Early and risky adolescent alcohol use independently predict alcohol, tobacco, cannabis and other drug use in early adulthood in Ireland: a longitudinal analysis of a nationally representative cohort

Table 4 Age at first alcoholic drink, AUDIT score at 17 years old and high-risk alcohol, tobacco, cannabis and other drug use at 20 years old examined using survey-weighted multivariable GEE regression models in multiply imputed data (N鈥�=鈥�4554)

听	High-risk alcohol*	听	Tobacco**	听	Cannabis***	听	Other drugs***
听	aOR (95% CI)	p	aOR (95% CI)	p	aOR (95% CI)	p	aOR (95% CI)	p
Age at first alcoholic drink	听	听	听	听	听	听	听	听
鈮も赌�14	Ref	听	Ref	听	Ref	听	Ref	听
15	0.88 (0.66鈥�1.18)	0.395	0.92 (0.69鈥�1.23)	0.573	0.95 (0.72鈥�1.24)	0.688	0.61 (0.46鈥�0.82)	0.001
16	0.68 (0.49鈥�0.95)	0.024	0.52 (0.40鈥�0.67)	&濒迟;鈥�0.001	0.75 (0.58鈥�0.97)	0.027	0.37 (0.28鈥�0.50)	&濒迟;鈥�0.001
17	0.70 (0.46鈥�1.05)	0.087	0.35 (0.26鈥�0.48)	&濒迟;鈥�0.001	0.45 (0.32鈥�0.64)	&濒迟;鈥�0.001	0.21 (0.14鈥�0.30)	&濒迟;鈥�0.001
鈮モ€�18/no alcohol****	0.19 (0.06鈥�0.55)	0.002	0.08 (0.04鈥�0.15)	&濒迟;鈥�0.001	0.19 (0.08鈥�0.42)	&濒迟;鈥�0.001	0.11 (0.05鈥�0.24)	&濒迟;鈥�0.001
AUDIT score at 17 years old	听	听	听	听	听	听	听	听
Low-risk	Ref	听	Ref	听	Ref	听	Ref	听
Increasing risk	3.08 (2.25鈥�4.22)	&濒迟;鈥�0.001	1.94 (1.57鈥�2.39)	&濒迟;鈥�0.001	1.28 (1.02鈥�1.61)	0.033	1.63 (1.27鈥�2.08)	&濒迟;鈥�0.001
High-risk	11.45 (7.04鈥�18.63)	&濒迟;鈥�0.001	2.07 (1.28鈥�3.33)	0.003	1.31 (0.85鈥�2.00)	0.218	2.53 (1.58鈥�4.06)	&濒迟;鈥�0.001
Tjur鈥檚 R2*	0.139	听	0.156	听	0.191	听	0.23	听
VIF *	All <鈥�4.25	听	All <鈥�2.3	听	All <鈥�3.5	听	All <鈥�3.35	听

AUDIT鈥�=鈥堿lcohol Use Disorders Identification Test categorised as 鈥�0鈥�7鈥� (low-risk), 鈥�8鈥�15鈥� (increasing risk) and 鈥�>15鈥� (high-risk), GEE鈥�=鈥塯eneralised estimating equations, aOR鈥�=鈥塧djusted odds ratio, CI鈥�=鈥塩onfidence intervals, p鈥�=鈥塸-value, R2鈥�=鈥塕 square, VIF鈥�=鈥塿ariance inflation factor
All models were adjusted for age at outcome, sex, Drumcondra test, extraversion, agreeableness, emotional stability, openness, conscientiousness, SDQ conduct, SDQ hyperactivity, parental monitoring, household social class, household structure, family AUD/drug-use & neighbourhood substance use
*The model examining high-risk alcohol use was additionally adjusted for parental education and peer drunkenness
**The model examining tobacco use was not adjusted for parental education, peer or parental tobacco use due to issues with multicollinearity
***The models examining cannabis and other drug use were additionally adjusted for parental education and peer cannabis use
**** Categories鈥夆墺鈥�18 (7% of participants) and no alcohol use (3%) were collapsed due to small numbers to achieve stable model estimates
AUDIT scores were categorised as 鈥�0鈥�7鈥� (low-risk), 鈥�8鈥�15鈥� (increasing risk) and 鈥�>15鈥� (high-risk)
* Tjur鈥檚 R2 and VIF estimated for analyses from each of ten imputed data frames and averaged

ISSN: 1471-2458