions sufficiently early. An appropriate period could be around 11 years of age, which roughly corresponds to the transition from primary school to the first phase of secondary education. Indeed, it has already been shown6 that this is the period when the risk of starting to smoke begins to increase.
As expected, the ‘Violent’ cluster was mostly composed of boys, although it also displayed a significant presence of girls (about a quarter); this means that careful attention should be paid to gender roles in the school environment. Moreover, this cluster had above-average scores on both ‘Smoking’ and ‘Alcohol drinking’. A highly significant and consistent association between bullying and smoking/drinking has also been documented among Italian adolescents.34 Since several health-compromising risk factors are difficult to identify in the school context, the violent behavior of students toward their peers could be regarded as a marker of other unhealthy behaviors, such as smoking or frequent alcohol consumption.34
About 11% of students fell into the ‘Screen passion’ group; inclusion in this cluster was also associated with high scores on ‘Alcohol drinking’ and ‘Smoking’. A similar pattern has been observed in previous research35–37 and may be explained by the fact that prolonged screen time increases exposure to media content that promotes such behaviors, thus encouraging alcohol and cigarette consumption. This may indicate that parental supervision of teens’ screen time, the activation of parental control options on TV, computer and mobile devices, and the engagement of children in open-air activities could be beneficial both in reducing smoking and alcohol consumption and in increasing levels of physical activity.
Our study showed that the healthiest cluster was the least numerous (7.8%). This group had a remarkable behavioral pattern; although students had very low scores on ‘Alcohol drinking’ and ‘Smoking’ behaviors, they also scored low on two health-enhancing behaviors, namely ‘Physical activity’ and ‘Healthy eating’. Clustering patterns of nutrition-related behaviors (such as consumption of raw fruits and vegetables) and alcohol drinking and smoking among Finnish and Swiss adolescents were investigated in a study by Karvonen et al.38 These authors discerned 3 clusters in both countries: healthy (healthy eating, no smoking, no drinking), unhealthy (unhealthy eating, smoking, drinking) and mixed (unhealthy eating, no smoking, no drinking). Notably, in Switzerland, inclusion in Cluster 3 was more typical of Italian-speaking teens than of German- and French-speaking ones. The third ‘mixed cluster’ almost corresponds to our ‘Quasi-healthy’ group; thus, our results confirm that this cluster is cross-culturally valid. Moreover, this finding also supports the conclusion of Wiefferink et al.19 that teenagers have a complex lifestyle paradigm, in that a substantial proportion of them have a somewhat mixed lifestyle that combines healthy and unhealthy behaviors.
What this study adds
Health-compromising behaviors are common among Tuscan teens, and their combination is thus highly plausible. In turn, these multiple health-related behaviors contribute to determining the lifestyle of adolescents that will reflect their health in the future. To our knowledge, the present study is among the first conducted in Italy to investigate clustering patterns of multiple health-related behaviors in a regionally representative sample of teenagers. Indeed, no studies carried out in Italy were identified in a scoping review of this topic.29 A down-to-earth appraisal of the clustering of health behaviors would enable integrated, multidimensional health-promotion strategies to be designed and implemented. The results of the present research may serve as a basis for more tailored school-based preventive interventions in Tuscany and, probably, at the supra-regional level.
Limitations of this study
Despite its large sample size, regional representativeness and analysis of various health-related behaviors, our study has some limitations. First of all, owing to its cross-sectional nature, it could not establish any temporal or cause-and-effect relationships; future longitudinal studies will help to highlight temporal trends in the clustering of health-related behaviors among adolescents as well as changes in the composition of the clusters over time. Second, the study was prone to the social desirability bias, in that some students might have under-reported their health-compromising behaviors. However, we believe that the impact of the social desirability bias was mitigated by the fact that research staff, rather than teachers, played a ‘dominant’ role in survey administration, since the close involvement of teachers may diminish students’ perceived anonymity.39 Third, Tuscany is a large Italian region situated in central Italy; owing to the well-known North–Center–South gradient among Italian regions in several parameters, including health-related ones40 (central regions being a transition area), it is unknown whether our results can be generalized to non-central Italian regions.
Conclusions
Early and mid-teenagers in Tuscany form distinct clusters and most of them engage simultaneously in >1 unhealthy behavior. This fact will undoubtedly contribute to compromising the health of the next generation in Italy. This means that parents, teachers, pediatricians, public health practitioners and other relevant stakeholders should urgently plan and implement health-promotion interventions that are not only multidisciplinary but also, considering the observed clustering pattern, equity-based. Given that traditional methods of health promotion often yield only modest benefits, and that today's adolescents readily take up novel technologies,41 eHealth- and mHealth-based interventions should be viewed as valuables complementary means. Indeed, exploiting appropriate media (such as social networks) to address the present generation of adolescents (the first to have ‘grown up online’)41 in their own language could improve their health in the future and ‘de-cluster’ the observed trend.
