Modeling the College Drinking Environment 1 Understanding the College Drinking Problem: Developing Mathematical Models to Inform Policy Ben G. Fitzpatrick, PhD Kate Angelis, PhD Tempest Technologies White Paper March 2014 www.tempest-tech.com
Modeling the College Drinking Environment 2 Contents 1. Introduction... 3 2. The Model... 5 3. Social Norms Marketing Research Project (SNMRP) Data... 6 4. Results and Discussion... 7 5. Conclusions... 9 6. References...10
Modeling the College Drinking Environment 3 1. Introduction A recent 2013 report has painted a grim picture of the effects of excessive alcohol consumption in the U.S. In addition to negative health and social consequences that include increased healthcare needs, elevated crime rates and lost worker productivity, the estimated economic cost of this public health threat reached $223.5 billion in 2006 1. Data related to excessive drinking among college students is far more worrisome and represents a dangerous public health problem for academic institutions across the country. Researchers at the National Institute on Alcohol Abuse and Alcoholism (NIAAA) have estimated that each year 1,825 college students die from alcoholrelated injuries and roughly 600,000 students are unintentionally injured (Table 1) 2. The phenomenon of college drinking is pervasive: more than 80% of college students drink alcohol and nearly 50% admit to binge drinking (defined as five (5) or more drinks for men and four (4) or more drinks for women in a two-hour period 3, 4 ) in the past two weeks 5. The five/four binge activity has been extensively studied and is strongly associated with negative outcomes 6, ranging from academic performance problems and date rape to serious injury and death. In previous research, campus violence has been tied to the physical availability of alcohol, often measured by alcohol outlet density (the number per square mile of establishments that sell alcohol) 7. In addition to the number of physical sources of alcohol in a given area, other environment features promote heavy episodic (binge) drinking, including residential setting, low price of alcohol and the unique drinking environment at a given college 6. Binge or heavy episodic drinking has been called arguably the No. 1 public health hazard and the primary source of preventable morbidity and mortality for the more than 6 million full-time college students in America 8. Understanding the environmental parameters that impact heavy episodic drinking is crucial to prevention efforts. The wetness parameter captures a combination of factors that promote heavy drinking in a college setting. Table 1. Consequences of excessive alcohol consumption by college students (ages 18-24). In order to add clarity to the college drinking problem, we are bringing the tools of systems science to bear to characterize the college drinking environment quantitatively. To that end, we recently introduced a mathematical model we call SimHED, which simulates the campus drinking environment 9. Our model defines the college landscape as a system in which drinking by college students is not only influenced by individual factors it is also impacted by social norms and social interactions. These influences are driven by a key parameter in the model that describes the relative wetness of a given campus 10. This wetness index is inferred from the drinking behavior of the students at a particular college. In the present document, we employ this model to study problematic outcomes associated with excessive drinking on academic campuses. Past studies have shown a correlation between negative outcomes and alcohol outlet density 7, 10, 11, suggesting a direct relationship between physical availability and problem behaviors. Our wetness index is related to outlet density 12, but we feel that it carries more information than just physical availability. In this study, we aim to understand which parameter namely wetness or alcohol outlet density - has a
Modeling the College Drinking Environment 4 stronger relationship with college student-specific behaviors and negative outcomes, namely (1) the proportion of binge drinkers, (2) disciplinary actions on campus related to alcohol, and (3) the incidence of sexual assaults. Based on the model outputs and subsequent analyses, we present novel data that suggests that wetness correlates more strongly with the above than outlet density, suggesting that a complex interplay of campus-specific factors and influences, in addition to the physical availability of alcohol, may play a role in influencing student behavior and subsequent outcomes.
Modeling the College Drinking Environment 5 2. The Model Our model, detailed in previous work 9, involves a systematic approach to quantifying college drinking dynamics, in which we (1) characterize the key processes that are believed to be responsible for driving the college drinking problem, (2) develop mathematical relationships that describe the dynamics of these processes, and (3) compare the outputs of the model with actual college drinking data. This approach led to three processes that define the rate parameters, including (1) individual characteristics that are associated with changes in drinking behavior, (2) social interactions that lead to changes in drinking behavior, and (3) social norms that affect drinking behavior. In addition to the above components we infused the model with essential patterns or styles of drinking on campus. Our construct has five drinking compartments, (1) abstainers, (2) light drinkers, (3) moderate drinkers, (4) problem drinkers, and (5) heavy episodic drinkers. A unique feature of the model is that it does not maintain detailed information about each individual drinker; rather, it accounts for the numbers of students following each of the five drinking styles. With five different compartments representing the drinking styles, and three possible transition paths that drinkers can take to move from one compartment to another, we are faced with a potentially large parameterization problem. Our research 9, 12 has led to a characterization of the transitions in terms of two parameters, misperception (social norms effect) and wetness, that model the campus environment. Misperception governs the amount of social norms pressure that moves students towards adopting heavier drinking styles. The wetness parameter captures a combination of factors that promote heavy drinking in a college setting. A conceptual diagram of our model and the interplay of the aforementioned parameters are illustrated in Figure 1 9. Figure 1. Conceptual model of college drinking.
