Alcohol Consumption Behavior among Undergraduate Students in Thailand: Development of a New Causal Relationship Model

Alcohol consumption among undergraduate students in Thailand is problematic. The aim of this study was to deepen our understanding of this problem by developing a causal relationship model for the alcohol consumption behavior of undergraduate students in Thailand, and to verify the model’s concordance with empirical data. Four latent variables were considered: alcohol consumption behavior, alcohol expectancy, drinking refusal self-efficacy, and health literacy. Participants included representative 1 – 5 year undergraduate students at the Thailand National Sports University, with 600 students being selected using stratified random sampling procedures. The descriptive statistics and the causal relationship model were analyzed using LISREL 8.80. The model developed was in good agreement with the empirical data (χ=228.66, df = 79, p > 0.05, χ/df=2.894, SRMR =0.07, RMSEA =0.06, CFI =0.99, and RFI =0.98), with all computed indices passing the stipulated criteria. On the basis of the coefficients of determination in the structural equation model, alcohol expectancy, drinking refusal self-efficacy, and health literacy together accounted for 80% of the variance in the student’s alcohol consumption behavior. These theoretically based causal factors provide new directions for future intervention work aimed at modifying the alcohol consumption behaviors of undergraduate students at the Thailand National Sports University. This can be accomplished by developing activities that are suitable and contextually sensitive to their needs.


Introduction
Excessive alcohol consumption causes more than 60 individual diseases, diseases that affect the lives of not only individual consumers, but also their families, friends, communities, and nations, making it a significant public health problem (WHO, 2014). This is especially noticeable in Thailand, where an estimated 52 liters of alcohol are consumed per person per year (WHO, 2014). Moreover, the number of new drinkers is estimated to increase by 250,000 people annually (Center for Alcohol Studies, 2014).
Undergraduate students in Thailand contribute to these negative trends, with 56.8% exhibiting harmful alcohol consumption behaviors (Gilles, Turk, & Fresco, 2005). Within certain geographical locations, the situation may even be worse. For example, Buasorn and Ratchadapunnathiku (2012), who studied the alcohol consumption behavior of undergraduate students in Bangkok, reported that 73.8% exhibited alcohol consumption behaviors up to a hazardous or harmful level. Consuming alcohol at a hazardous level has been shown to increase risk-taking behavior by as much as 25 times in comparison to ordinary undergraduate students (Ham & Hope, 2003;Karam, Miller, 1992;Petraitis, Flay, & Miller, 1995;Ham & Hope, 2003). While multi-level influences are important, it is also acknowledged that environmental and social factors are subject to changing social dynamics and mores. Furthermore, such factors reside beyond an individual's control and are difficult to change or intervene upon (Glanz, Rimer, & Viswanath, 2015). As such, identifying actionable mechanisms of behavior change at the individual level is important (Ponrachom, Boonchuaythanasit, & Cardinal, 2020).
Toward this end, Ponrachom et al. (2020) identified three relevant variables from their narrative literature review. Namely, alcohol expectancy and drinking refusal self-efficacy (Oei & Morawska, 2004;Oei & Jardim, 2007), and health literacy (Rundle-Thiele, Siemieniako, Kubacki, & Deshapnde, 2013;Chisolm, Manganello, Kelleher, & Marshal, 2014;Chang, Miao, Lee, Chen, Chiu, & Lee, 2016). Through the identification of these variables, Ponrachom et al. proposed a new paradigm focused on strengthening individual immunity through empirically supported psychological constructs. While the variables in their model have been individually acknowledged as important, their collective potential in understanding the alcohol consumption behavior of undergraduate students in Thailand is unknown. As a necessary first step in closing this gap, the aim of the present study was to develop a new causal relationship model for the alcohol consumption behavior of undergraduate students in Thailand, and to verify the model's concordance with empirical data.

Participants
Ethical approval to conduct this study was obtained from the Human Research Ethics Committee, Kasetsart University, Bangkok, Thailand (COA No. COA59/003; approved February 2, 2017). All study participants were at least 18 years old and they signed an informed consent form prior to their participation.
On the basis of the criteria of Hair, Black, Babin, Anderson, and Tatham (2006), a representative sample of 1 st -5 th year undergraduate students at the Thailand National Sports University were identified (N = 600). The two steps in this process were as follows.
Step 1: Thailand was divided into four geographical regions (i.e., Central, Northeastern, Northern, and Southern), with the primary sampling unit within regions being the campuses of the Thailand National Sports University. At each campus the secondary sampling unit was at the Faculty level (i.e., Faculty of Education, Faculty of Liberal Arts, and Faculty of Sport and Health Sciences). From there a tertiary sampling unit was the students' education level (i.e., 1 st year, 2 nd year, 3 rd year, 4 th year, and 5 th year students).
Step 2: On the basis of this stratification process, representative samples were selected using simple random sampling derived from a table of random numbers.

