Condomless sex and psychiatric comorbidity in the context of constrained survival choices: a longitudinal study among homeless and unstably housed women

Abstract

We sought to identify the prevalence and independent correlates of condomless sex within a cohort of community-recruited homeless and unstably-housed cisgender adult women who were followed biannually for three years (N=143 HIV+, N=139 HIV-). Nearly half (44%) of participants reported condomless sex in the 6 months before baseline, which increased to 65% throughout the study period. After adjusting for having a primary partner, longitudinal odds of condomless sex among women with HIV were significantly higher among those reporting <daily use of alcohol or cannabis (AOR=2.09, p=.002, and 1.88, p=.005, respectively) and PTSD (AOR=1.66, p=.034). Among women without HIV, adjusted longitudinal odds of condomless sex were significantly higher for those reporting <daily methamphetamine use (AOR=2.02, p=.012), panic attack (AOR=1.74, p=.029), and homelessness (AOR=1.67, p=.006). Associations were slightly attenuated when adjusting for sex exchange. Targeted HIV/STI programs for unstably housed women should address anxiety and trauma disorders, infrequent substance use, and housing challenges.

INTRODUCTION

The role of constrained life choices in health risks

Despite extensive services for HIV prevention and treatment, low-income women residing in cities with high costs of living continue to have significantly elevated risks for HIV and sexually transmitted infections (STIs) (1, 2). This elevated risk suggests that the presence of resources in well-resourced areas, such as health care, education, and behavior change programs, cannot be equated with “access” to those programs and may not equally benefit those lacking resources. Social and structural barriers that limit options and constrain choices have been posed as major contributors to elevated STI risk, along with conditions that are more common in impoverished populations such as mental illness and addiction (3, 4). Related to specific barriers, a lack of sufficient resources is a major predisposing factor for experiencing homelessness, which in turn places people at greater risk for both acquiring STIs and experiencing barriers in accessing appropriate medical care (5). Within this causal pathway, exchange of sex for money, food, drugs, shelter, or other unmet needs contributes to STI and HIV risk (6–8).

HIV and STI risk among unstably housed women

Much of the existing literature reports broad categories of poverty and mental illness as correlates of sexually transmitted infections (9–12). Our prior work, conducted exclusively among women with a history of housing instability (referred to from here on as “unstably housed women”), indicates disproportionately high levels of STIs and unique factors that predict STIs. For example, 88% of unstably housed women test positive for herpes simplex virus type 2 (compared to 21–24% of women in the general population), which differs in prevalence and correlates by HIV status and alcohol use (2). In addition, 12% test positive for trichomonas vaginalis (compared to 3% of women in the general population), which is associated with recent homelessness (1). Likewise, our previous work assessing mental health conditions and related issues also shows associations unique to this population of unstably housed women. For example, nearly every woman (97%) assesses positive for at least one psychiatric diagnosis, with a median count of 8 diagnoses and recent violence predicting lower mental health status. (13, 14) Additionally, both violence and unstable housing predict stimulant use among women who do not use at baseline, (14) which has been shown in other vulnerable populations to be associated with condomless sex (15, 16).

HIV and mental illness

People with severe mental illness (SMI) are disproportionally affected by and at risk for HIV and other STIs (17–22) and prevalence of HIV is significantly higher among individuals with SMI (23). SMI includes chronic recurring psychiatric disorders such as major depressive disorder, post-traumatic stress disorder (PTSD), schizophrenia, substance use disorders, and other conditions that result in substantial functional and cognitive impairment. People living with HIV/AIDS are also more likely to have mental health disorders, particularly depression, and HIV treatment outcomes are often worse for those with SMI (23–26). Previous research has found significant associations between alcohol use and HIV in settings characterized by limited resources to cover health care costs (27, 28). Research also shows significant associations between methamphetamine use and condomless sex among men who have sex with men (29), but less research has focused on risks of methamphetamine use among women (30). Similarly, less is known about how specific psychiatric comorbidities and frequency of substance use may affect health risks distinctly among women.

With regard to sexual risk behaviors that transmit STIs (namely unprotected or condomless sex), some studies find lower rates of sexual activity among those with SMI, but higher rates of risk behaviors among those who are sexually active (17, 18). However, other studies show that, rather than psychiatric symptoms alone, it is intimate partner violence, substance use, and constrained survival choices in the context of lower socioeconomic status that drive sexual risk behavior (18, 23). Nevertheless, few studies have examined these issues exclusively among low-income women, which would facilitate a better understanding of issues that are gender-specific and germane in low-income populations but rare in the general population.