Modeling the College Drinking Environment 6 3. Social Norms Marketing Research Project (SNMRP) Data In order to validate the model we compared outputs with data from student surveys. These student-level data were obtained from the Social Norms Marketing Research Project (SNMRP) conducted by DeJong and colleagues 13. This group-randomized study involved the examination of social norms marketing campaigns aimed at reducing student alcohol consumption at 32 colleges and universities across the U.S. (58.1% public institutions and 41.9% public institutions). The sampling of academic centers represented all four U.S. census regions (31% Northeast, 31% North Central, 16% West and 22% South) 7. To ensure privacy and confidentiality of all study participants, names of the colleges and universities have not been disclosed; each institution is referred to as School 1 through School 32. Using annual surveys, a random sample of 300 students per institution, per year were analyzed. The data was stratified by year in school to produce proportional representation. Based on survey answers related to drinking habits (quantity of drinking, frequency of drinking, etc.), students were categorized into one of the five drinking groups described earlier. As an example, Table 2 summarizes data for School 17 12. Table 2. Survey data for School 17. School 17 Abstainers Light Drinkers Moderate Drinkers Problem Drinkers Heavy Episodic Year 1 18 30 51 36 61 Year 2 21 46 56 46 71 Year 3 19 59 66 30 84 Year 4 19 45 69 39 82
Modeling the College Drinking Environment 7 4. Results and Discussion We used our model to understand the relationship between wetness and alcohol density and (1) the proportion of binge drinkers, (2) disciplinary actions on campus related to alcohol misuse, and (3) the incidence of sexual assaults. Based on these regression analyses (Figure 2), wetness is more strongly correlated to negative outcomes than is outlet density. Figure 2. (A) Proportion of heavy episodic drinkers, (B) liquor law disciplinary actions per 1000 students, and (C) rapes per 1000 students, as a function of wetness (left) and outlet density (right).
Modeling the College Drinking Environment 8 To explore the statistical significance of the relationships, we performed a set of regression analyses on a number of negative outcome variables, using wetness and outlet density as independent variables. In Table 3, we show the regression results. Note that wetness plays a significant role in the rates of violent crimes, liquor law disciplinary actions, rapes, and heavy episodic drinking. Outlet density correlates significantly with burglaries, but its statistically significant effects on liquor law arrests and on heavy episodic drinking are paradoxically negative, an outcome that may be associated with the positive relationship with wetness. Table 3. Regression analyses of outlet density and wetness. Wetness Coefficient Outlet Density Coefficient Wetness p Value Outlet Density p Value Violent Crimes 0.9546 0.0402 0.000191 0.694591 Liquor Law Arrests 1.5886-1.0562 0.099900 0.018800 Liquor Law Disciplinary Actions 3.4559-0.4003 0.000029 0.216000 Rapes 1.6463-0.0806 0.000010 0.569300 Robberies -0.6407 0.2470 0.261350 0.338520 Assaults 0.6706 0.1071 0.216000 0.660000 Burglaries 0.1345 0.4113 0.746400 0.035900 Proportion of Heavy Episodic Drinkers 0.5920-0.0746 0.000000 0.005720 We have also conducted pairwise comparison of the individual correlations between these negative outcomes and our two independent variables. In particular, we considered the null hypothesis that outlet density correlation is larger than wetness correlation, with a statistically significant p value providing evidence for wetness being more the strongly correlated quantity. As summarized in Table 4, for liquor law disciplinary actions, rape, and heavy episodic drinking, correlation with wetness is significantly larger. Table 4. Correlation analyses comparing outlet density and wetness. Correlation with Wetness Correlation with Outlet Density Z-Statistic p Value Violent Crimes 0.7189 0.4659-1.5247 0.063663 Liquor Law Arrests 0.0477-0.3101-1.4024 0.080395 Liquor Law Disciplinary Actions 0.6862 0.2712-2.1423 0.016084 Rapes 0.7527 0.3883-2.1682 0.015071 Robberies -0.1282 0.0666 0.7446 0.771732 Assaults 0.3307 0.2581-0.3029 0.380983 Burglaries 0.3275 0.4818 0.7058 0.759839 Proportion of Heavy Episodic Drinkers 0.8750 0.3277-3.8609 0.000056