Measures
All data were obtained via self-report questionnaire. Three of the questionnaires were previously validated measures, and one was developed and validated specifically for the purposes of this study.
To assess the participants' alcohol consumption behavior, the Alcohol Use Disorder Identification (AUDIT) questionnaire was used (Saunders, Aasland, Babor, de la Fuente, & Grant, 1993). This measure has well-established psychometric properties and has been used in several countries around the world for decades. It has high sensitivity and specificity in diagnosing alcohol consumption.
Permission to use the Alcohol Expectancy Questionnaire from The Comprehensive Effect of Alcohol (CEOA) (Fromme, Stroot, & Kaplan, 1993) and the Drinking Refusal Self-Efficacy Questionnaire (DRSEQ) (Oei & Borrow, 2000) were obtained from the copyright holders. These instruments were further developed in accordance with the tool creation and quality test standards set forth by Maxwell (1996). The Alcohol Expectancy Questionnaire had an overall Cronbach alpha coefficient of .93, and values for each sub-domain ranging from .62-.81, while the DRSEQ had an overall Cronbach alpha coefficient of .95, and values for each sub-domain ranging from .87-.95. All internal reliability coefficients were deemed satisfactory (Ponterotto & Ruckdeschel, 2007).
The final measure used in this study was developed specificially by the researchers for purposes of this research; it was the Health Literacy Questionnaire. The questionnaire attained a content validity index of .78-1.00, was verified by nine experts, and it was quality tested by 60 representative samples from the population using people who were not involved in the study sample. It achieved Cronbach alpha coefficient values ranging from .70-.91, which indicated that this instrument satisfied the required standard of reliability necessary in research (Ponterotto & Ruckdeschel, 2007).

Procedures
The Faculty at the Thailand National Sports University assisted with data collection at their respective campuses. Participants were asked to complete the questionnaires as distributed. Questionnaires were coded, encrypted, and verified prior to data analysis.

Analysis
Data were analyzed using descriptive statistics (e.g., mean [M], standard deviation [SD], skewness, and kurtosis) and the causal relationship model's goodness of fit was measured using LISREL 8.80.

Descriptive Summary
The largest majority of students cae from the Northen campus (27.5%). They were dispersed across each Faculty as follows: Education (36.3%), Health and Sport Sciences (33.5%), and Liberal Arts (30.2%). Most commonly they were in their 3 rd year (31.5%). Their M age was 20.83 (SD = 1.50) years, with the majority being males between 18-30 years of age (68.5%). The largest majority had a grade point average between 2.51-3.00 (37.0%).
Half of the participants tried alcohol for the first time when they were between 16-18 years of age (M = 16.39, SD = 2.15). The top three reasons given for first trying alcohol were trying on their own will (89.3%), friend's incitement (47.3%), or as a challenge (15.8%). The top three alcohol beverages consumed by the participants were beer (92.0%), liquors/red liquors (inclusive of Thai-Foreign brandy) (49.8%), and white liquors/local liquors (34.5%).
The pattern of their alcohol consumption behavior was categorized on the basis of the measurement standard of the World Health Organization (Saunders, Aasland, Babor, de la Fuente, & Grant, 1993). The most common pattern found was hazardous drinking (38.5%), followed by low risk drinking (30.8%), harmful drinking (19.5%), and alcohol dependence (11.2%).

Preliminary Analysis: Check for Multicollinearity
There were 135 pairs of Pearson's correlation coefficient values. They ranged in size from -.752 to .844 (all p < 0.01), with 64 of those pairs being positive (i.e., ranging from .290 to .844). The analysis of the correlation coefficients reflects that the observed variables were suitable for inclusion in the structural equation model, as none of the pairs was >.85, suggesting that multicollinearity was not a major issue (Kline, 2005).

