Homelessness in Pregnancy and Increased Risk of Adverse Outcomes: A Retrospective Cohort Study

Abstract

Limited data indicates that homelessness during pregnancy is linked to adverse outcomes for both mothers and newborns, but there is an information gap surrounding pregnant individuals struggling with homelessness. In a landscape of increasing healthcare disparities, housing shortages and maternal mortality, information on this vulnerable population is fundamental to the creation of targeted interventions and outreach. The current study investigates homelessness as a risk factor for adverse obstetrical, neonatal, and postpartum outcomes. We reviewed more than 1000 deliveries over 1 year at a large public hospital in New York City, comparing homeless subjects to a group of age-matched, stably housed controls. Multiple outcomes were assessed regarding obstetrical, neonatal, and postpartum outcomes along with social stressors. Homeless pregnant individuals were more likely to experience numerous adverse outcomes, including cesarean delivery and preterm delivery. Their neonates were more likely to undergo an extended stay in the intensive care unit and evaluation by the Administration for Children’s Services, suggesting that they may be at an increased risk for family separation. After delivery, patients were less likely to exclusively breastfeed or return for their postpartum visit. Regarding personal history, they were more likely to endorse a history of violence or abuse, use illicit substances, and carry a psychiatric diagnosis. These findings indicate that homelessness is linked to numerous adverse obstetrical, neonatal, and postpartum outcomes that worsen health indices and exacerbate pre-existing disparities. Initiatives must focus on improved outreach and care delivery for homeless pregnant individuals.

Introduction

The urgency of the housing crisis in the United States cannot be overstated. In 2022, more than 580,000 people experienced homelessness on any given day, with rising numbers of pregnant individuals and their children [1]. Over the past 3 years, COVID-19-related affordable housing shortages have exacerbated an already dire situation [2, 3]. New York hosts up to 10% of the country’s homeless population, and by one estimate, one in every hundred babies born in New York City is brought home to a shelter [4]. Over the past two decades, experts in perinatal health have noted numerous difficulties faced by homeless pregnant individuals, often exacerbated by a lack of targeted healthcare services [5–8]. Furthermore, challenges faced by homeless families, such as psychiatric illness, hunger, and substance abuse, are linked to poor health outcomes for the entire family [9, 10]. Finally, pregnancy is associated with difficulties in exiting the shelter system [4] and foster care involvement [11] and may contribute to cyclic inequalities.

Despite these known disparities, pregnant homeless individuals remain largely invisible in medical literature. A recent systematic review noted an increase in adverse perinatal outcomes for homeless families, but identified a need for ongoing study [12]. Most prior studies focused on pediatric outcomes, specifically preterm birth and infant birthweight [6, 12], but less is known about maternal outcomes. Overall, there is very little data on pregnant individuals who are homeless during pregnancy and at delivery [5, 8, 13]. Furthermore, most existing studies relied on data that was extracted from large databases and birth certificates, which may be imprecise, and often lack clarification regarding definitions of homelessness [8, 12]. Further study is needed to adequately address the needs of this population. Our objective was to examine the perinatal outcomes for a cohort of clearly defined homeless patients compared to an age-matched group of stably housed controls. Different terms have been used to describe individuals who lack stable housing, but for this manuscript, we have chosen to use the world “homeless,” which encompasses a range of varying states of housing instability as defined by the US Department of Housing and Urban Development [14].

Methods

This was a retrospective 1:1 matched cohort study. Review was performed for charts of all patients ages 12–65 years who delivered at a large, academic, urban public hospital between January 1, 2017, and December 31, 2017. The age range was chosen to account for both adolescent pregnancies and pregnancies in older patients from artificial reproductive technology. Patients were excluded if their ages fell outside the stated range, or if information regarding their housing status could not be gathered from the electronic medical record (EMR). Institutional Review Board (IRB) approval was obtained.

Study Population

Electronic medical records of all deliveries during the dates of interest were reviewed to identify which patients were identified as homeless by the social work team. Our definition of homelessness is consistent with the specifications published by the US Department of Housing and Urban Development (HUD), which defines homelessness under four categories: literally homeless, or lacking fixed nighttime shelter; at risk of imminent loss of shelter; experiencing persistent housing instability; fleeing from, or attempting to flee from domestic violence [14]. We categorized patients as homeless when they were housed on the street, a shelter (including domestic violence shelters), or were living in temporary arrangements, for example, sleeping on the couch of a friend. Our study sample therefore includes individuals who were both homeless or at imminent risk of losing their housing at time of delivery. These homeless subjects were then age-matched to non-homeless controls. Multiple variables were examined with regard to prenatal care, obstetrical outcomes, postpartum care, and neonatal outcomes. Demographic and descriptive data was also collected.

At our institution, social workers perform extensive evaluations of every patient admitted for delivery. We classified a subject as homeless if the social worker’s note documented homelessness at any point during the pregnancy or at the time of delivery. Race and ethnicity, as expressed by the patient, were documented by the social worker. Obstetrical care was provided by resident trainees in obstetrics and gynecology, attending obstetrician-gynecologists, and certified nurse-midwives. Care for neonates was provided by resident trainees in pediatrics, attending pediatricians, physician assistants, and nurse practitioners. Documentation was reviewed from intrapartum admission(s), and prenatal and postpartum visits, if available.