Study objectives

We determined the degree to which factors related to poverty and mental health were independently associated with condomless sex over time within a cohort comprised entirely of community-recruited unstably housed women. We examined this behavior among all women enrolled in the study, not just among those who reported sexual activity, as abstinence has been promoted as a strategy to prevent transmission of HIV and STIs (31). Abstinence is also a practice with parallel implications to intentional social isolation, which has been reported as one strategy to prevent sexual victimization (4). Building on prior research, we hypothesized that the longitudinal odds of engaging in condomless sex would be higher among women with a greater number of mental health disorders, more frequent substance use, and recent homelessness.

METHODS

Data collection procedures

The analysis described here used longitudinal cohort data from the “Shelter, Health, and Drug Outcomes among Women” study (14). Three hundred women were recruited between 2008 and 2010 from free meal programs, homeless shelters, and a probability sample of single room occupancy (SRO) hotels in San Francisco, California, and completed study visits in private at a storefront research center in the Tenderloin neighborhood, where many participants lived. Participants were followed biannually for up to three years (for a total of 7 visits). Eligibility criteria included being a cisgender woman, at least 18 years old, and having a history of housing instability (slept on the street, in a homeless shelter, or in someone else’s place because there was no other place to sleep). Participants provided informed consent, which included being asked to restate their understanding of voluntary participation. HIV testing occurred during study screening and women with HIV were oversampled (50%) to meet HIV-specific study aims regarding violence and HIV risk behavior. Participants were reimbursed $15 for each interview completed and $5 for each monthly check-in visit to the storefront research center or via telephone during which participants updated contact information.

Questionnaires were administered by trained interviewers in a private setting. Drug use questions were asked using Audio Computer-Assistant Self-Interviews (ACASI). Given the lengthy time of assessment of psychiatric disorders via the Diagnostic Interview Schedule (DIS), these were assessed at a separate visit within one month of the baseline visit to reduce interview fatigue. Except for HIV status, all measures were self-reported. Study procedures were approved by the Institutional Review Board of the University of California, San Francisco.

Measures

All measures were time varying and assessed at each time point, except for race, highest level of education, and psychiatric diagnoses, which were only assessed at baseline.

Primary outcome: Condomless sex

The main outcome of interest was sex without a condom, assessed with the question: “Thinking about vaginal and anal sex only, were there times when you did not use a condom during the past 6 months?”

Primary exposures: Psychiatric comorbidity and substance use

Comorbid psychiatric conditions were assessed with the U.S. National Institute of Mental Health Diagnostic Interview Schedule (DIS) (32). This computerized tool (33) was used to assess the presence of up to 39 current mental health disorders using DSM-IV diagnostic criteria. In this analysis we focused on six mental health diagnoses (major depression, manic episode, PTSD, panic attack, generalized anxiety, schizophrenia) and diagnoses related to five substances (alcohol, opioid, amphetamine, cocaine, cannabis) that were present in at least 10% of this population and excluded phobias. Because the diagnosis of substance use disorder may not capture less frequent or episodic use, we also assessed the frequency of substance use. Three-level indicator variables (no use, less than daily use, daily use) were used to measure the average level of use over the past 6 months for multiple substances (alcohol, heroin, non-medical prescription opioid painkillers, methamphetamine, crack cocaine, cannabis) at each study visit.

Other exposures of interest: Constrained survival choices and social determinants of health

Other factors that have been associated with condomless sex and STI risk in previous research were considered as independent variables here, and included (within the past 6 months) having a primary intimate partner; experiencing violence; and reporting exchange of sex for money, drugs, housing, food, or protection (7, 14). We also examined social determinants of health including age, race/ethnicity, high school completion, and housing stability (i.e., slept on the street or in a shelter during the prior six months) and unmet subsistence needs in the previous 6 months as potential confounders of the relationship between psychiatric diagnoses and condomless sex. Primary partner was assessed as “one person who you feel closest or most committed to.” Experiences of physical, sexual, and emotional violence were assessed using questions based on the Severity of Violence against Women Scales (34) that had been previously tested in this population (14). Race and ethnicity were self-reported and grouped for analysis purposes into African-American, White/Caucasian, Asian/Pacific Islander, Hispanic/Latina, and Other/Mixed. Although all women in the study had a lifetime history of housing instability, women who reported sleeping in a shelter or public place in the past 6 months were considered to have recently experienced homelessness. Unmet subsistence needs included insufficient access to food, clothing, a restroom, a place to wash, or a place to sleep (35).

