The Unexpected Public Health Effects of the 21 Drinking Age Richard A (Rick) Grucza, PhD MPE 12 th Annual Guze Symposium on Alcoholism Missouri Alcoholism Research Center Washington University, St. Louis, Missouri February 16, 2012 Disclosure Statement Richard A Grucza Washington University School of Medicine The 12th Annual Guze Symposium on Alcoholism February 16, 2012 Source of research support: NIH (NIDA, NIAAA), American Foundation for Suicide Prevention, Washington University Consulting relationships: None Stock equity: No healthcare or pharmaceutical related interests. Speaker s Bureaus: None 1
Outline of Presentation I. Secular Trends in Binge drinking and the (im)moderating effect of college II. The 21 policy Delay or Prevention? III. College style drinking: Distinguishing heavy (binge) drinking from non heavy drinking IV. The 21 Policy and drinking patterns History of the U.S. Minimum Legal Drinking Age (MLDA) 1930s 1940s 1950s 1960s 1970s 1980s 1990s Prohibition Ends Most States (37) keep MLDA=21 30 States Lower MLDA First Reports of Increased MV Fatalities Some States Voluntarily Raise MLDA Vietnam, 26 th Amendment National Uniform Drinking Age Act (1984) Louisiana Loophole De Facto National MLDA=21 2
I. Secular Trends in Binge Drinking Acknowledgements Survey data from 500,000+ subjects, 20 administrations Binge = 5+ Drinks @ one time (in 30 days prior to survey) NSDUH (f/k/a NHSDA) longest running measure of binge drinking in both youth and adults Analyses focused on ages 12 34. Relative risks for by age, gender and college attendance J. Am. Acad. Child. Adolesc. Psychiatr. July 2009 3
Boys/Men Girls/Women Ages 21 23 Ages 18 20 Ages 21 23 Ages 15 17 Ages 18 20 Ages 12 14 Ages 15 17 Ages 12 14 Back calculated estimated prevalence based on trends in relative risk assuming: (1) More recent estimates=most accurate (2) No major change in prevalence at ages 26 35 (supported by data from BRFSS). Macro secular trends: 1979 2006 Monitoring the Future: 1975 2010 Some States Voluntarily Raise MLDA National Uniform Drinking Age Act (1984) Johnson, O,Malley, Bachman and Schulenberg, MTF report, 2010 4
Changes in Binge Drinking Relative Risk Risk, 1979 2006 (Assessed on past 30 basis: relative to 24 35 y/o reference group) Trends in Binge Drinking: Men Ages % Change p (1979-2006) 21-23 Student -8% 0.60 21-23 Non-Student -24% 0.03 18-20 Student -24% 0.09 18-20 Non-Student -56% <0.001 Trends in Binge Drinking: Women Ages % Change p ( 1979-2006) 21-23 Student +150% <0.001 21-23 Non-Student +49% 0.02 18-20 Student +20% 0.34 18-20 Non-Student -2% 0.91 Men: Smaller reductions in binge drinking among students vs. non students (Significant among 18 20) Women: Larger increases in binge drinking among students vs. non students in (Significant among 21 23) Macro secular trends: 1979 2006 As a whole, binge drinking among youth / young adults decreased This happened in an era of national 21 MLDA and other policy initiatives But not apparent in girls/women: Narrowing gender gap College age student status seems to buffer against larger societal trends 5
II. The 21 Drinking Age: Delay or Prevention? Documented effects of the 21 MLDA Reduced alcohol consumption among <21 Reduced DUI fatalities Smaller effects: Reduction in homicides among ages 18 20 Reduction in suicides among ages 18 20 Reduction in unintentional injuries ages 18 20 Jones, Pieper, and Robertson, 1992; Birckmayer and Hemenway, 1999; Hingson, Meerigan, and Heeren, 1985; Links, 2000; Wagenaar & Toomey (review), 2002 6
Global Distribution of Drinking Ages 120 Number of Countries 100 80 60 40 20 0 Drinking Age First Intoxication Before Age 13 Adolescents Aged 15 16 40 35 30 25 20 15 10 5 0 Sources: Hibell, et al., 2004; Johnston, O Malley, Bachman, & Schulenberg, 2004; Slide Courtesy of Jim Fell, Pacifica institute 7
Early Use of Alcohol, Drugs, Tobacco Robust predictor of subsequent dependence 1985 ECA (Robins & Przybeck) many, many replications Lifetime prevalence of DSM IV Alcohol Dependence (NESARC): Began drinking 19 or younger: 30% Began drinking at 20 or older: 10% 15 Does delaying alcohol use decrease alcohol related problems in adulthood? On the one hand: On the other hand: 8
