The association between largest dose and frequency of alcohol consumption, and employment situation
Roberto Ternes Arrial
School of improvement for education professionals, Brasília, Brazil
All data were obtained from NESARC, the National Epidemiologic Survey on Alcohol and Related Conditions. It consists “a nationally representative study, funded by the National Institute on Alcohol Abuse and Alcoholism (NIAAA), that explores substance use and mental health disorders of adults” (NESARC Codebook, 2013).
Population under investigation was restricted to individuals between 20 and 70 years old.
All variables were collected from a computer-assisted survey, conducted on households or workplaces of respondents. Variables used on this study were as follows:
Estimate quantity of alcohol consumed/month: quantitative. Number of alcohol drinks per month, estimated from number of days in the week when consumed alcohol, and how many drinks consumed per day. See annex for details
Type of drinker: categorical. Categorization of the above variable in three categories: light, medium and heavy drinkers.
Ever sought help because of drinking: categorical.
Presents situation includes working full time: categorical.
Unknown or blank data, as defined on the Codebook, were disregarded and ignored.
All statistical analyses were conducted on software SAS OnDemand for Academics. One graph was edited on software Microsoft Excel 2007. Some graph labels were edited on the Microsoft Paintbrush software.
Alcohol consumption is a worldwide, multicultural and common everyday activitity. Along the commonplace consumption, its use also has numerous meanings: religious, social, for fun, or even as a means to alter one´s state of mind (as a drug). Although its use is widely accepted while consumed with parcimony, the abuse is often observed, because alcohol also induces dependency. In such cases, abusers are prone to commit irrational acts such as crime, involvement on car accidents, vandalism, social/financial problems, among other unacceptable actions. Nowadays alcohol abuse is considered a psychiatric disease described on DSM-IV as alcoholism (Wikipedia, 2013).
One can expect to detect social disadjustments on individuals that abuse alcohol. If this drug induces behaviors that are not desirable in many societies, the abusers may be excluded from certain social activities, even to an extent that they may be considered a social pariah. On the other hand, another kind of alcohol abusers are people with anxiety disorders, which are unable to control drinking craves once they start (Wikipedia, 2013). This second type of alcoholics are not considered in the present study.
While it makes sense to consider that alcohol abuse may indeed impair social experience, it often is useful to determine which social interations are affected, and the frequency and amount of alcohol consumption that correlates with these behavior modifications.
In this work it is intended to seek a correlation between alcohol usage and the hability to keep a full-time job, which may be considered an important social behavior. Also, a relation between frequency and largest quantity of alcohol consumed is searched.
Around 60% of the sample reports as employed
Light, medium and heavy drinkers are nearly equally distributed (each around 30% of the sample)
Around 6% report having seek help because of drinking issues
Alcohol frequency was average 22.7 (s.d. 47.0)
Contrary to the expected, chi-square test showed that for the selected sample there is a significant difference of employement rate relative to type of drinker (X2 = 138.6, 2 df, p < .0001). Light drinkers are in average 60.6% employed, medium are 67.3% employed and heavy drinkers are 69% employed, as shown on Figure 1.
FIGURE 1. Full-time Employement rate according to drinking frequency of each individual.
Significant values were found between light and medium, and light and heavy drinkers, but not between medium and heavy drinkers. In other words, the more an individual drinks, more likely it is that he is employed.
FIGURE 2. Largest number and frequency of alcohol consumption among individuals who did not sought help because of drinking.
FIGURE 3. Largest number and frequency of alcohol consumption among individuals who did sought help because of drinking.
All analyses conducted on this work were performed using SAS Enterprise OnDemand for Academics version 4.3, kindly provided by the Wesleyan University/COURSERA. This work is part of Passion Driven Statistics Online Course for attaining certification of conclusion, under supervision of Dr. Lisa Diercker and coworkers.
What might the results mean?
Correlation between a high frequency of drinking and a large number of drinks consumed is expected, as alcohol causes dependency. It is an intuitive rationale that this work provides some support with its statistical results. Also noteworthy that considering the variable for seeking help because of drinking slightly improves the mentioned correlation. One may expect that, once engaged on frequent and abusive drinking, the individual is more likely to seek help. However, this relation may not be considered a statistical moderation, since it was significative even before it was taken into account.