Multivariate Analysis: Structural Equation Model
The structural equation model is shown in Figure 1. The model developed was in good agreement with the empirical data (χ 2 =228.66, df = 79, p > 0.05 , χ 2 /df=2.894, SRMR =0.07, RMSEA =0.06, CFI =0.99, and RFI =0.98), with all computed indices passing the stipulated criteria. As can be seen, alcohol expectancy was directly influenced by drinking refusal self-efficacy, with a path coefficient of -0.59 (p < 0.05), and health literacy, with path coefficient of -0.17 (p < 0.05). Furthermore, alcohol expectancy was indirectly influenced by health literacy, with a path coefficient of -0.29 (p < 0.05).
Looking at alcohol consumption behavior, it was directly influenced by drinking refusal self-efficacy, with a path coefficient of -0.55 (p < 0.05), as well as directly influenced by health literacy and alcohol expectancy, with path coefficients of -0.41 and 0.09, respectively (both p < 0.05). In addition, alcohol consumption behavior was indirectly influenced by health literacy, with a path coefficient of -0.31 (p < 0.05), as well as indirectly influenced by drinking refusal self-efficacy, with a path coefficient of -0.05 (p < 0.05).
Finally, drinking refusal self-efficacy was directly influenced by health literacy, with a path coefficient of 0.48 (p < 0.05).
Having considered the coefficient of determination from the structural equation model, health literacy described 48% of the variance of drinking refusal self-efficacy, health literacy and drinking refusal self-efficacy together describes 23% of the variance of alcohol consumption behavior, and health literacy, alcohol consumption behavior, and drinking refusal self-efficacy combined described 80% of the variance in alcohol consumption behavior (Table 1).

Discussion
In this stratified random sample of 600 undergraduate students at the Thailand National Sports University, 38.5% were drinking alcohol at a hazardous level, 30.8% at a low risk level, 19.5% at a harmful level and another 11.2% were alcohol dependent. Clearly, there is an urgent need for educational and social interventions aimed at mitigating the alcohol consumption behavior of undergraduate students at the Thailand National Sports University.
In this study, their alcohol consumption behavior was well explained by three factors: alcohol expectancy, drinking refusal self-efficacy, and health literacy. Health literacy provides a direct negative influence on alcohol consumption behavior, as well as an indirect influence via alcohol expectancy and drinking refusal self-efficacy.
The negative relationship between health literacy and alcohol consumption behavior indicates that undergraduate students with a high level of health literacy exhibited declining levels of alcohol consumption behavior. This was observed in the amount consumed, the frequency of consumption, and the seriousness of alcohol drinking problems. This is an important discovery, as health literacy skills can be taught and learned.
Alcohol expectancy was found to provide a direct positive influence on alcohol consumption behavior. In other gjhs.ccsenet.org Global Journal of Health Science Vol. 13, No. 3;2021 words, the higher the degree of alcohol expectancy the students have, the more alcohol they consume. This is consistent with the findings of others (Chisolm, Manganello, Kelleher, & Marshal, 2014;Gilles, Turk, & Fresco, 2005;Oei & Morawska, 2004).
Drinking refusal self-efficacy provides a direct negative influence on alcohol consumption behavior, meaning that the higher the degree of drinking refusal self-efficacy the students have, the more likely they are to consume less alcohol. Others have also observed this phenomenon (Gilles, Turk, & Fresco, 2005;Oei & Jardim, 2007;Oei & Morawska, 2004). This is because drinking refusal self-efficacy involves the confidence or courage to refuse to consume alcohol across a variety of situations. In addition, drinking refusal self-efficacy also provides an indirect influence on alcohol expectancy, which can help mitigate some of the alcohol expectancy effects.
Like all studies, this study is not without limitations. First, the causal relationships observed and tested in the model were derived from cross -sectional data. Second, all data were self-reported. Self-report data are subject to a variety of biases, such as item interpretation, recall, and social desirability. Third, the data were obtained from undergraduate students from all regions of Thailand; however, they were all part of the multi-campus Thailand National Sports University. As such, the results may not generalize to all university students in Thailand. These limitations should all be kept in mind.

Conclusion
While three factors were simultaneously considered in this study, the one that stands out most prominently as an immediately actionable intervention variable is health literacy. Those with higher health literacy drank less alcohol. Moreover, those with higher health literacy score had increased drinking refusal self-efficacy and tempered alcohol expectancies. In other words, they not only drank less, but they were better prepared for a variety of drinking opportunities that might otherwise emerge in their lives (e.g., peer pressure, social setting), and they were more aware and informed about the deleterious effects of alcohol. The new causal relationship model of alcohol consumption behavior developed in this study can be further examined in other institutions and settings. Moreover, the model provides a framework for educational interventions.