Cases in the deCODE sample met criteria for lifetime DSM-III-R or DSM-IV cannabis abuse or dependence or DSM-5 cannabis use disorder according to diagnoses made at the National Center of Addiction Medicine in Iceland, whereas controls were derived from the general population of Iceland . Exposure data were not available for some large groups ; therefore, controls were defined regardless of lifetime cannabis exposure across all datasets.Adverse mental health and social outcomes of cannabis use have been reported for individuals and societies . Cannabis use was associated with the use of other drugs , difficulties to reach life goals , adverse educational outcomes in adolescents , legal issues, and traffic accidents . On the one hand, there are genetic and neuro developmental risk factors ; on the other hand, there are potentially modifiable environmental risk factors of cannabis use . Parenting styles , substance use of parents and peers, academic and school related factors , and risk perception have been described as relevant psychosocial risk factors . In 2016, a global estimate by UNODC based on data from 130 countries estimated that 5.6% of the population aged 15–16 years had used cannabis at least once in the past year . In different regions of the world, the sale of cannabis has been legalized, leading to renewed interest in how this may affect cannabis use and associated factors . In Chile, a law is currently under discussion in Congress aiming to legalize home cannabis cultivation for personal recreational and/or medical use. In the past decade, there has been a public debate about legalization and important changes in legal practice to decriminalize cannabis cultivation. An increase of the prevalence of adolescent cannabis use in Chile was reported for the year 2013 compared to the years 2001–2011 . The prevalence of adolescent cannabis use in Chile was reported to be the highest in the Americas /InterAmerican Drug Abuse Control Commission , 2019. In line with the normalization theory , adolescents in countries with high prevalence of substance use are less likely to report risk factors than in countries with low prevalence.Therefore, risk and protective factors may have changed in prevalence and/or strength of association constituting new challenges for targeting prevention.
Risk and protective factors of substance use had been reported to be consistent between 1976 and 1997 in the US : several variables such as religiosity, political beliefs, truancy, and frequent evenings out were consistently linked to substance use over time among high school students. However, in the current context of marked changes of prevalence in adolescent cannabis use in Chile, the assessment of prevalence factors and their effect size over time may allow reaching a better understanding of the factors underlying the process in which the substance use is changing over time and then contributing to adjust prevention strategies and exploring if the factors associated to substance use vary across years. Factors associated with substance use in adolescents had been reported from Chile for one single year , but how these factors vary over time and in their strength of association with the prevalence of cannabis grow set up use has not been previously addressed. In Argentina, Chile and Uruguay, an increase of cannabis use in adolescents in recent years has been reported, and the association between risk perception and use has decreased. Meanwhile, perceived availability remained strongly associated with cannabis use, but other potential risk factors have not been investigated . The quantification of potential risk factors and their trends over time may allow targeting prevention strategies . The aim of the present research was to identify prevalence trends and associated factors of cannabis use in the past years among adolescents, and to assess trends of associated factors and the strength of association over time.Study participants were adolescent high-school students. The Chilean National Service of Drugs and Alcohol Use Prevention and Rehabilitation carries out the nationwide school-based survey in students from 8th to 12th grade every two years, with a probabilistic, representative at regional and nationwide level, stratified , multistage sampling design in clusters . The rate of reached sample was around 80% of the theoretical sample size. The detailed methodology is presented by SENDA in each survey report available online with stability across the years from 2003 to 2017 and minimal variations. We obtained data from SENDA for the years 2003 to 2017 , 2018. SENDA offers the option of a self-administered questionnaire and a face-to-face interview. In the self-administered version, the students are supervised by a surveyor.
Once the schools and classes were defined, random samples of 20 students were selected from each classroom to participate in the survey /Inter-American Drug Abuse Control Commission , 2019.The survey questions included socio-demographic data, several types of substance use , tobacco and alcohol, perceived risk of substance use, satisfaction with school, school attendance, grades, relations with peers, teachers, and parents and extracurricular activities among others. We selected items that were consistently present across the years, relevant in practice and representing different areas of risk. The following variables were included in the analyses: 1) Cannabis use in the past year ; 2) Funding of the school: public vs. private or mixed ; 3) gender; 4) age; 5) Use of alcohol in the past month ; 6) Use of tobacco in the past month ; 7) Age at first use of alcohol; 8) Age at first use of tobacco; 9) Unexcused absence from school in the past year ; 10) School performance based on self-report was dichotomized as low vs. high; 11) Sport activities, as the number of days per week doing sports as extracurricular activity; 12) Educational level of parents with three alternative categories: uncomplete secondary level, complete secondary level and complete higher education; 13) Marital status of the parents; 14) Parental acquaintance with friends was assessed with the question, “In general, would you say that your parents know your closest friends very well, fairly well or little?” ; 15) Parental rejection of alcohol use ; 16) Parental rejection of cannabis use ; 17) Having friends who regularly use alcohol ; 18) Having friends who regularly use cannabis ; 19) Perceived risk of cannabis use .Descriptive statistics were calculated for each year of the surveys, and for the variables: gender, age, school funding, cannabis use prevalence, alcohol and tobacco use prevalence; 95% confidence intervals were calculated for prevalence rates. Mixed effects logistic regressions were performed for data at individual level, with data nested at the school level , and nested at the level of funding source of the schools. The multilevel logistic binomial regressions were conducted with cannabis use in the past year as dependent variable for each year separately. Adjusted odds ratios were calculated for each variable. The variables at the individual level were: Gender; age; age at first alcohol use; age at first tobacco use; alcohol use in the past month; tobacco use in the past month; perceived cannabis use risk; school performance; truancy; days of sport activities in a week; friends regularly using alcohol; friends regularly using cannabis; educational level of father and mother; parents’ marital status; parental acquaintance with friends; parental alcohol use rejection; parental cannabis use rejection.