Acknowledgements
The authors wish to thank the Authorities of the Regional Health System, the Local Health Authorities and the Schools. The authors thank Dr Bernard Patrick for his linguistic review of the manuscript.
Funding
This project was supported by grants from the Region of Tuscany (Resolution No. 800 of 13/10/2008 Gaining Health in Tuscany: facilitating healthy choices).
Authors’ contributions
GL designed and coordinated the study; RS collected data, made quality control checks and collaborated in writing the manuscript; AP performed the statistical analyses and evaluated the results; GL, DP, AD, LA, MVG, DA and RG discussed and interpreted the results and wrote the manuscript. All authors read and approved the final manuscript.
References
- 1. Kipping RR, Campbell RM, MacArthur GJ et al. Multiple risk behaviour in adolescence. J Public Health (Oxf) 2012;34(S1):1–2. [DOI] [PubMed] [Google Scholar]
- 2. Kann L, Kinchen S, Shanklin SL et al. Youth risk behavior surveillance-United States, 2013. MMWR Surveill Summ 2014;63(S4):1–168. [PubMed] [Google Scholar]
- 3. Health Behaviors in School-aged Children (HBSC) study , Italy 2010. http://www.salute.gov.it/imgs/c_17_newsaree_1286_listafile_itemname_15_file.pdf (1 October 2015, date last accessed).
- 4. Burke V, Milligan RA, Beilin LJ et al. Clustering of health-related behaviors among 18-year-old Australians. Prev Med 1997;26(5Pt1):724–33. [DOI] [PubMed] [Google Scholar]
- 5. Flotta D, Micò R, Nobile CG et al. Consumption of energy drinks, alcohol, and alcohol-mixed energy drinks among Italian adolescents. Alcohol Clin Exp Res 2014;38:1654–61. [DOI] [PubMed] [Google Scholar]
- 6. Panatto D, Amicizia D, Domnich A et al. Tobacco smoking among students in an urban area in Northern Italy. J Prev Med Hyg 2013;54(2):97–103. [PMC free article] [PubMed] [Google Scholar]
- 7. Piovani D, Clavenna A, Bonati M, Interregional Italian Drug Utilisation Group . Drug use profile in outpatient children and adolescents in different Italian regions. BMC Pediatr 2013;4:13–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Graziano F, Bina M, Giannotta F et al. Drinking motives and alcoholic beverage preferences among Italian adolescents. J Adolesc 2012;35:823–31. [DOI] [PubMed] [Google Scholar]
- 9. Dumith SC, Muniz LC, Tassitano RM et al. Clustering of risk factors for chronic diseases among adolescents from Southern Brazil. Prev Med 2012;54:393–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Poortinga W. The prevalence and clustering of four major lifestyle risk factors in an English adult population. Prev Med 2007;44:124–8. [DOI] [PubMed] [Google Scholar]
- 11. Chiolero A, Wietlisbach V, Ruffieux C et al. Clustering of risk behaviors with cigarette consumption: a population-based survey. Prev Med 2006;42:348–53. [DOI] [PubMed] [Google Scholar]
- 12. Fine LJ, Philogene GS, Gramling R et al. Prevalence of multiple chronic disease risk factors. 2001 National Health Interview Survey. Am J Prev Med 2004;27:18–24. [DOI] [PubMed] [Google Scholar]
- 13. Laaksonen M, Prattala R, Karisto A. Patterns of unhealthy behaviour in Finland. Eur J Public Health 2001;11:294–300. [DOI] [PubMed] [Google Scholar]
- 14. Schuit AJ, Loon AJ, Tijhuis M et al. Clustering of lifestyle risk factors in a general adult population. Prev Med 2002;35:219–24. [DOI] [PubMed] [Google Scholar]
- 15. Spring B, Moller AC, Coons MJ. Multiple health behaviours: overview and implications. J Public Health (Oxf) 2012;34(S1):i3–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Conry MC, Morgan K, Curry P et al. The clustering of health behaviours in Ireland and their relationship with mental health, self-rated health and quality of life. BMC Public Health 2011;6:11–692. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Gorman BS, Primavera LH. The complementary use of cluster and factor analysis methods. J Exp Educ 1983;54:165–8. [Google Scholar]
- 18. Newby PK, Tucker KL. Empirically derived eating patterns using factor or cluster analysis: a review. Nutr Rev 2004;62:177–203. [DOI] [PubMed] [Google Scholar]
- 19. Wiefferink CH, Peters L, Hoekstra F et al. Clustering of health-related behaviors and their determinants: possible consequences for school health interventions. Prev Sci 2006;7:127–49. [DOI] [PubMed] [Google Scholar]