Modeling the College Drinking Environment 9 5. Conclusions Mathematical models that allow the simulation and characterization of communities, as well as the individuals within a community, are powerful tools that enable policy creation, examination and implementation. The processes of model construction and calibration encourage us to investigate the system under study deeply and quantitatively. The dynamic SimHED model we have constructed provides a succinct encoding of drinking behavior on a campus and a tool for gaining additional insights. SimHED s driving parameter wetness can be estimated from observations of the drinking styles of the students. Most interestingly, this inferred parameter is highly correlated with negative outcomes such as rape and assault, outcomes that are not used in building the model. Taking the model to a finer level of detail, we will seek a deeper understanding of the details of the campus environment that lead to higher wetness indices, at which point we will be better able to address heavy episodic drinking and its negative consequences. Comments on this paper? Please visit us at www.tempest-tech.com
Modeling the College Drinking Environment 10 6. References 1. Sacks JJ, Roeber J, Bouchery EE, Gonzales K, Chaloupka FJ, and Brewer RD. State costs of excessive alcohol consumption, 2006. Am.J.Prev.Med. 45:474 (2013). 2. Hingson RW, Zha W, and Weitzman ER. Magnitude of and trends in alcohol-related mortality and morbidity among U.S. college students ages 18-24, 1998-2005. J.Stud.Alcohol Drugs. Suppl12 (2009). 3. Johnston LD, O'Malley PM, Bachman JG, and Schulenberg JE. Monitoring the future. National survey results on drug use, 1975-2012: Volume 2, College students and young adults ages 19-50, Ann Arbor: Institute for Social Research, The University of Michigan (2013). 4. Wechsler H and Nelson TF. Binge drinking and the American college student: what's five drinks? Psychol.Addict.Behav. 15:287 (2001). 5. National Institute on Alcohol Abuse and Alcoholism, College Drinking Fact Sheet. http://pubs.niaaa.nih.gov/publications/collegefactsheet/collegefactsheet.pdf (2013). 6. Wechsler H and Nelson TF. What we have learned from the Harvard School of Public Health College Alcohol Study: focusing attention on college student alcohol consumption and the environmental conditions that promote it. J.Stud.Alcohol Drugs. 69:481 (2008). 7. Scribner RA, Mason KE, Simonsen NR, Theall K, Chotalia J, Johnson S, Schneider SK, and DeJong W. An ecological analysis of alcohol-outlet density and campus-reported violence at 32 U.S. colleges. J.Stud.Alcohol Drugs. 71:184 (2010). 8. Wechsler H, Dowdall GW, Davenport A, and Castillo S. Correlates of college student binge drinking. Am.J.Public Health. 85:921 (1995). 9. Scribner R, Ackleh AS, Fitzpatrick BG, Jacquez G, Thibodeaux JJ, Rommel R, and Simonsen N. A systems approach to college drinking: development of a deterministic model for testing alcohol control policies. J.Stud.Alcohol Drugs. 70:805 (2009). 10. Scribner R, Mason K, Theall K, Simonsen N, Schneider SK, Towvim LG, and DeJong W. The contextual role of alcohol outlet density in college drinking. J.Stud.Alcohol Drugs. 69:112 (2008). 11. K.P.Theall FP, Scribner R, Cohen D, Bluthenthal RN, Schonlau M, Lynch S, and Farley TA, The neighborhood alcohol environment and alcohol-related morbidity. Alcohol Alcohol. 44:491 (2009). 12. Ackleh AS, Fitzpatrick BG, Scribner R, Simonsen N, and Thibodeaux JJ. Ecosystem Modeling of College Drinking: Parameter Estimation and Comparing Models to Data. Math.Comput.Model. 50:481 (2009). 13. DeJong W, Schneider SK, Towvim LG, Murphy MJ, Doerr EE, Simonsen NR, Mason KE, and Scribner RA. A multisite randomized trial of social norms marketing campaigns to reduce college student drinking. J.Stud.Alcohol. 67:868 (2006).