Intraclass correlation coefficients were calculated from a null model for both the school level clusters and school funding level clusters. The command glmer of the lme4 package was used in R software to estimate the mixed effects logistic regressions. Variables with odds ratios on average higher than 1.5 across the entire time series were retained for further analyses. This threshold was introduced due to the large size of the data set and to avoid retaining significant odds ratios close to 1.0 that may be clinically irrelevant and irrelevant for prevention planning. Odds ratios smaller than 1.5 can be considered as small effect size and larger than 1.5 as moderate or large effect size . Adjusted odds ratios were calculated for the retained variables for each year of data collection to assess changes of the association over time. Also, interactions between year and each one of the retained variables were analyzed by multilevel mixed effects logistic regressions for all pooled data to assess how the associations between variables and outcomes were affected by time in each survey cycle, using the first year of the series as reference. The prevalence of the retained variables was described as trends over time. Trends of prevalence data and odds ratios over time were plotted for the retained variables and each trend was tested for its fit to linear or higher models, and the F-statistic, degrees of freedom , R-squared and p values were reported. As quality control, before the analyses, data points of participants who answered in at least two occasions in an inconsistent way for each substance were eliminated , for instance, inconsistent answers about date of last use of cannabis, lifetime use and/or use in the past month.The interaction of each one of the retained prevalence factors and year of data collection was calculated, with the year 2003 as reference. specific differences over time in the association of each factor with cannabis use were observed. For the use of alcohol in the past month, we observed a significant negative interaction from 2007 to 2017 showing a decrease of the association with cannabis use over time. For the use of tobacco in the past month, a similar pattern of negative interactions was observed between 2007 and 2015. For the factor friends who regularly use cannabis, we observed negative interactions from 2007 to 2017. For truancy, negative interactions were seen from 2009 to 2017. For low outdoor cannabis grow risk perception, the interaction for 2009 was negative, but thereafter positive. In contrast, low parental cannabis rejection was the only factor that showed positive interactions from 2007 over time until 2017 . Table 3 shows the interactions observed between years and prevalence factors.Our research showed that cannabis use among adolescents increased substantially from 2003 to 2017.
We identified the factors most strongly associated with adolescent cannabis use and present prevalence estimates over time for those prevalence factors. Furthermore, we inform the strength of association over time for each of the most important factors. Although having friends who regularly use cannabis decreased in the strength of association with cannabis use, the variable continued to have the strongest effect size. An important increase in the magnitude of association with cannabis use was seen for low parental rejection of cannabis use. Interaction analyses for each year with each of the factors associated with cannabis use, showed trends since 2007 with a decrease of the association between cannabis use and the factors alcohol use in the past month, tobacco use in the past month, cannabis use in friends, and since 2009 for truancy. However, we observed an increase in the association between cannabis use and low parental cannabis rejection since 2007. Overall, the most important prevalence factors show significant changes in the strength of association since 2007 compared to the reference year 2003. Interestingly, this precedes the major increase in the prevalence of cannabis use observed between 2011 and 2013.This study comprised nationwide survey data of more than a decade with large sample sizes. We show for the first time trends for the prevalence and the strength of association with cannabis use of possible risk factors in a Latin American country. The study also has limitations: even though the surveys were presented in a consistent way over the years, the data were based on self-reporting. Repeated cross-sectional data do not allow establishing causal links between the increase of cannabis use and the associated variables. The variables assessed in this research were mainly on the individual level and limited to the items continuously included in the national surveys over the years.In the US, the prevalence of cannabis use among adolescents in the past year increased between 1991 and 2015, while the prevalence of alcohol use decreased, and the prevalence of any other illicit substance use also decreased . In Europe from 2000 to 2015, the prevalence of cannabis use in the past month among adolescents showed heterogeneous trends in different regions: decrease in Northern Europe in linear trends, increase in Southern Europe in linear trends, decrease in Eastern Europe in a concave trend and increase in the Balkans in a convex trend.