This study was part of a larger randomized controlled trial evaluating the effectiveness of a 14 session counseling intervention designed to increase medication adherence in a sample of Latino adolescents living in the San Diego–Tijuana border region. Peer counselors delivered the sessions during the first 4 months of a 9-month treatment. Contingency management and goal setting comprised the content of the intervention sessions. The control group completed the same number of sessions, but the content was unrelated to medication adherence. For the current study, baseline data from the trial regarding demographics, social environmental variables, and A&T use were used to: (1) Assess the association between neighborhood characteristics and distance to the nearest A&T retailer; and (2) Test the influence of variables from both the social and built environment on A&T use.
Screening, Recruitment, and Informed Consent
The sample of 225 Latino adolescents was recruited from seven high schools in south San Diego County. These schools were chosen because of their proximity to the United States (US)–Mexico border and were likely to result in greater likelihood for identifying students with latent tuberculosis infection (LTBI). Free tuberculosis (TB) screenings were offered at all seven schools. The medication adherence trial was presented to all students that had a positive tuberculin skin test. In order to be eligible for the trial, participants had to test positive for LTBI, have a residential address in the US, and plan to receive their LTBI treatment in the US. Data were collected between 2004 and 2005. After obtaining informed consents and assents, trained bilingual staff completed a baseline interview in the participant’s home.
The study sample was 50.4% female and 43.1% foreign-born. The participant mean age was 15.9 years (SD = 1.2; Range 13–19). Almost two-thirds of participants were bicultural (63.7%), 27.4% were non-acculturated or Hispanic, and 8.8% were assimilated.
The primary dependent variable was a created index reflecting lifetime use of both A&T. Dichotomous interview questions separately asked respondents if they had ever used A&T. These variables were summed to form an index ranging from 0 to 2, with higher scores indicating more drugs used (Cronbach’s Alpha = 0.76).
Independent variables included age, gender, acculturation, parental consistency, school truancy, and peer modeling of A&T. The Bidimensional Acculturation Scale for Hispanics  was used to measure acculturation. This standardized acculturation measure (Cronbach’s Alpha = 0.86 in current sample) consists of 24 questions primarily assessing language competence, including language use (e.g., How often do you speak English/Spanish?), linguistic proficiency (e.g., How well do you read in English/Spanish?), and electronic media use (e.g., How often do you listen to music in English/Spanish?). Each question had four possible responses: very poorly, poorly, well, or very well. The questions were separated into two domains, Hispanic (all items about Spanish usage) and non-Hispanic (all items about English usage), with 12 items in each. For each cultural domain, an average of the 12 items was calculated, obtaining a mean range of scores between 1 and 4. Scores on both domains were used to determine the level of acculturation. Acculturation categories were computed using a 2.5 cutoff score to indicate low or high level of adherence to each cultural domain. Individuals scoring higher than 2.5 in both domains were considered bicultural . Adolescents scoring high in the Hispanic domain, but low in the non-Hispanic domain were coded non-acculturated. Likewise, a high score in the non-Hispanic domain combined with a low score in the Hispanic domain resulted in a coding of assimilated.
Parents’ child management procedures were measured by adolescents’ perceived reports of the consistency of parents’ use of rewards and punishment. This scale included three questions about the level of consistent use of parental rules, punishment, and rewards (Cronbach’s α = 0.65). Individual item responses were based on a 5-point scale, 1 = very inconsistent to 5 = very consistent, resulting in a parental consistency scale ranging from 3 to 15.
A measure of school truancy was also constructed. Respondents were asked how many times in the last year they had skipped or ditched school. This variable was dichotomized (0 = never, 1 = ever) in order to correct a positive skew.
Peer modeling of A&T use was measured with the following two items: “How many of your close friends have ever used cigarettes?” and “…ever used alcohol?” These two variables were summed to create a peer modeling scale (Cronbach’s α = 0.75).
