Table 2.

Mixed Effects Generalized Linear Models Estimating the Adjusted Effect of Intervention on Number of Self-Reported SSTIs, Total Number of Injections, Number of Uncleaned Skin Injections, and The Relative Frequency of Injections without Skin Cleaning.^a

Model Outcome	IRR/exp(b)b	95% CIc	z (p =)c
# Self Reported SSTIsd	0.65	(0.35; 1.22)	−1.34 (.179)
Total # of Injectionsd	0.61	(0.36; 1.02)	−1.90 (.058)
# of Uncleaned Skin Injectionsd	0.34	(0.20; 0.60)	−3.78 (<.001)
Relative Frequency of Skin Cleaning when IDUe	1.82	(0.77; 4.39)	1.36 (.173)

^aCovariates included in all models were indicator variables for month of assessment, number of infections reported in the year prior to baseline, years age, gender, race (White), ethnicity Latinx) and # of nights spent on the street or in a shelter during 3-months prior to baseline.

^bIRR reported for # self-reported SSTIs, total # of injections, # of uncleaned skin injections; exp(b) reported for relative frequency of skin cleaning when IDU

^cConfidence interval estimates and tests of significance were based on the robust Huber-White variance estimator.

^dModel specified with log link function and negative binomial error distribution. To adjust for time at risk the models also included log(number of days at risk during each assessment) as an offset with coefficient fixed to 1. The n was 213 persons observed on 669 occasions.

^eOperationally defined as ((times cleaned skin / total injections) * 100). Model specified a logit link function with binomial family error distribution. The n was 182 persons observed on 460 occasions.).