Table 3.

Relative Risk From Marginal Structural Cox Proportional Hazard Regression^a for All-Cause, Drug-Related, and HIV-Related Mortality Rates by Incarceration/Homelessness Trajectory Groups Among 15,620 Adults Who Had Been Both Incarcerated and Homeless, New York City, 2004–2005

Trajectory	All-Cause Death		Drug-Related Death		HIV-Related Deathb
	RR	95% CI	RR	95% CIc	RR	95% CI
Temporary	1.91	1.00, 3.68	7.80	1.07, 56.86	3.32	1.14, 9.67
Transition to incarceration	1.21	0.49, 2.98	3.52	0.32, 38.75	3.41	0.95, 12.28
Transition from incarceration	1.99	0.88, 4.53	9.86	1.00, 97.30
Transition to homelessness	1.27	0.58, 2.78	7.58	0.90, 63.60	1.00	Referent
Transition from homelessness	1.88	0.86, 4.14	8.90	1.04, 76.22
Continuously homeless	1.00	Referent	1.00	Referent

Abbreviations: CI, confidence interval; HIV, human immunodeficiency virus; RR, relative risk.

^a We used inverse probability of treatment weight to control for bias due to confounding. Potential confounders included age, sex, race/ethnicity, nativity, neighborhood poverty level, a proxy measure of substance use, a proxy measure of mental illness, and criminal charges for drug sales, violent crimes, weapon possession, public administration, property crimes, quality-of-life crimes, and sex crimes.

^b Because there were 0 deaths attributed to HIV in the continuously homeless pattern, 3 trajectories of homelessness (transition to homelessness, transition from homelessness, and continuously homeless) and 2 trajectories of incarceration (transition to incarceration and transition from incarceration) were collapsed into homelessness and incarceration groups, respectively.

^c Wide confidence intervals reflect a small number of outcome cases.