NEW PSYCHOACTIVE SUBSTANCE USE IN YOUNG POLY SUBSTANCE

NEW PSYCHOACTIVE SUBSTANCE USE IN YOUNG POLY SUBSTANCE USING EUROPEANS Harry Sumnall

ACKNOWLEDGEMENTS CO-AUTHORS • Liefman H – SW; Atkinson A – UK; Henriksson C – SW; Kraus L – DE/SW; Franelić IP – HR; Balakireva O – UA; Rupsiene L – LT; Sturua L - GE; Strizek J – AT; Steriu A – GBM; Clancy L – IE; Nociar A – SK; Florescu S - RO • We would like to acknowledge the members of the ESPAD group who collected the national data (http: //www. espad. org/report/acknowledgements ) and the funding bodies who supported the international coordination of ESPAD: the government of Sweden, the EMCDDA and the Pompidou Group. Special thanks are due to the schoolchildren, teachers and national funding bodies who made this project possible • The production of this paper was supported by EMCDDA contract CT. 15. EPI. 0080. 1. 0. • The ESPAD Group (2016). ESPAD Report 2015. Results from the European School Survey Project on Alcohol and Other Drugs. Luxembourg: Publications Office of the European Union

BACKGROUND • International interest in rise of NPS, health and social harms, and new policy approaches • Limited European data in young people; little comparative data between Member States (cf USA) • Existing data estimates NPS prevalence is relatively low, but use is part of broader risk profile, including polysubstance use • Developing and targeting prevention and education programmes requires better understanding of potential recipient groups • NPS-specific programmes • Multiple risk and vulnerabilities • Insights into policy impact e. g. Hanneman et al. , 2017; Palamar, 2015; Sumnall et al.

NPS USE IN ESPAD 2015 The ESPAD Group (2016)

The ESPAD Group (2016)

MAIN RESEARCH QUESTIONS • What is the relationship between use of NPS and other drugs in young Europeans responding to the ESPAD survey (aged 15/16)? • Is NPS use associated with other ‘risk’ behaviours? • Do individual and national level variables predict substance use? • Does the predictive utility of these variables differ between countries? (exploratory analysis) Sumnall et al. , in prep

METHODOLOGY – MAIN POINTS • Secondary analysis of ESPAD - European School Survey Project on Alcohol and Other Drugs (http: //www. espad. org/). • Large school-based substance use survey; sample size 2015 = 96046 in 35 countries • Analytical sample = 88, 206 cases (91. 8%) • Latent Class Analysis (LCA; Level 1 – Pupils) (Vermunt & Magidson, 2000) • Multilevel Latent Class Analysis (MCLA; Level 1 and 2 [countries]) (Henry & Muthén, 2010) • Covariate analysis – describing the derived Classes • Distal variable analysis – predictive utility of Modelled Classes

U 1 U 2 Z U 3 U 4 U 5 C U 6 U 7 Y Level 1 Within Country Multilevel Latent Class Model with Three Level 1 Latent Classes— Parametric Approach with Level 2 Factor(s) on Random Latent Class Intercepts C#1 FC U 0 C#2 Level 2 Across Country W

VARIABLES IN THE MODELS • Level 1 indicators • LTP NPS; LMP Tobacco; Alcohol (LTP; LYP; LMP abstainer); LTP cannabis; LTP ecstasy; LTP inhalants; LTP tranquilisers • Level 1 covariates • Sex; Perceived risks of trying cannabis, ecstasy, occasional cigarettes; HED on weekend; family income; parental education; rules setting and monitoring; warmth and care; emotional support • Level 1 distal variables • Lifetime occasions of cigarette use; age of first drunkenness; frequency of HED in last 30 days; ease of accessing cannabis; gambling frequency in previous year; time on Internet; harms from i) own use of alcohol ii) others’ use of alcohol • Level 2 covariates • GINI coefficient; population size; adult population drug use prevalence; GDP; urbanisation of population; NPS ‘blanket ban’; NPS laws short of ‘blanket ban’

MAIN FINDINGS 1 – LCA SOLUTION Indicator Class 1 - Alcohol users and abstainers 69. 5% Class 2 – CAT users 24. 0 % Class 3 – Polysubstance Users 6. 6% Wald statistic No Yes 0. 97 0. 03 0. 39 0. 61 0. 24 0. 76 7853. 00*** No Yes 0. 99 0. 01 0. 59 0. 41 0. 08 0. 92 3401. 7*** No Yes 1. 00 0. 00 0. 99 0. 01 0. 70 0. 30 3920. 25*** No Yes 0. 96 0. 04 0. 90 0. 10 0. 65 0. 35 3809. 48*** No Yes 0. 96 0. 04 0. 91 0. 09 0. 66 0. 34 3604. 00*** No Yes 0. 99 0. 01 0. 28 0. 13 0. 27 0. 32 0. 98 0. 02 0. 01 0. 02 0. 16 0. 80 0. 34 0. 66 0. 03 0. 02 0. 10 0. 85 6295. 47*** 6591. 47*** Last month use of cigarettes Lifetime use of cannabis Lifetime use of ecstasy Lifetime use of inhalants Lifetime use of tranquilisers Lifetime use of NPS Alcohol use status 1 Abstainer Used in lifetime Used in last year Used in last month Level 1 LCA – 3 Class Solution; BIC = 521773; classification error 10%.