Data Analysis

Given a standing tradition of separating HIV prevention programs by serostatus (e.g., “prevention for positives” and prevention for negatives) (36), which is based on prior research showing differences in HIV risk behaviors by HIV serostatus (37), and given prior women-specific research showing differences in condomless sex by HIV status (37), as well as our own research showing differences in prevalence and correlates of herpes simplex virus type 2 by HIV status (2), we restricted all analyses by HIV status.

Descriptive statistics at baseline were examined with chi-squared and Wilcoxon rank-sum tests to determine differences in prevalence or median by HIV status, respectively. We used generalized estimating equation (GEE) logistic regression analyses to estimate population level associations between exposures of interest and condomless sex for women with and without HIV, accounting for repeated measures with an exchangeable correlation matrix. Analyses were performed in R 3.2.2 (38) using the geepack package (39).

GEE logistic regression analyses were conducted separately for women with and without HIV given different implications of condomless sex (e.g., transmission to others vs. acquiring HIV) and a significant interaction (p=.022) between HIV status and having a primary partner on condomless sex. Psychiatric diagnoses were assessed at baseline and the effect of diagnoses at both baseline and over time (by interacting diagnosis with time) were modeled. After estimating unadjusted GEE models with each of the independent variables listed above and condomless sex, we conducted adjusted GEE analyses with all variables significantly associated with the outcome in unadjusted models at p<.25 and removed variables by largest p-value until all p-values were less than .05 (40).

Sex exchange has been consistently associated with condomless sex in previous research (7, 41). Because sex exchange lies in the causal pathway between housing instability and condomless sex, we first assessed independent associations with psychiatric diagnoses and substance use without controlling for sex exchange using adjusted GEE logistic regression models as described above. We then compared these adjusted analyses to identical adjusted analyses but also controlling for sex exchange.

RESULTS

Participant Characteristics

Of 300 study participants, 9 did not complete psychiatric diagnosis assessments at baseline and an additional 8 had missing data on condom use throughout the study, leaving an analysis sample of 283. The median number of completed visits was 6 (IQR = 4–7). During the course of the study, 8.8% of this analysis sample were lost to follow up, 3.9% passed away, and 1.8% withdrew from the study. Among analysis sample participants, the median age was 48 (IQR = 42–53) and 69.8% were women of color (44.5% African American, 5.3% Latina, 3.5% Asian American/Pacific Islander, 19.1% Other/Mixed). While all women had a lifetime history of housing instability, almost half of participants (47.7%) reported sleeping in the street or a shelter in the past 6 months at baseline, which declined substantially to 14.7% in the final 36 month/7^th visit, as women experiencing chronic homelessness were more likely to be lost to follow up. Sex exchange was reported by 24.4% of women at baseline and by 27.9% of women during the study period, but also declined substantially to 12.0% at the final visit. At baseline, 55.8% of participants reported having a primary intimate partner, which remained consistent over time at the population level (52.9%−60.0% at follow up visits 2–7). [Table I]

Table I:

Social Determinants, Psychiatric Diagnoses, and Substance Use Frequency at Baseline Visit among Unstably Housed Women (N = 283)