On the other hand, an unexpected result emerges when considering employment and drinking habit. Light drinkers seem to be more unemployed than medium or heavy drinkers. This can be interpreted in many ways; one explanation may be that heavy drinkers may be less demanding in their jobs, that is, they assume lower income jobs, which are more common and easy to find.
The NESARC database is both large and trustworthy, nationally representative for the U.S.
While this work provides some insights correlating alcohol use and social behavior disadjustments, it is imprescindible to consider that, as any other human traits, the response variables considered are certainly influenced by other aspects of the individual´s life, even the use of other drugs. These other invisible, innumerable variables, which were not considered in this particular work, may be contributing to observed social behavior or influencing even more than alcohol drinking itself.
Recommended future research
Other social variables may be investigated further to clarify whether alcoholism indeed influences several social characteristics of an individual.
Also, in-depth studies may identify alcohol consumption quantity that induces detrimental social behaviors. By pinpointing the alcohol consumption limit that separates healthy social drinking from pathological consumption, one may expect to more easily identify when the line is being crossed and start seeking an advice or treatment. Also, after engaged on detrimental alcohol abuse, it may also be beneficial to know which areas are most influenced so that the help provided can be more precise and efficient.
Inputs for the SAS software, as used in this work, were as follows:
libname mydata “/courses/u_coursera.org1/i_1006328/c_5333” access=readonly;
DATA new; set mydata.nesarc_pds;
LABEL AGE = “Age”
MARITAL = “Current marital status”
SEX = “Sex” /*Not sure if will use sex, but keep it anyway*/
AGEGROUP = “Grouped age- 20-35, 36-48, 49+”
S1Q6A = “Higher grade or year of school completed”
EDUCATION = “Studies - low, intermediary, higher”
S1Q7A1 = “Presents situation includes working full time”
S1Q239 = “Had problem with neighbor, friend, or relative in last 12 months”
S1Q2310 = “Experiences major financial crisis, bankruptcy, or been unable to pay bills on time in last 12 months”
S1Q2311 = “You or family member had trouble with police, got arrested, or sent to jail in last 12 months”
ALCABDEP12DX = “Alcohol abuse/dependence in last 12 months”
ALCABDEPP12DX = “Alcohol abuse/dependence prior to the last 12 months”
S2AQ1 = “Drank at least 1 alcoholic drink in life”
S2AQ2 = “Drank at least 12 alcoholic drinks in last 12 months”
S2AQ3 = “Drank at lest 1 alcoholic drink in last 12 months”
S2AQ8A = “How often drank any alcohol in last 12 months”
S2CQ1 = “Ever sought help because of drinking”
NUMALCMO_EST = “Estimate number of alcohol drinks in a month”
ALCBYDAY_EST = “Grouping from NUMALCMO_EST - light, medium or heavy drinker”;
/*CONDITIONALS: Analyse only males because to reduce variability on full-time job. Aged on an age one expects to have social “obligations”.*/
/*IF SEX=1; — This will limit only to males. Would be interesting because of JOB; but may induce some kind of bias?*/
IF AGE GE 20 AND AGE LE 70;
/* Select only those who drink OR smoke*/
/*IF S2AQ1 = 1 OR S3AQ1A = 1;*/
/*Dealing with the unknown data: attribute a specific coding*/
IF S2AQ2 = 9 THEN S2AQ2 = .;
