Demographic characteristics of the adult alcohol dependent sample are listed in . Most of the alcohol dependent sample was male (66%) and white (68%), older than 35 years of age, and had undergraduate school and graduate school (48%) as highest education level. The most prevalent illegal drug of use was marijuana (41%), followed by non-prescription use of opioid analgesics (23%), cocaine (19%), non-prescription use of sedatives and tranquilizers (12%), non-prescription use of stimulants (7%), ecstasy (5%) and other hallucinogens (5%).
Characteristics of Adult Alcohol Dependent Individuals (n=6,059), NSDUH 2005–2007
An identified latent class model indicated that a five class model was optimal. Fit statistics were computed for the 1 to 6 class models. The number of classes was chosen using theoretical reasoning, fit statistics and practical consideration. The fit statistics of AIC (28,405.2), BIC (28,666.8) and ABIC (28,542.9) were lower for the 5-class compared to the 4-class model (AIC=28,490.5; BIC=28,698.5; ABIC=28,600.0) and similar to the 6-class model (AIC=28,376.2; BIC=28,691.6; ABIC=28,542.2). The BIC was lowest for the 5 class model as opposed to the 6 class model suggesting that the minimum fit index value occurred for the 5 class solution (Nylund, 2007
). Furthermore, the ABIC value was not differentiated in the 5 and 6 class model. The parsimonious 5 class solution was chosen over the 6 class solution due to production of the minimum BIC fit statistic. In addition to picking the model with good fit statistics, the five class model was chosen over the four class model because the five class model differentiated between high concurrent drug users who had high probabilities of all illegal drug use except other hallucinogens and high concurrent drug users who had high probabilities of all illegal drugs except ecstasy and stimulants. The entropy value for the 5 class model was approximately, 0.7.
plots the past year illegal drug use on the x-axis and the probability of past year use of each drug for the five classes as indicated on the y-axis. The probabilities of endorsing past year illegal drug use are plotted for the seven illegal drugs for each of the five classes. The five class solution included a ‘no illegal drug use’ class (class 1) consisting of 65% of the adult alcohol dependent sample and included participants with near zero probabilities of past year illegal drug use. Class 2, consisting of 7% of the sample of adults with alcohol dependence, had medium probabilities of marijuana use, medium probabilities of non-prescription sedative/tranquilizer use and high probabilities of non-prescription analgesic use. Class 3, consisting of 21% of the sample, included individuals with high probabilities of marijuana use and medium probabilities of cocaine use. Two high concurrent-drug use classes emerged: Class 4 and Class 5. Class 4 included adults with alcohol dependence (6%) with high probabilities of marijuana, cocaine, non-prescription tranquilizers/sedatives and non-prescription analgesic use. Class 5 was the smallest class consisting of 1% of the sample and was characterized as adults with alcohol dependence with high probabilities of each illegal drug except other hallucinogens.
Conditional Probabilities of Adult Alcohol Dependent Participants Based on a 5-Class Latent Class Analysis (n=6,059)
demonstrates the sample characteristics of the adults with alcohol dependence by latent class, given the five class solution. Participants in Class 1 are less likely to report lifetime MDD, GAD, deviant behavior, having an STI and been incarcerated compared to the concurrent drug using classes. Also, participants in Classes 2, 4 and 5 were most likely to report having had lifetime MDD and GAD compared to participants in Class 1. Participants in Class 5 were most likely to report GAD (26%), MDD (40%) and deviant behavior (68%). Compared to the other concurrent-drug using classes participants in class one are less likely to have been in treatment for alcohol or drug use (27%). Although all individuals in the study met alcohol dependence criteria, individuals in class 1 were more likely to report meeting 3 criteria and were less likely to report meeting 5+ criteria compared to all other classes. Furthermore, the heaviest concurrent-drug use class, Class 5, was more likely to have been in treatment for alcohol or drug use compared to all other classes.
Characteristics of Adult Alcohol Dependent Individuals by Concurrent Drug Use Class, NSDUH 2005 to 2007
Results from the simple and multiple multinomial regression models are presented in which lists the OR and AOR given that class 1 or the ‘no past year illegal drug use’ class is the referent group. Simple multinomial regression models indicate that individuals with alcohol dependence reporting GAD during their lifetime had 1.9 (95% CI: 1.3, 2.7) the odds of being in class 4 compared to class 1. Although not always statistically significant, the odds of being in any of the concurrent drug using classes compared to class 1 ranged from 1.2 to 2.1 for participants reporting lifetime GAD. In alcohol dependent participants reporting a history of MDD the odds of being in class 4 were 1.6 (95% CI: 1.1, 2.4) and being in class 5 were 2.3 (95% CI: 1.2, 4.7) compared to class 1. Concurrent drug users (Classes 2 to 5) were also more likely to report deviant behavior, having ever had an STI, having been booked or incarcerated, meeting more (5+) alcohol dependence criteria and having been in treatment for an alcohol or drug problem compared to participants with practically no concurrent drug use in the past year (Class 1).
Crude Odds Ratios (OR), Adjusted Odds Ratio (AOR) and 95% Confidence Intervals (CI) of Adult Alcohol Dependent Individuals, NSDUH 2005, 2006, and 2007
Results of the multiple multinomial regression model demonstrated that adults with alcohol dependence who reported deviant behavior were more likely to be in the concurrent drug using classes (Classes 2, 3, 4 and 5) compared to class1. Alcohol dependent participants reporting a history of incarceration had 1.7 (95% CI: 1.2, 2.5) the odds being in class 4 compared to class 1. Alcohol dependent participants meeting 5 or more of the alcohol dependence criteria compared to 3 dependence criteria had odds ranging from 1.7 to 2.2 for being in the concurrent drug using classes compared to class 1. Reporting ever having been in treatment for an alcohol or drug problem was more likely for class 3 (AOR=1.7, 95% CI: 1.2, 2.3) and class 4 (AOR=1.7, 95% CI: 1.1, 2.7) compared to class1.