	Overall		HIV+		HIV-
	N	%	N	% of 144	N	% of 139
Social Determinants & HIV Sexual Risk Factors (past 6 months)
Age (median, IQR)	48	(42, 53)	48	(42, 53)	48	(42, 54)
Race/Ethnicity +
Caucasian	87	30.7	39	27.1	46	33.1
African American	126	44.5	71	49.3	52	37.4
Latina	15	5.3	5	3.5	10	7.2
Asian American/Pacific Islander	10	3.5	2	1.4	8	5.8
Other/Mixed	54	19.1	27	18.8	23	16.5
Completed High School ***	184	65.0	77	53.5	107	77.0
Slept on street/in shelter1 ***	135	47.7	46	31.9	89	64.0
Unmet subsistence needs **	118	41.7	48	33.3	70	50.4
Condomless sex	123	43.5	57	39.6	66	47.5
Sex exchange	69	24.4	33	22.9	36	25.9
Any violence2 **	170	60.1	74	51.4	96	69.1
Primary partner	158	55.8	78	54.2	80	57.6
HIV+ ***	144	50.9	144	100.0	0	0.0
Psychiatric Diagnoses (past 2 weeks)
Number of Diagnosis (median, IQR)	8	(5, 11)	8	(5, 11)	7	(4, 10)
Major Depressive Disorder	185	65.4	89	61.8	98	70.5
Manic Episode	68	24.0	33	22.9	36	25.9
Post-Traumatic Stress Disorder (PTSD)	144	50.9	72	50.0	74	53.2
Panic Attack	103	36.4	53	36.8	50	36.0
Generalized Anxiety	32	11.3	13	9.0	19	13.7
Schizophrenia	55	19.4	25	17.4	30	21.6
Alcohol Use Disorder	178	62.9	93	64.6	85	61.2
Opioid Use Disorder	111	39.2	63	43.8	48	34.5
Amphetamine Use Disorder	106	37.5	57	39.6	49	35.3
Cocaine Use Disorder +	182	64.3	100	69.4	82	59.0
Cannabis Use Disorder	44	15.5	24	16.7	20	14.4
Substance Use Frequency (past 6 months)
Alcohol use: none	114	40.3	71	49.3	43	30.9
< daily *	122	43.1	52	36.1	70	50.4
daily	47	16.6	21	14.6	26	18.7
Heroin use: none	236	83.4	120	83.3	116	83.5
< daily	32	11.3	14	9.7	18	12.9
daily	15	5.3	10	6.9	5	3.6
Opioid painkiller use: none	210	74.2	111	77.1	99	71.2
< daily	58	20.5	24	16.7	34	24.5
daily	15	5.3	9	6.3	6	4.3
Methamphetamine use: none	255	90.1	117	81.3	108	77.7
< daily	53	18.7	23	16.0	30	21.6
daily	5	1.8	4	2.8	1	0.7
Crack cocaine use: none	159	56.2	83	57.6	76	54.7
< daily	85	30.0	38	26.4	47	33.8
daily	39	13.8	23	16.0	16	11.5
Cannabis use: none	139	49.1	70	48.6	69	49.6
< daily	97	34.3	46	31.9	51	36.7
daily	47	16.6	28	19.4	19	13.7

⁺p<.1,

^*p<.05,

^**p<.01,

^***p<.001

¹All women have a lifetime history of homelessness; this measure refers to the previous 6 months

²Includes physical, sexual, and emotional violence

In accordance with over-sampling of women with HIV in the original study design, half the women (N=144) had HIV. While women with HIV were less likely to have completed high school, they were also less likely to have experienced recent (past 6 months) violence, unmet subsistence needs, or homelessness at baseline.

Condomless sex was reported by 43.5% of participants at baseline and by 65.0% at any time point during the 3 year study period. Differences in the proportions of participants reporting condomless sex between women with HIV (39.6%) and women without HIV (47.5%) at baseline and at any time point (63.1% of women with HIV and 66.9% of women without HIV), did not differ significantly (chi-square p=.18 at baseline, p=.51 over time). Over the 3 year follow up period, however, the unadjusted longitudinal odds of reporting condomless sex was marginally lower (OR=0.70, p=.06) for women with HIV vs. women without HIV.

At baseline, women had a median of 8 (IQR: 5–11) psychiatric diagnoses (13, 14), with the most common mental health disorders being major depression (65.4%) and PTSD (50.9%). Cocaine (64.3%) and alcohol (62.9%) use disorders were the most common substance use disorders, and alcohol (59.7%) and crack cocaine (44.3%) were the most commonly used substances in the past 6 months. Less than daily use was more commonly reported than daily use for all substances. There were no significant differences in prevalence of current psychiatric diagnoses or substance use in women with HIV vs. women without HIV, with the exception that women without HIV were more likely to report less than daily alcohol use vs. no use (p < .05).

Unadjusted Associations with Condomless Sex

In bivariate analysis among women with HIV, PTSD was the only psychiatric diagnosis significantly associated with condomless sex (OR=1.72, p=.033). Among women without HIV, manic episode (OR=2.33, p=.002), panic attack (OR=1.98, p=.011), and alcohol use disorder (1.76, p=.047) were significantly associated with condomless sex. [Table II] No significant associations were found between groups with respect to condomless sex for number of diagnoses, major depression, generalized anxiety, schizophrenia, or other substance use disorders. (Interactions between baseline diagnoses and time were not significant, indicating that the decrease in proportion of women reporting condomless sex over time was not related to any specific diagnosis.)