MAIN FINDINGS 2 – COVARIATE ANALYSIS Class Covariate CAT users Sex SE Polysubstance Users Femalea SE Wald χ2 0. 01 0. 03 0. 14 0. 05 9. 67** Perceived risk scores Trying Cannabis -0. 81 0. 02 -1. 40 0. 04 3336. 06*** Trying Ecstasy 0. 56 0. 02 0. 23 0. 03 706. 77*** Occasionally smoking cigarettes -0. 38 0. 02 0. 10 0. 03 444. 91*** HED -0. 26 0. 02 -0. 17 0. 03 261. 32*** Male Family income better than average Noa Yes -0. 25 0. 04 -0. 52 0. 06 92. 66*** College/University education (father) Noa Yes -0. 05 0. 03 0. 17 0. 05 11. 41** College/University education (mother) Rules score Warmth and support score Noa Yes -0. 07 0. 03 0. 05 5. 98 -0. 09 0. 01 0. 00 -0. 17 -0. 02 0. 01 1016. 05*** 19. 10*** 3 -Step covariate analysis - Level 1. For Class 3, 16% of the total variance was explained by country level effects

MAIN FINDINGS 3 – DISTAL VARIABLE ANALYSIS Dependent variable Cluster CAT users Lifetime cigarette use 1. 32 0. 02 Polysubstance Users 1. 58 Age of first drunkenness (reversed scored) 1. 26 0. 01 1. 33 0. 02 9330. 6*** Frequency of HED in previous 30 days 1. 30 0. 02 1. 45 0. 02 6023. 1*** Ease of accessing cannabis 0. 79 0. 01 1. 44 0. 03 7514. 6*** Frequency of gambling in previous 12 months 0. 34 0. 01 0. 55 0. 01 1565. 4*** Weekday 0. 26 0. 01 0. 28 0. 02 1154. 1*** Weekend 0. 15 0. 01 0. 13 0. 02 373. 57*** Own use of alcohol 1. 20 0. 03 1. 26 0. 03 1892. 7*** Others’ use of alcohol 0. 35 0. 01 0. 60 0. 01 2798. 0*** Hours on the Internet Harm scores 3 -Step distal variable analysis – Level 1 SE SE Wald χ2 0. 03 7394. 2***

MAIN FINDINGS 4 – COUNTRY LEVEL VARIABLES (EXPLORATORY ANALYSIS) Covariate Class CAT users SE Polysubstance SE Wald χ2 Users GINI coefficient -3. 03 1. 4 2. 08 2. 20 5. 04 Population (log transformed) 0. 47 0. 08 0. 36 0. 18 40. 63*** Prevalence of adult drug use 0. 02 0. 10 0. 03 21. 32*** GDP (log transformed) -0. 50 0. 21 -1. 39 0. 36 25. 67*** % urbanization -0. 01 5. 31 Blanket ban on NPS (reference = no) -0. 18 0. 14 -0. 21 0. 24 4. 62 National NPS legislation (reference = no) -0. 15 0. 12 -0. 35 0. 16 6. 55* 3 -Step covariate analysis – Association between Level 2 covariates and Level 1 Classes

4 % of sample reported NPS use. Most likely to be members of Class 3 (Polysubstance Users; 6. 6% of respondents). • This group better characterised by use of multiple substances (not enough data to clarify nature of polysubstance use) • Reported lower substance-use risk perception, lower family income, and lower reports of parental warmth. • More likely to self-report additional risk behaviours such as heavy episodic drinking, gambling, more frequent cigarette use, and alcoholrelated harm. Between country variance in latent distribution • Exploratory analysis suggested Class 3 (Polysubstance Users) members more likely to live in more populous countries with lower per capita GDP, high adult prevalence of drug use, and had introduced NPS legislation short of a blanket ban. • Only simple policy indicators utilised SUMMARY OF MAIN FINDINGS

CONCLUSIONS • NPS use only reported by a small proportion of young Europeans in the ESPAD survey • Associated with a broader profile of risk • Important between-country variation in NPS use and profile of NPS use • Universal NPS specific (or substance specific) prevention/education unlikely to be cost effective – address NPS use as part of broader programme of activities.

STRENGTHS AND WEAKNESSES First pan-EU analysis of patterns of NPS use in young people Limited nature of NPS questions (compared to other substances) Inherent strengths & weaknesses of ESPAD Missing data and missing countries Causality Exploratory heuristic not reified groups of young people

FURTHER ANALYSES? • Focus on NPS users – individual vs group level predictors of use • Smaller sample of countries of higher use prevalence - which ones? Justification? Which variables? • ESPAD 2021 – ‘Longitudinal analysis’ (using propensity score matching’, including effects of introduction of legislation) • LCA at Level 2 (i. e. grouping at country level)

Contact: Harry Sumnall h. sumnall@ljmu. ac. uk Public Health Institute Liverpool John Moores University @profhrs @euspr