Table 3B.
Zero-inflated negative binomial regression for the count of risk exposures in the past 30 days for housed and homeless individuals.
| Outcomes | Variable | Coefficient (SE) | Odds | t | p-value |
|---|---|---|---|---|---|
| Number of times used crack, coke or heroin in last 30 days | |||||
| Zero count logistic model | Experienced violence in the place I live | −0.89 (0.44) | 0.41 | −2.01 | .044 |
| Likelihood of living in the same place in 3 months | −0.22 (0.10) | 0.79 | −2.21 | .027 | |
| Any positive reasons for moving | −0.75 (0.28) | 0.47 | −2.71 | .007 | |
| Living on the street/in shelter | −1.36 (0.39) | 0.25 | −3.48 | <.001 | |
| Living with family, sex partner of friend | −1.10 (0.29) | 0.33 | −3.80 | <.001 | |
| Longest homeless period (months) | −0.02 (0.01) | 0.98 | −2.55 | .011 | |
| Positive count negative binomial model | Male | 0.54 (0.18) | 2.96 | .003 | |
| Age (years) | −0.03 (0.01) | −2.57 | .010 | ||
| Total number of sex partners in last 30 days1 | |||||
| Positive count negative binomial model | Age (years) | −0.01 (0.01) | −0.63 | .53 | |
| Diagnosed with mental illness | 0.52 (0.84) | 0.62 | .53 | ||
| Feel safe in the place I live | −0.15 (0.12) | −1.26 | .21 | ||
| Likelihood of living in the same place in 3 months | 0.19 (0.06) | 3.31 | <.001 | ||
| Longest prison sentence | 0.04 (0.01) | 2.53 | .011 | ||
| Living with family, sex partner of friend | −0.02 (0.21) | −0.09 | .93 | ||
| Diagnosed with mental illness * Age | −0.05 (0.02) | −3.05 | .002 | ||
| Diagnosed with mental illness * Feel safe in the place I live | 0.43 (0.14) | 3.05 | .002 | ||
| Longest prison sentence * Feel safe in the place I live | −0.01 (.004) | −2.63 | .008 | ||
| Longest prison sentence * Living with family, sex partner of friend | 0.03 (0.01) | 2.82 | .005 | ||
| Number of different people having sex with without condoms in last 30 days 2 | |||||
| Zero count logistic model | Age (years) | 0.21 (0.07) | 1.24 | 3.31 | <.001 |
| Less than HS education | 1.71 (0.82) | 5.53 | 2.09 | .037 | |
| Ever told by a doctor that I have HIV/AIDS | 2.82 (0.89) | 16.84 | 3.16 | .002 | |
| Number of moves in the past 6 months | 1.12 (0.44) | 3.06 | 2.54 | .011 |
Outcome variables: Number of times smoked crack, or sniffed/injected coke, or sniffed/injected heroin in the last 30 days (Fitted model: Likelihood Ratio Chi-square test=156.0, df=8; p=<.0001; Schwarz Bayesian Criterion(SBC)=2556); Total number of sex partners in last 30 days (Fitted model: Likelihood Ratio Chi-square test=114.64, df=10; p<.0001; SBC=962); Number of different people having sex with without condoms in the last 30 days (Fitted: Likelihood Ratio Chi-square test=53.31, df=4; p<.0001; SBC=557).
SE = Standard Error of coefficient estimate.
Odds = Exponentiated coefficients of the logistic model for zero inflation represent the odds of a zero count.
Number of cases =374 (excludes 18 living in SRO or half-way houses).
Housing status groups: Living on street or in shelter; living with family, sex partner of friend; living in own apartment with a rental subsidy or supportive housing; living in own apartment without a rental subsidy (reference).
Predictors considered at beginning of the model: age (years); black race (y/n); male gender (y/n); less than HS education (y/n); diagnosed with mental illness (y/n); told by doctor that I have HIV/AIDS (y/n); feel safe in the place I live (scale); perceived gang activity in neighborhood (scale); experienced violence in the place I live (y/n); likelihood of living at the same place in 3 months (scale); number of moves in past 6 months; number of positive reasons for moving; number of negative reasons for moving; longest prison sentence (months); longest homeless period (months); living on the street or in a shelter (y/n); living with family, sex partner of friend (y/n); living in own apartment with a rental subsidy or in supportive housing (y/n).
A zero countlogistic regression component was not fit because there is no evidence of an excess in number of zero counts for the number of sex partners in last 30 days
There is an insufficient number of individuals who report having condomless sex with more than one partner in the last 30 days to enable fitting a negative binomial regression model to the count.