International Journal of Drug Policy - 2014

Volume 25 Issue 3 May 2014

Issue link: http://digitalreprints.elsevier.com/i/364061

Contents of this Issue

Navigation

Page 117 of 153

618 C.E. Sterk et al. / International Journal of Drug Policy 25 (2014) 616–623 Methods Data The data for this paper were collected for People and Places, a large-scale community-based cross-sectional study. The study was designed to achieve a better understanding of multiple levels of influences on health and health-related behaviors. In the present study, because the focus is on recent crack cocaine use, only those who reported having used crack at least once in the 90 days prior to the interview were included. Data were collected from 461 cur- rently using respondents recruited through active (e.g., community outreach or street intercept methods where people are approached in their neighbourhoods and informed about the study) and passive (e.g., posting flyers) methods between May 2009 and March 2012. The study area includes 70 neighbourhoods (census block groups). Eligibility criteria included self-identification as African Amer- ican, black, or African, being 18 years or older, and having lived in the same census block group within the study area for at least 12 months prior to the interview. The added eligibility criteria for recent crack cocaine use required respondents to have used the drug at least once in the last 90 days. Non-probability quota sampling was employed to ensure sufficient variation in gender and age among the respondents by census block group with the goal of having the ability to make comparisons rather than to pro- vide a representative sample. The Institutional Review Board at Emory University approved informed consent procedures. Inter- views were conducted in a private room at a centrally located research site in one of the study neighbourhoods by trained inter- viewers using computer-assisted technology. The survey included domains such as demographic characteristics, psychosocial mea- sures, health history, alcohol and drug use history, criminal justice involvement, and neighbourhood perceptions. At the completion of the interview, each respondent was paid $30 for participating in the study as well as offered referrals to local health/social service agencies. Dependent variable Frequency of crack cocaine use was operationalized as the num- ber of days a respondent used the drug during the 90 days preceding the interview. Micro-system Socio-demographic characteristics Age at first use and Age were measured in years. Gender was coded as male (0) and female (1). Education ranged from "no schooling" (0) to "doctorate or equivalent degree" (23). Income was measured in dollars per month. This variable was skewed and a square root transformation was applied prior to analysis in order to have it conform to a more normal distribution. Relationship status was measured as not having a partner (0) versus being partnered (1). Stability of housing situation was categorized as unstable hous- ing (e.g., homeless, temporary housing, or staying in someone else's house) (0) versus stable housing (e.g., renting or owning) (1). Neigh- bourhood 5 + years reflected if a respondent had been living in the neighbourhood less than five years (0) or for five or more years (1). Exo-system Perceived neighbourhood disorder Perceived neighbourhood disorder was measured using 8 items from the Ross and Mirowsky (1999) neighbourhood disorder scale. Items were summed such that higher scores corresponded to greater perception of neighbourhood disorder and ranged from 2 to 32. Internal consistency as measured by Cronbach's alpha reflected good reliability at 0.81. Meso-system Crack cocaine use-related practices Involvement in drug distribution was measured with the question "During the past year, did you sell, distribute, or help to make illegal drugs?" and was coded as no (0) or (1) yes. Having traded sex was measured with the question "During the past year, did you trade sex for food, drugs, or money?" and was coded as no (0) or yes (1). Social context of crack cocaine use Setting of use was assessed with the question "During the past 90 days, at what places have you used crack?" Responses were used to generate the measure as any (semi-)public places (including in a crack house, in a shooting gallery/get off house, in an abandoned building, in a car, outdoors and elsewhere) (0) and private residence (including only at home, the home of relatives and the home of friends) (1). People with who use occurred was assessed with the question "During the past 90 days, with what other people have you used crack?" Responses were used to generate the measure as alone only (0), ever with individuals with who the respondent has per- sonal connections (including a sex partner, one or more relatives or one or more friends) and alone but not with acquaintances (1) and ever with acquaintances (including acquaintances, co-workers, neighbors, a drug dealer or strangers) (2). Analyses Descriptive statistics were computed for the independent and dependent variables. For all inferential statistics, adjustments were made to parameters and standard errors to account for clustering using generalized estimating equations (GEE) in IBM SPSS Statistics 20. GEE is used to adjust standard errors to account for the clus- tered nature of data collected by census block group. Associations between each independent variable and the outcome were tested with regression models using a negative binomial distribution and a log link. Negative binomial regression was used to account for the over dispersion of the dependent variable (Hilbe, 2008). These regressions model the log of the expected number of days of crack cocaine use in the last 90 days. An exponentiation of the coefficients yields an incident rate ratio (IRR) (Hilbe, 2008), or an estimate of the percentage change (IRR − 1) in the outcome for a one-unit change in each independent variable (Long, 1997). The multivariable regression models of the effect of perceived neighbourhood disorder adjusted for socio-demographics (Model 1) and the full model of perceived neighbourhood disorder, crack cocaine use-related practices and the social context of crack cocaine use (Model 2) also employed a negative binomial distribution and log link and included all predictors that were significant at the level of p < .05 in bivariate regressions. Regressions were used to test for the mediation of perceived neighbourhood disorder by crack cocaine use-related practices and the social context of crack cocaine use using the criteria outlined by Baron and Kenny (1986). Evidence of mediation was determined if perceived neighbour- hood disorder was associated with the indicator of crack cocaine use-related practices and the social context of crack use in analy- ses controlling for socio-demographics, perceived neighbourhood disorder was associated with frequency of crack use controlling for socio-demographics and if the effect of neighbourhood disor- der was reduced to non-significance when the indicator of crack cocaine use-related practices and the social context of crack use was included in the model. Regressions with dichotomous out- comes employed a binomial distribution with a logit link and the regression predicting people with who use occurred employed a

Articles in this issue

view archives of International Journal of Drug Policy - 2014 - Volume 25 Issue 3 May 2014