Table II:

Unadjusted and Adjusted Longitudinal Correlates of Unprotected Sex among unstably housed women living with HIV (N=144) and without HIV (N=139)

	HIV+							HIV-
	Unadjusted			Adjusted (M1)		Adjusted - with sex exchange (M2)		Unadjusted		Adjusted (M3)		Adjusted – with sex exchange (M4)
	OR	p-value		AOR	p-value	AOR	p-value	OR	p-value	AOR	p-value	AOR	p-value
Social Determinants & HIV Sexual Risk Factors (past 6 months, all visits)
Age	0.93		<.001	0.94	<.001	0.94	<.001	0.96	0.008
Race/Ethnicity (ref: Caucasian)
African American	1.57		0.170					0.73	0.312
Latina	0.55		0.422					0.32	0.027
Asian/Pacific Islander	5.17		0.255					0.25	0.014
Other/Mixed	2.29		0.038					0.94	0.879
Completed High School	1.21		0.446					0.85	0.567
Slept on street/in shelter1	1.42		0.149					1.63	0.002	1.67	0.006	1.56	0.014
Unmet subsistence needs	1.11		0.594					1.13	0.437
Any violence2	1.30		0.106					1.12	0.492
Primary partner	2.77		<.001	2.76	<.001	2.73	<.001	5.96	<.001	5.88	<.001	6.25	<.001
Sex Exchange	2.81		<.001	--	--	2.73	<.001	3.00	<.001	--	--	3.13	<.001
Psychiatric Diagnoses (past 2 weeks, baseline visit)
Number of Diagnoses	1.00		0.968					1.05	0.121
Major Depression	0.90		0.686					1.24	0.464
Manic Episode	1.26		0.429					2.33	0.002
PTSD	1.72		0.033	1.66	0.034	1.58	0.056	1.42	0.191
Panic Attack	1.47		0.140					1.98	0.011	1.74	0.029	1.67	0.043
Generalized Anxiety	0.70		0.476					1.43	0.327
Schizophrenia	1.50		0.217					0.62	0.143
Alcohol Use Disorder	1.04		0.883					1.76	0.047
Opioid Use Disorder	0.68		0.141					0.90	0.704
Amphetamine Use Disorder	1.56		0.087					1.23	0.461
Cocaine Use Disorder	0.96		0.869					0.89	0.686
Cannabis Use Disorder	1.28		0.488					1.34	0.403
Substance Use Frequency (past 6 months, all visits), Ref: no use
Alcohol use, < daily	2.17		<.001	2.09	0.002	1.98	0.002	1.13	0.628
Alcohol use, daily	1.45		0.179	1.50	0.208	1.52	0.203	1.05	0.796
Heroin use, < daily	1.34		0.334					1.34	0.364
Heroin use, daily	1.22		0.617					1.26	0.726
Opioid painkiller use, < daily	1.28		0.234					1.37	0.152
Opioid painkiller use, daily	0.80		0.531					1.72	0.066
Methamphetamine use, < daily	1.42		0.219					1.95	0.009	2.02	0.012	1.56	0.013
Methamphetamine use, daily	1.73		0.274					0.97	0.954	0.76	0.696	0.73	0.650
Crack cocaine use, < daily	1.75		0.013					1.63	0.033
Crack cocaine use, daily	1.15		0.617					1.96	0.096
Cannabis use, < daily	2.10		<.001	1.88	0.005	1.73	0.012	1.17	0.487
Cannabis use, daily	1.47		0.118	1.02	0.937	1.01	0.944	1.28	0.420

Bold p<.05, Italics p<.1

¹All women have a lifetime history of homelessness; this measures refers to the previous 6 months

²Includes physical, sexual, and emotional violence

With respect to substance use frequency, when compared to no use, less than daily use of alcohol (OR=2.17, p<.001), crack cocaine (OR=1.75, p=.013), and cannabis (OR=2.10, p<.001) were associated with an increased odds of condomless sex over time among women with HIV, while less than daily use of crack cocaine (OR=1.95, p=.009) and methamphetamine (OR=1.63, p=.033) were associated with an increased odds of reporting condomless sex over time among women without HIV. Compared to no use, daily use of substances were not significantly associated with condomless sex in this population. [Table II]

