Sample
This study used a large sample (n = 8837) of Danish citizens aged 15 to 65 from a Danish national survey conducted in 2011 (55, 56). Participants from two rural and two urban Danish municipalities (i.e., Holstebro, Lolland, Copenhagen and Aarhus) were randomly selected by the Danish Civil Registration System by Simple Random Sample. Persons younger than 19 years were oversampled for this survey, as youths were a group of particular interest for the overall research project. Of the sample, 33.4% were aged 15–18 years, 19.9% were aged 19–30 years, and 46.8% were aged 31–65 years.
Measures
The survey consisted of a battery of self-report questionnaires designed to assess demographic characteristics, alcohol and drug use patterns.
Substance Use Indicators
To construct latent profiles of substance use, nine dichotomous indicators of alcohol and drug use in the past year were created: 1) low alcohol consumption past year, 2) regular alcohol consumption past year, 3) drunk past month, 4) cannabis lifetime, 5) cannabis month, 6) stimulants lifetime, 7) opioids lifetime, 8) other drugs lifetime, 9) illegal drugs past month. The two indicators of alcohol use and the two indicators of cannabis use were constructed as mutually exclusive variables in order to prevent nesting (57):
Indicators of alcohol use were created by combining two questionnaire items “How often have you used alcohol during the past year?” and “how many times have you been drunk or very drunk within the past month?”. Respondents who reported using alcohol “a few times a year”, “once” or “not at all” during the past year, were placed in the low alcohol consumption category. Respondents who reported “4 to 5 times a week”, “2 to 3 times a week”, “once a week”, “2 to 3 times a month”, or “once a month”, but did not report being drunk within the past month, were placed in the “regular alcohol consumption” category. Respondents who reported being drunk one time or more within the past month were placed in the “drunk past month” category.
Indicators of cannabis use were created based on two measures: “Have you ever used Cannabis“ (0 = no and 1 = yes) and “have you used cannabis within the past month?” (0 = no and 1 = yes). Respondents who reported cannabis use in their lifetime, but not within the past month, were placed in the “lifetime Cannabis use” category, and respondents who reported cannabis use within the past month were placed in the “past month cannabis” category.
An indicator of lifetime use of stimulants was created by using the item “have you ever used party drugs such as amphetamine, cocaine, or ecstasy?”. Respondents who reported any use of stimulants were placed in the “lifetime stimulants use” category.
An indicator of lifetime use of opioids was created by using the item “have you ever used heroine, opium, methadone or similar drugs?”. Respondents who reported use in this item were placed in the “opioids life” category.
An indicator of lifetime use of other drugs was created by combining the three items “have you ever used mushrooms, LSD or similar?”, “have you ever sniffed glue, gasoline, lighter fluids, or other solvents?” and “have you ever used other drugs such as fantasy, khat, or ketamine”. Respondents who reported lifetime use of any of these other drugs were placed in the “lifetime use of other drugs” category.
An indicator of use of other drugs than cannabis within the past year was created by combining five items on substance use in the past year “Within the past year have you used”: 1) party drugs such as amphetamine, cocaine, or ecstasy, 2) heroine, opium, methadone or similar drugs, 3) mushrooms, LSD or similar?, 4) glue, gasoline, lighter fluids, or other solvents?, and 5) other drugs such as fantasy, khat, or ketamine”. Respondents who reported any use within the past year of these other drugs were placed in the “other drugs year” category.
Register Based Outcomes: Criminal Convictions And Substance-related Healthcare
Denmark has multiple national registers with individual-level data on the entire population, covering population, cohabitation, marital status, healthcare use, criminal justice contacts, treatment for substance use disorders, sources of income, and a range of other information. These register data can be linked to survey participants via a unique personal identification number (58). For the present study, different registers were used to obtain the two outcomes of interest (1) criminal convictions, and (2) substance-related healthcare. For each outcome, survey respondents were tracked using registers until the event had occurred, or death, whichever occurred first. The maximal follow-up time was seven years.