MLDA Policy as a Natural Experiment What if we could randomly assign one group of young people to an environment where they had easy access to alcohol, and another group to an environment where they didn t? MLDA differed according to state and birth year. People are randomly assigned to when and where they are born 30 28 b. <1953 b. 1954 1961 b. <1953 b. 1954 1961 Prevalence of Hypothetical Bad Outcome 26 24 22 20 18 16 14 12 IL MO 10 <1953 1954 1961 9
# Of States Permitting Sales to Under 21 (MLDA < 21) 10
Does MLDA @ 18 20 predict Alcohol Use Disorder later, in adulthood? Defined as past year DSM IV abuse and/or dependence Prevalence = 8.5% in 2001 Data from two national surveys (NLAES, NESARC, 1991, 2001). Subjects Age 21 53 at time of survey. 33,869 informative subjects (born 1948 1970) Based on state of residence, estimate exposure to MLDA < 21 at ages 18 20 All models include fixed effects for state and birth year cohort, as well as basic demographics (age, sex, race). Extended model includes other risk factors. Similar results for all demographic groups 11
III. Distinguishing heavy from non heavy drinking Annual Surveys of That Query Alcohol Use 3 Most frequent items: On how many days in the past (year) have you drank How many drinks, on average, do you consume on days that you drink On how many days in the past (year) have you had five or more drinks on one occasion? 12
Annual Surveys of That Query Alcohol Use Drinks per day = # Drinking Days * Drinks per Drinking Day # of Days in (week / mo / year The J-Shaped Curve Di Castelnuovo, A. et al. Arch Intern Med 2006;166:2437-2445. Relative risk of total mortality (95% confidence interval) and alcohol intake extracted from 56 curves using fixed- and random-effects models 13
Annual Surveys of That Query Alcohol Use # Drinking Days = # Heavy (5+) Days +# Non Heavy (<5) Days # Heavy (5+) Days +# Non Heavy (<5) Days Drinks per day R 2 =0.70 1997 2001 National Health Interview Survey: Ages 18 64, 12+ drinks in past year N=65,402 14
Mortality Analysis NHIS mortality follow up through 2006 1997 2001 samples: N=128,233 Decedents=6,416 Survival Analysis: Heavy days, Non heavy days, BMI, smoking, demographics Fractional Polynomials Best curve fit on two primary variables Mortality Risk: Through 2006 (Hazard Ratios vs. Lifetime Abstainers) Heavy Drinking Days Non Heavy Drinking Days Drinking Days Per Week 15
MLDA as a Predictor of Drinking Status Data from two national surveys (NLAES, NESARC, 1991, 2001). Subjects Age 21 53 at time of survey. (Same sample as AUD analyses) MLDA < 21 as risk factor: Outcome N OR (95% CI) p Ever drink vs. Lifetime Abstention 39,412 1.02 (0.93, 1.11) 0.66 Among Lifetime Drinkers Any Past Year drinking vs. None 25,240 1.02 (0.92, 1.12) 0.72 Weekly drinking vs. Less Frequent 25,240 0.99 (9.90, 1.09) 0.81 MLDA as a Predictor of Drinking Status Data from two national surveys (NLAES, NESARC, 1991, 2001). Subjects Age 21 53 at time of survey. (Same sample as AUD analyses) MLDA < 21 as risk factor: Outcome N OR (95% CI) p Frequency of Heavy (Binge) Drinking 24,253 2+ / Month or Higher n=7,202 1.15 (1.04, 1.28) 0.009 <2 / Month but >0 n=4,508 1.01 (0.88, 1.16) 0.91 None in Past Year n=12,813 1.0 (ref) 16
MLDA as a Predictor of Drinking Status Data from two national surveys (NLAES, NESARC, 1991, 2001). Subjects Age 21 53 at time of survey. (Same sample as AUD analyses) MLDA < 21 as risk factor: Outcome N OR (95% CI) p Frequency of Heavy Drinking 24,253 2+ / Month or Higher n=12,457 0.86 (0.74, 1.00) 0.053 <2 / Month but >0 n=8,045 0.81 (0.71 0.94) 0.005 None in Past Year n=3,588 1.0 (ref) Tying It All Together Prevalence of binge drinking among youth / young adults has gone down since late 1970s Campus environment may be buffering broader societal trends. 21 drinking age may have persistent protective effects against alcohol use disorders 21 may prevent binge drinking and promote moderate drinking 17
Acknowledgements Collaborators Laura Bierut Karen Norberg (NBER) Patty Cavazos Rehg Arpana Agrawal Andrew Plunk Husham Syed Mohammed Funding (RG) R01 AA017444 R21 DA026612 R01 DA031288 Collaborators MARC ARTSS (AA11998) T32 DA07313 18