IF S2AQ3 = 9 THEN S2AQ3 = .;
IF S2AQ8A = 99 THEN S2AQ8A = 0;
IF S2AQ8B = 99 THEN S2AQ8B = 0;
IF S2AQ8C = 99 THEN S2AQ8C = .;
IF S2CQ1 = 9 THEN S2CQ1 = .;
IF S3AQ1A = 9 THEN S3AQ1A = .;
IF CHECK321 = 9 THEN CHECK321 = .;
IF S3AQ3B1 = 9 THEN S3AQ3B1 = .;
IF S3AQ3C1 = 99 THEN S3AQ3C1 = .;
IF S1Q239 = 9 THEN S1Q239 = .;
IF S1Q2310 = 9 THEN S1Q2310 = .;
IF S1Q2311 = 9 THEN S1Q2311 = .;
/*Making sense of alcohol drink coding. Variable ALCFREQMO is coded more user friendly than S2AQ8A*/
/*Inspired by Prof. Decker´s suggestion on smoking variables*/
IF S2AQ8A = 1 THEN ALCFREQMO = 30;
ELSE IF S2AQ8A = 2 THEN ALCFREQMO = 22;
ELSE IF S2AQ8A = 3 THEN ALCFREQMO = 14;
ELSE IF S2AQ8A = 4 THEN ALCFREQMO = 8;
ELSE IF S2AQ8A = 5 THEN ALCFREQMO = 4;
ELSE IF S2AQ8A = 6 THEN ALCFREQMO = 10;
ELSE IF S2AQ8A = 7 THEN ALCFREQMO = 1;
ELSE IF S2AQ8A = 8 THEN ALCFREQMO = 0.75;
ELSE IF S2AQ8A = 9 THEN ALCFREQMO = 0.375;
ELSE IF S2AQ8A = 10 THEN ALCFREQMO = 0.125;
ELSE IF S2AQ8A = 0 THEN ALCFREQMO = 0;
/*Estimate of #drinks consumed in a month — adapted from Professor´s suggestion*/
/*This can be considered somewhat similar to a quantitative variable*/
NUMALCMO_EST = ALCFREQMO * S2AQ8B;
/*Antisocial indicators are all categorial, so let´s just categorize the drinkers*/
/*How did I define these limits? Just by looking at frequency distribution. I chose to split the data in 3 almost equal parts*/
IF NUMALCMO_EST GE 0 AND NUMALCMO_EST LE 1 THEN ALCBYDAY_EST = 1; /*light drinker*/
ELSE IF NUMALCMO_EST GT 1 AND NUMALCMO_EST LT 16 THEN ALCBYDAY_EST = 2; /*medium drinker*/
ELSE IF NUMALCMO_EST GE 16 THEN ALCBYDAY_EST = 3; /*heavy drinker*/
/*Transforming unknown data into missing quantity i.e., How many drinks last year? When asked to someone who never drunk. */
IF S2AQ8A = . THEN S2AQ8A = 11;
IF S3AQ3B1 = . THEN S3AQ3B1 = 10;
PROC SORT; by IDNUM;
/*PROC FREQ; TABLES NUMALCMO_EST S1Q7A1 ALCBYDAY_EST S2CQ1;*/
PROC FREQ; TABLES ALCBYDAY NUMALCMO_EST S1Q7A1 S2CQ1 S3AQ3B1;
PROC UNIVARIATE; VAR NUMALCMO_EST ALCBYDAY_EST;
PROC ANOVA; CLASS S1Q7A1;
MODEL ALCBYDAY_EST = S1Q7A1;
/*How good it the usual drinking qqty of alcohol a good indicator of what is the largest qtty of drinks taken?*/
/*Now it is moderated by whether the individual did sought help due to drinking*/
PROC SORT; by S2CQ1;
PROC CORR; VAR NUMALCMO_EST S2AQ8C; by S2CQ1;
Then I decided to moderate last Correlation, that is, largest number of drinks as a function of frequency of drinking. Moderation variable was categorical S2CQ1 (“Ever sought help because of drinking?”).
For those who did sought help, r = 0.65, so the moderation did improved variability prediction (p still is < .0001).
If individual did not sought help, r = 0.60, moderation in this case has deteriorated but still significative (p still is < .0001).
Interesting results indeed!
Question: Is the usual amount of alcohol consumed a good indicator to the largest number of drinks a person takes?
Let´s analyse, as usual, people between 20 and 70 years.
Correlation between estimate amount of alcohol taken per month (explanatory quantitative variable NUMALCMO_EST - see previous post for clarification) and largest number of drinks in last 12 months (response quantitative variable) is 0.62 (p < .0001), suggesting that 0,3844 of the variance of largest number of drinks can be explained by amount of drinks taken per month.
That´s quite a finding, suggests that people who drink often also tend to drink more.