The variable operationalizing contact with the criminal justice system was based on the Danish Central Crime Register (59). Events of interest were convictions concerning all penal code violations from 2011 to 2018, including driving under the influence, the Euphoriants Act, and the Firearms Act. Speeding violations and minor traffic offenses, such as missing bicycle lights, were omitted from the analyses due to their relative normalization in the Danish society.
For Substance-related healthcare, registered healthcare from 2011 to 2018 were extracted from four different registers. Firstly, SIB and NAB, which are the official registers of people enrolled in treatment for substance use disorders. Secondly, these registers were supplemented with data from the National Patient Register and its separate database of psychiatric treatment. For somatic treatment the following fully alcohol or drug use-related ICD-10 codes were selected (B18.2, E24.4, G31.2, G62.1, G72.1, I42.6, K29.2, K70.0-K70.4, K70.9, K86.0, O35.4, P04.3, P04.4, P96.1, Q86.0, R78.0, T40.0-T40.5, T51.0, T51.1, T51.9, X45, X65, Y15, Z71.4, Z71.5). For psychiatric treatment all F1 ICD-10 codes (mental and behavioral disorders due to psychoactive substance use) were used.
Sociodemographic Measures (Covariates)
In order to control and test for potential confounders and important influential factors, sociodemographic variables were included in the analyses (i.e., sex, municipality, level of education, and immigrant status). Sex was coded as a binary variable (male, female). The municipality variable was coded as a categorical variable corresponding to the four included municipalities (Aarhus, Copenhagen, Holstebro, Lolland). Highest level of education was coded as a categorical variable into 1) primary school or less, 2) secondary education, 3) short-term higher education, 4) medium or long-term higher education, and other (not defined). Information on immigrant status was retrieved from the population register and coded as a binary variable (immigrant and descendants of immigrants = 1, dane = 0)
Procedures
In 2011, a total of 13,157 Danish citizens aged 15 to 60 years were invited by e-mail to participate in a study on substance use and related factors. The survey was available via a link in the e-mail sent to potential study participants. In order to increase the response rate, individuals who did not respond after receiving e-mails were contacted by phone by trained employees from statistics Denmark. In total, 8,837 individuals completed the questionnaire (67.2% response rate). Of these, 6,557 respondents (74.2%) filled out the questionnaire online and 2,280 (25.8%) responded to the questionnaire via phone interviews.
Data Analyses
Descriptive statistics were conducted in IBM SPSS v. 26. The latent class analyses (LCA) were conducted using Mplus 8.4 software, employing maximum likelihood estimation with robust standard errors (MLR). First three LCAs were performed to determine the number of heterogeneous groups with homogeneity within each group based on nine substance use indicators across the three age groups (i.e., 15 to 18 years, 19 to 30 years, and 31 to 65 years). LCA estimates the posterior probabilities of class membership or size of the class (57). Better fitting models are reflected by significant p values for the Lo-Mendell-Rubins likelihood ratio test (LMR), and the Bootstrap likelihood ratio test (BLRT), lower values on the Akaike Information Criteria (AIC), the Bayesian Information Criteria (BIC), and the sample size adjusted BIC (SSABIC), and higher entropy values indicate clearer classification (57). Furthermore, model fit and the resultant class solution should be judged based on substantive meaningfulness of the classes, i.e., the classes should be distinct and meaningful (57).
After obtaining the latent classes, class membership was converted into an observed variable and chi-square tests were conducted to examine the associations between class membership and legal convictions and receiving treatment for substance use or for a psychiatric disorder related to substance use. Across the three age groups, univariate Cox regressions were performed to examine the associations between the substance use class memberships and two outcomes namely, criminal convictions and substance-related healthcare. Next, multivariate Cox regression analyses were performed using the survival package in R (R version 4.2.1, package version 3.3-1) to evaluate the associations between class membership (class 1 as the reference group), and the two outcomes. These regressions were adjusted for age, sex, education, municipality, psychiatric diagnosis at the time of survey participation, and immigrant status.