In examining social determinants of health and known risk factors for condomless sex, there were some notable differences with respect to age and race/ethnicity: among women with HIV, younger women and mixed/other race women had higher odds of reporting condomless sex, whereas among women without HIV, younger, Latina, and Asian/Pacific Islander women had lower odds of reporting condomless sex. We also found that the odds of reporting condomless sex was almost three times higher among women both with and without HIV reporting sex exchange (OR=2.81 and OR=3.00, respectively, p<.001) and among women with HIV reporting a primary partner (OR=2.77, p<.001). These odds were six times higher for women without HIV reporting a primary partner (OR=5.96, p<.001). [Table II]

Independent Correlates of Condomless Sex

Adjusting for age and primary partner, the longitudinal odds of having condomless sex were significantly greater for women with HIV who reported less than daily of use of alcohol (AOR=2.09, p=.002) and cannabis (AOR=1.88, p=.005) vs. no use, and with PTSD (AOR=1.66, p=.034). [Table II] Also adjusting for primary partnership, the odds of having condomless sex were significantly greater for women without HIV who used methamphetamine less than daily compared to no use (AOR=2.02, p=.012), who experienced panic attacks (AOR=1.74, p=.029), and who recently slept on the street or in a shelter (AOR=1.67, p=.006). Independent associations were attenuated slightly after adjustment for sex exchange, but associations remained strong and significant. [Table II]

DISCUSSION

In a population previously reported to experience disproportionately high rates of herpes simplex virus-2 and Trichomonas vaginalis, (1, 2) two-thirds of homeless and unstably housed women reported condomless sex in a three-year period and this did not differ significantly by HIV status over time. Condomless sex was consistently associated with less than daily substance use relative to no use for alcohol and cannabis in women with HIV and for methamphetamine in women without HIV. Higher frequency use (vs. no use) and specific substance use disorder diagnoses did not reach conventional thresholds of significance. Even after adjusting for substance use and having a primary partner, PTSD was a significant correlate of condomless sex among women with HIV, while panic attack and recent homelessness were significant correlates of condomless sex among women without HIV. Finally, even after adjusting for sex exchange, the odds of condomless sex were greater among women without HIV experiencing homelessness compared to those with stable housing.

Results reported here are consistent with previous work with unstably housed San Francisco adults, which found that methamphetamine use was associated with sex exchange in women but not men (7). The results presented here extend this research by suggesting that women’s risk surprisingly lies in less frequent use. Moreover, this unexpected finding was also observed for use of alcohol and cannabis. The fact that daily use of these substances was not significant is noteworthy and has clinical implications. Less frequent substance use may present differently in a clinical setting than a substance use disorder, reinforcing the importance of substance use screening. Results suggest that including substance use assessment and counseling in interventions and risk reduction messages may nevertheless address this modifiable risk factor for STI transmission.

While much research has focused on depression as a predictor of HIV risk and treatment outcomes (23, 25, 42), less research has focused on anxiety or trauma-related disorders as distinct from depression in the context of condomless sex (43, 44). The few extant studies that have done so did not include women and did not highlight unstable housing (43, 45–47). Our findings show only slightly attenuated independent associations with panic attack or PTSD in the presence of sex exchange. While absolute differences in adjusted odds are less than 10%, they may suggest evidence of partial mediation of the relationship between these disorders and condomless sex through sex exchange. Alternatively, participation in sex exchange may lead to panic attack and PTSD symptoms that affect efficacy for condom use. Thus, low-income women faced with constrained survival choices also experience worry and fear that may impact their STI risk.

The finding of an association between PTSD and condomless sex over time among women with HIV indicates a need to address traumatic life experiences in this population. Chu et al. have proposed that traumatic events, especially in childhood, are linked to dissociative symptoms (as a coping mechanism) that can make it harder for adult women to assess or evaluate risk in dangerous situations (48, 49). Previous research with this cohort of unstably housed women found that 43% of participants have elevated levels of dissociation at baseline, which was significantly associated with recently experiencing physical or sexual violence (50). These results reiterate the calls for trauma-informed care in primary care, OB-GYN, psychiatric, and substance use disorder treatment settings (51–56), especially in an HIV treatment context (57). The Substance Abuse and Mental Health Services Administration (SAMHSA) has specified six principles of trauma-informed care: safety; trustworthiness and transparency; peer support; collaboration and mutuality; empowerment, voice, and choice; and cultural, historic, and gender issues (58).

