MACC is a longitudinal study measuring adolescent tobacco use patterns, perceptions, and attitudes. The state of Minnesota was divided into group-level units based on geographical and political (GPU) boundaries thought to reflect local tobacco control environments. These units were constructed based on the following four criteria: 1) the boundaries of the GPUs resemble existing geographic and/or political limits; 2) the boundaries of the GPUs reflect patterns of local tobacco program activities; 3) there is similar variation in youth smoking behavior across GPUs; and 4) a sufficient number of teens reside in each GPU to meet the sample size requirements. Using these criteria, Minnesota was first divided into 129 GPUs that included a mixture of counties, municipalities, school districts, urban neighborhoods and local planning districts. We then selected a stratified random sample of 60 GPUs. Of the 60 GPUs, 28 were rural (46%), 21 suburban (35%), 3 small city (5%), and 8 urban (14%). For more information about how the GPUs were selected see Chen et al. (under review). For this paper we hypothesized that youth living in the same GPUs would have a common tobacco access environment.
Five age cohorts were initially established (ages 12, 13, 14, 15, and 16) and approximately 60 participants, 12 in each age cohort, participated from each of 60 GPUs. The MACC study originally recruited a Minnesota cohort of 3,637 teens at baseline, approximately 725 each at ages 12, 13, 14, 15, and 16. Additionally, 587 12-year-olds were recruited one year after baseline. A combination of probability and quota sampling methods (to assure equal age distribution) was used to establish the six age cohorts. Surveys times were staggered such that surveys were administered to approximately one-sixth of the cohort members each month and then each youth was followed up at six month intervals. This meant that the baseline survey was administered between October 2000 and March 2001, the first follow-up survey was administered from April 2001 through September 2001, etc.
We used data from the first six MACC surveys (October 2000 – September 2003), restricting this sample to youth who were under age 18 at the time of the survey (100% of sample at survey 1, 72% of the sample at survey 6). At each of the 6 surveys only youth who reported smoking at least once in the previous month were included in this analysis. Our analytical sample ranged from 340 to 454 (). Subjects for this study were drawn from both the original baseline Minnesota cohort and the sample of 12-year-olds who were recruited one year after baseline. The new recruits were included to reduce the potential for age confounding and increase the sample size as the sample aged over time.
Characteristics of the previous month smoker youth sample at surveys 1 – 6 for the analytical sample and the whole Minnesota MACC cohort at baseline. Minnesota Adolescent Community Cohort, 2000-2003.
The survey was designed so that adolescents could answer potentially sensitive questions over the phone with non-revealing answers, e.g. a letter corresponding to a categorical response, in order to minimize bias that may result if a teen was concerned that a nearby parent or guardian may be listening to the phone conversation.
Smoking frequency, age, sex, and ethnicity were ascertained by youth self report. For ethnicity, adolescents were asked to indicate if they were: African American or Black, American Indian or Alaskan Native, Asian, Hispanic or Latino, White, or “Or something else?” (Interviewer recorded response). For these analyses we collapsed the categories of “Hispanic or Latino” and “Or something else” and labeled this category “other” due to small numbers. Using a series of survey questions all youth in the MACC study were classified as being at one of six smoking stages at each interview. The six MACC smoking stages were:
- Never = Never smoked a cigarette, not even a few puffs.
- Trier = Having smoked one whole cigarette or less (not including Never smokers).
- < Monthly = smoked a whole cigarette but not smoked on any of the last 30 days.
- Experimenter = Smoked less than 20 of the last 30 days and none of the last 7 days.
- Regular = smoked less than 20 cigarettes in the last 30 days and smoked at least once in the past 7 days.
- Established = Smoked more than 20 of the last 30 days.
For the purposes of this paper, we labeled those adolescents classified at stage 4, 5, or 6 as “monthly smokers”. “Weekly smokers” are those classified as a 5 or 6.
Social and commercial access measures were ascertained from the questions in . An individual would be coded as using a particular access method if he or she answered affirmatively to any of the questions relating to that type of source of cigarettes in the previous month. If this youth did not answer “yes” to any of the access method questions he/she would be coded as not having used that access method.
MACC Adolescent Phone Survey. Guide to how survey answers were used to classify access. Minnesota Adolescent Community Cohort 2000-2003.
The analyses allow for repeated measures, individuals nested within GPUs, and more than one source of random variation. We used a logistic model (using SAS PROC GLIMMIX (SAS Version 9.1)) to calculate probabilities and confidence intervals for several different tobacco access patterns reported in the month preceding the survey: those who used commercial access exclusively, those who used any commercial access, those who used social access exclusively, and those who used any social access. We also examined whether there was a longitudinal association between probability of social or commercial access method at the GPU-level and the aggregate smoking intensity in the GPUs.