- 20. Leech RM, McNaughton SA, Timperio A. The clustering of diet, physical activity and sedentary behavior in children and adolescents: a review. Int J Behav Nutr Phys Act 2014;11:4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Cuenca-García M, Huybrechts I, Ruiz JR et al. Clustering of multiple lifestyle behaviors and health-related fitness in European adolescents. J Nutr Educ Behav 2013;45:549–57. [DOI] [PubMed] [Google Scholar]
- 22. Ottevaere C, Huybrechts I, Benser J et al. Clustering patterns of physical activity, sedentary and dietary behavior among European adolescents: the HELENA study. BMC Public Health 2011;17:11–328. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Al-Hazzaa HM, Al-Nakeeb Y, Duncan MJ et al. A cross-cultural comparison of health behaviors between Saudi and British adolescents living in urban areas: gender by country analyses. Int J Environ Res Public Health 2013;10:6701–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Lazzeri G, Azzolini E, Pammolli A et al. Factors associated with unhealthy behaviours and health outcomes: a cross-sectional study among Tuscan adolescents (Italy). Int J Equity Health 2014;13:83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Lazzeri G, Giacchi MV, Dalmasso P et al. The methodology of the Italian HBSC 2010 study (Health Behavior in School-aged Children). Ann Ig 2013;25(3):225–33. [DOI] [PubMed] [Google Scholar]
- 26. Roberts C, Freeman J, Samdal O et al. The Health Behavior in School-aged Children (HBSC) study: methodological developments and current tensions. Int J Public Health 2009;54S2:140–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Yen LL, Chiu CJ, Wu WC et al. Aggregation of health behaviors among fourth graders in northern Taiwan. J Adolesc Health 2006;39:435–42. [DOI] [PubMed] [Google Scholar]
- 28. Cerny BA, Kaiser HF. A study of a measure of sampling adequacy for factor-analytic correlation matrices. Multivar Behav Res 1977;12:43–7. [DOI] [PubMed] [Google Scholar]
- 29. McAloney K, Graham H, Law C et al. A scoping review of statistical approaches to the analysis of multiple health-related behaviours. Prev Med 2013;56:365–71. [DOI] [PubMed] [Google Scholar]
- 30.European Commission The Use of Alcohol in Europe. http://ec.europa.eu/health-eu/doc/alcoholineu_chap4_en.pdf http://ec.europa.eu/health-eu/doc/alcoholineu_chap4_en.pdf
- 31. Ferraroni M, Decarli A, Franceschi S et al. Validity and reproducibility of alcohol consumption in Italy. Int J Epidemiol 1996;25:775–82. [DOI] [PubMed] [Google Scholar]
- 32. Simpura J, Karlsson T. Trends in Drinking Patterns in Fifteen European Countries, 1950–2000: A Collection of Country Report. Helsinki: Stakes, 2001. [Google Scholar]
- 33. Pyörälä E. Trends in alcohol consumption in Spain, Portugal, France and Italy from the 1950s until the 1980s. Br J Addict 1990;85:469–77. [DOI] [PubMed] [Google Scholar]
- 34. Vieno A, Gini G, Santinello M. Different forms of bullying and their association to smoking and drinking behavior in Italian adolescents. J Sch Health 2011;81:393–9. [DOI] [PubMed] [Google Scholar]
- 35. Pedrosa AA, Camacho LA, Passos SR et al. Alcohol consumption by university students. Cad Saude Publica 2011;27:1611–21. [DOI] [PubMed] [Google Scholar]
- 36. Saffer H. Alcohol advertising and youth. J Stud Alcohol Suppl 2002;14:173–81. [DOI] [PubMed] [Google Scholar]
- 37. Dias PJ, Domingos IP, Ferreira MG et al. Prevalence and factors associated with sedentary behavior in adolescents. Rev Saude Publica 2014;48:266–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Karvonen S, Abel T, Calmonte R et al. Patterns of health-related behaviour and their cross-cultural validity—a comparative study on two populations of young people. Soz Praventivmed 2000;45:35–45. [DOI] [PubMed] [Google Scholar]
- 39. Bjarnason T. Administration mode bias in a school survey on alcohol, tobacco and illicit drug use. Addiction 1995;90:555–9. [DOI] [PubMed] [Google Scholar]
- 40. de Waure C, Bruno S, Furia G et al. Health inequalities: an analysis of hospitalizations with respect to migrant status, gender and geographical area. BMC Int Health Hum Rights 2015;15:2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Amicizia D, Domnich A, Gasparini R et al. An overview of current and potential use of information and communication technologies for immunization promotion among adolescents. Hum Vaccin Immunother 2013;9:2634–42. [DOI] [PMC free article] [PubMed] [Google Scholar]