While some symptoms of panic attack may be attributable to stimulant use, the DIS diagnostic assessment process excludes panic attack symptoms related to substance use. In addition, the association between panic attack and condomless sex persisted even after adjusting for stimulant use. The frequent victimization of unstably housed women (13, 14) as well as constrained choices due to limited housing and healthcare options for women without HIV may contribute to the increased risk for panic attack. In addition to addressing interpersonal violence and unmet subsistence needs, results suggest that promotion of in-the-moment stress reduction techniques, coping and coping-efficacy interventions (59), cognitive behavioral (60) and compassion focused therapy (61), and medication (62) for anxiety may potentially decrease condomless sex among impoverished women.

Principles of trauma-informed care may also be applied to settings where unstably housed women receive social services. For example, in supportive housing, this may include having flexible curfews and guest policies (63), as well as features that facilitate privacy (such as private rooms and locks) and maximizing choice whenever possible (64). At a homeless shelter this may mean providing a daytime refuge (but not forcing people to leave during the day) and at a free meal program this may mean having women-only mealtimes or providing meal delivery services. Women-only spaces (that include staff and management) in all service provision settings may facilitate both agency and social support that expand choices, safety, and recovery (63).

Our finding that women without HIV who experienced recent homelessness had greater longitudinal odds of condomless sex than housed women, even after adjusting for sex exchange and having a primary partner, provides additional support for the importance of housing in promoting wellbeing of impoverished women (3, 7, 65). The lack of such an association in women with HIV may be due to individual abstaining, serosorting, or other STI prevention behaviors. It may also be a reflection of the greater linkage to housing, healthcare, counseling, and case management (14) that HIV-infected individuals receive via the Housing Opportunities for Persons with AIDS (HOPWA) program and other programs (66–68), and that subsequently increase women’s choices and options. Related clinical implications include the importance of assessing and documenting patients’ homelessness or risk of becoming homeless (69), understanding the competing needs for survival that may negatively influence health behaviors, and being prepared to coordinate referrals to social services case managers and local legal and civil rights advocates (3).

The high prevalence of comdomless sex reinforces prior calls for heightened attention to STI prevention for unstably housed adults, including inclusion of homeless women as an HSV-2 high-risk population in testing guidelines (2), and integrating sexual health care into housing and support services (70). This may also include individual-level social approaches such as skill-based interventions to improve condom use (71) and structural interventions to improve educational and economic prospects in resource-poor communities (72). An additional implication of findings is that many women in this study would meet the American College of Obstetricians and Gynecologists and Centers for Disease Control and Prevention candidacy recommendations for pre-exposure prophylaxis (PrEP) (73, 74). This underscores the importance of improving PrEP uptake and education among impoverished women, particularly those facing housing challenges (75, 76).

Limitations

These findings should be considered in light of several limitations. A major limitation for HIV implications is that we do not have information on serosorting behavior by women either with or without HIV, so we do not know to what degree condomless sex was reported when women had the same HIV status as their partner. Another limitation is that all behaviors except HIV status are self-reported. As behaviors may be under-reported, our estimates of associations are likely at least as strong as reported here. A related consideration is that there may have been some social desirability bias for women with HIV to deny condomless sex. Additionally, we may have been underpowered to detect differences in unprotected sex when comparing daily substance use to no substance use. Finally, we urge caution in over-interpreting the impact of individual diagnoses, as they were not confirmed by more formal clinical assessment and there is a high level of comorbidity, which may have resulted in confounding. Additionally, the Diagnostic Interview Schedule used in assessment is reflective of the DSM-IV, and not the updated DSM-5, which categorizes PTSD as a trauma and stressor disorder rather than as an anxiety disorder (77). Nevertheless, notable strengths of this study include use of longitudinal data to determine correlates of condomless sex over time and a detailed assessment of mental health (rather than report of mental health services utilization).

Conclusions

Assessing and addressing infrequent substance use and anxiety- and trauma-related disorders may reduce subsequent condomless sex among homeless and unstably housed women. In addition, more targeted approaches are needed to address homelessness in women, particularly women without HIV who have fewer resources.