Two variables representing the built environment were utilized in this study, proximity of the school to the border and distance to the nearest A&T retailer from the participant’s home. Proximity of the school to the US/Mexico border distinguished between respondents enrolled at schools closest to the border, versus students from the remaining schools. All participating schools were within 10.5 miles of the border, but the three closest averaged 4.3 miles in distance, while the remaining four averaged 7.9 miles. This variable was included in analyses because of the high likelihood that those attending schools closest to the border also visit Mexico more frequently and are therefore exposed to less difficulty obtaining alcohol or tobacco [24, 25].
The participant residential address was geocoded in ArcView 9.2. Geocoding refers to the process of creating a point along a roadway segment that defines the location of any given address. A quarter-mile street network buffer was then created around each participant’s residential location or point. This buffer was intended to reflect the “walking neighborhood,” or those locations where the participant could easily walk to access nearby A&T retailers. Currently no standard exists to define a buffer size that appropriately reflects “neighborhood.” However, given the typically limited travel choices of adolescents, the area within a 5-min walk of his/her home can reasonably be considered a highly accessible area. The quarter-mile distance was developed assuming a walking speed of 3.4 miles/h . Buffers created using distances along the street network, such as that employed in the current study, exclude areas of the urban environment that are not accessible via roadways.
US Census Bureau data were obtained from San Diego Geographic Information Source (SanGIS), and used to identify neighborhood characteristics. Items representing neighborhood characteristics were selected using an adaptation of an approach employed by Sampson et al. [27, 28] and primarily represent indicators of neighborhood poverty. The values used in this study were: (1) percentage of families living below the poverty level; (2) percentage of unemployment; (3) percentage of adults (25 and older) with a high school diploma; (4) percentage of owner occupied homes; (5) percentage of the population under 18 years of age; (6) percentage of homes headed by a single mother; and (7) percentage of Hispanics.
The neighborhood characteristic variables from SanGIS were available by Census Block Groups (CBGs), a census geography that reflects aggregations of several Census Blocks. Since participant neighborhood buffers were irregular and did not fall exactly on the boundaries of the CBGs, it was necessary to estimate Census Bureau values within each participant’s buffer using a method referred to as “apportioning”. This procedure involves calculating the proportion of each CBG that overlaps with a neighborhood buffer and then using that percentage to factor each respective Census variable. For example, if a participant’s neighborhood buffer included 25% of one CBG, 55% of another, and 20% of a third CBG, then these percentages were used to weight the census values associated with each CBG to develop a unique value more closely aligned with the boundaries of the neighborhood buffer. This approach to adjusting Census data (available only in limited geographies) reflects a unique, non-census geography.
The San Diego County Department of Environmental Health, Food & Housing Division maintains a database of all county retailers that apply for food permits. This study analyzed retailers from the 2004 database. Stores that did not sell A&T were removed. The retailer address was geocoded using ArcView 9.2 and then used to measure the distance to the nearest retailer from each participant’s residential location. Distance to the nearest retailer was calculated using the Network Analyst function in ArcView, which is capable of finding and then measuring the distance of the shortest roadway path between a given participant’s residential point and the nearest retailer point. This variable demonstrated a non-normal distribution (positive skew) and required square root transformations to reach normality.
All statistical analyses were conducted using SPSS version 15.0. Pearson correlations were used to estimate the association between neighborhood indicators (community characteristics) and the distance from the participant’s home and the nearest retailer. Ordinal regression was used to explore the influence of the built environment on A&T use, while controlling for hypothesized social predictors. A sample size of 205 (regression sample) was used for regression analyses and excluded adolescent responses with missing data, whereas other analyses utilized the entire sample of n = 225.
Two interactions including distance to the nearest retailer were tested in the regression analyses: (1) parental consistency by distance to the nearest retailer; and (2) peer use of A&T by distance to the nearest retailer. The latter was not significant and therefore excluded from the final analysis.