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 18 of 153

A.N. Martinez et al. / International Journal of Drug Policy 25 (2014) 516–524 519 Table 1 Characteristics of unduplicated UHS study participants between 2004 and 2005 (N = 1084) by activity space variables. Total sample N = 1084 Activity space distance N = 989 SEP Accessibility N = 96 Percentage Mean, standard deviation Percentage Health-related outcomes HIV seropositivity 12% 0.87 (2.4) 8.5% Syringe sharing in past 6 months 35% 1.8 (2.9) 8.8% Non-fatal overdose in the past 12 months 9% 2.3 (3.6) 10.4% Individual-level variables Male 75% 1.6 (2.5) 10.1% Female 25% 1.5 (2.7) 7.7% Age Under 30 5% 1.9 (3.0) 15.4% 30–49 62% 1.6 (2.7) 9.4% 50 and over 33% 1.3 (2.3) 9.3% Considers self homeless 59% 1.6 (2.6) 7.9% Race/ethnicity Black 40% 1.2 (1.9) 13.1% White 42% 1.5 (2.6) 8.4% Latino 8% 2.8 (3.4) 6.0% Other 10% 1.4 (2.0) 4.7% Received government assistance in past 30 days 33% 1.4 (2.2) Illegal source of income in past 30 days 36% 1.7 (2.7) 7.7% Traded sex for drugs or cash in past 6 mo 16% 1.4 (2.0) 9.6% SEP use in past 6 months No use in past 6 months 12% 1.9 (2.5) 11.7% Less than once a week 41% 1.5 (2.5) 10.2% Once a week or more 47% 1.5 (2.7) 9.3% Arrested in past 6 months 30% 1.8 (2.9) 8.6% Residential transience Number of locations slept in past 6 mo 3.6 (4.1) 2.0 (3.0) 2.4 (2.7) Slept in >3 locations in past 6 months 31% 7.9% Drug-related variables Smoking crack cocaine in past 6 mo 71% 1.6 (2.6) 8.8% Heroin injection in past 6 months 71% 1.7 (2.7) 9.2% Methamphetamine injection in past 6 months 32% 1.2 (2.1) 8.7% Injection drugs <10 years 15% 1.6 (2.4) 9.6% Census tract-level variables Concentrated poverty 61% 1.1 (2.1) 9.5% buffer of 50 meters is used around each activity space route. Data for locations of all SEPs during the same time period (2004–2005) were identified previously for another analysis (Wenger et al., 2011). Seventeen unique SEP locations were mapped during the two-year time period. ArcGIS created a dichotomous variable with a '1' denot- ing if a participants' route intersects with any SEP location. Census tract-level measures Poverty level is the only Census tract variable included in the sta- tistical analysis of all three outcomes. Poverty level in the United States is defined as a set of money income thresholds that vary by family size and composition to determine who is poor. In 2000, a family of four, with two related children under the age of 18, will count as poor if the total family income is less than $17,463 (cite: http://www.census.gov/prod/2001pubs/p60-214.pdf). We dichotomized the poverty threshold into low and high categories. Census tracts with more than 20% of households reporting income below the federal poverty level are classified as high poverty and tracts with less than 20% of households reporting income below the poverty level are classified as low poverty. Concentrated poverty is defined in the literature as a range of 40–20% of Census tract households living in poverty (Alexassensoh, 1995; Lichter, Parisi, & Taquino, 2012). The mean and median percentage of households in poverty in San Francisco is 22% and 21%, with a maximum of 52%. Because only 0.8% of the tracts (n = 7) met the definition of concen- trated poverty, we used 20% as a threshold that results in enough variation to compare the tracts in each category. The dichotomous measure of concentrated poverty was linked to the individual-level data using the Census tract where participants usually sleep. Statistical analysis Depending on the level of variable measurement, we calcu- lated frequencies, means and standard deviations, medians and interquartile ranges, as appropriate. Logistic regression models were estimated to assess the magnitude of the association between activity space variables and the three outcomes of HIV seropositi- vity, syringe sharing, and overdose. First we fit logistic regression models to estimate the crude association between activity space variables and individual level and Census tract level covariates. Second we fit three logistic regression models to estimate the association between activity space variables (distance and SEP accessibility) and outcomes of HIV serostatus, syringe sharing, and overdose after adjusting for all covariates. Covariates selected for model building were based on P values below .10. We fit each model using methods of backward elimination. The parameter estimates of each outcome in multivariate analysis are associated with a one- unit change in the activity space distance variable. Each estimated coefficient is the expected change in the log odds of an outcome for a unit increase in the corresponding continuous independent variable holding the other covariates constant. No efforts were made to categorize activity space distance into groups due to the exploratory nature of this measure and its skewed distribution. Given the multiple levels of data included in the multivari- ate analysis, potential clustering of participants at the level of

Articles in this issue

Links on this page

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