Data Collection

Data was extracted from the electronic medical record by a single investigator. Study data were cataloged and managed using REDCap electronic data capture tools hosted at our institution [15].

Data Analysis

Data were analyzed using Stata v.14 (StataCorp, College Station, TX). Percentages were calculated using descriptive statistics. Statistical analysis was conducted using the chi-square test for categorical variables, two-sample t-test to compare means of normally distributed independent variables, and Wilcoxon-Mann–Whitney test to compare means of non-normally distributed independent variables.

Results

There were 1425 total deliveries at our institution between January 1, 2017, and December 31, 2017. Sixty-seven patients (4.7%) were identified as homeless based on chart review. These patients were compared to a group of 67 age-matched controls (4.9% of total housed population) drawn from the same time period, for a final cohort size of 134. Table 1 displays demographic information for homeless patients and controls.

Table 1.

Demographics

	Homeless n (%) (N = 67)	Housed n (%) (N = 67)	p
Age (years) (SD)	29.6 (6.7)	29.6 (6.7)
Racial/ethnic/nationality-based background*
White/Caucasian American	7 (10.4)	10 (14.9)	0.436
Black/African American/African	25 (37.3)	10 (14.9)	0.003
Asian	1 (1.5)	6 (8.9)	0.052
Native American	0	0
Indigenous/Pacific Islander	0	0
Hispanic/Latina	31 (46.2)	41 (61.2)	0.083
Middle Eastern	2 (2.9)	2 (2.9)	1.000
Mean number of living children (SD)	1.52 (1.32)	0.98 (1.25)	0.007
Mean body mass index (BMI) on admission (kg/m2) (SD)	30.6 (6.1)	30.1 (4.2)	0.670
Description of homelessness
Street	4 (6.0)
Homeless shelter	39 (58.2)
DV shelter	3 (4.5)
Sleeping on couch of relative or friend	8 (11.9)
Short-term apartment/housing	4 (6.0)
Transient (moving between unstable sites)	8 (11.9)
Other**	1 (1.5)

^*Race totals include patients who were identified as mixed or multiple races

^**Other was listed as located in temporary housing as refugee from hurricane in Puerto Rico

Information on specifics of homelessness can be seen in Table 1. The majority of homeless patients, 58.2%, were shelter-housed, while 5.9% were living on the street during pregnancy. Others, 11.9% (n = 8), were moving between housing sites, such shelter and street; this group was combined under “transient.” One patient’s housing site was unstable due to Hurricane Maria in Puerto Rico in 2017.

Obstetrical and neonatal outcomes are shown in Table 2. Homeless pregnant patients had a lower mean hematocrit on admission. Their delivery was more likely to be performed by a physician than by a certified nurse midwife (CNM) (83.6% vs 65.6%, p = 0.01). They were more likely to have a preterm delivery (23.9% vs 9.0%, p = 0.02), and their mean gestational age at delivery was lower, at 37.6 weeks gestational age vs 39 weeks for housed patients (p < 0.001). They were more likely to receive antenatal corticosteroids (19.4% vs 5.9%, p = 0.019). They were more likely to have a cesarean delivery (34.3% vs 16.4%, p = 0.017) and to have a positive test for group B streptococcus (23.8% vs 16.4%, p = 0.06). There were no differences in mean neonatal birthweight when controlling for gestational age, and homeless patients were not more likely to have an intrauterine fetal demise (IUFD) or receive a diagnosis of chorioamnionitis or endometritis.

Table 2.

Obstetrical and neonatal outcomes

Outcome	Homeless n (%) (N = 67)	Housed n (%) (N = 67)	p
Obstetrical
Mean weight on admission (kg) (SD)	80.3 (17.1)	76 (11.9)	0.174
Mean hematocrit on admit (g/dL) (SD)	34.5 (3.4)	36.1 (2.9)	0.002
Delivery performed by MD	56 (83.6)	44 (65.6)	0.010
Gestational age at delivery (weeks)	37.6	39	< 0.001
Preterm delivery (< 37 weeks)	16 (23.9)	6 (9.0)	0.020
Admission for spontaneous labor	25 (37.3)	36 (53.7)	0.056
Received antenatal corticosteroids	13 (19.4)	4 (5.9)	0.019
IUFD	1 (1.5)	1 (1.5)	1.000
Spontaneous vaginal delivery	41 (61.1)	52 (77.6)	0.058
Operative vaginal delivery	3 (4.5)	3 (4.5)	1.000
Cesarean delivery	23 (34.3)	11 (16.4)	0.017
GBS positive status	16 (23.8)	11 (16.4)	0.060
GBS negative, expired, or unknown	51 (76.1)	56 (83.6)
Diagnosis of chorioamnionitis	6 (9.0)	5 (7.5)	0.753
Diagnosis of endometritis	1 (1.5)	2 (3.0)	0.559
Neonatal
Mean neonatal birthweight (g)	2864	3174	0.334*
Apgars at 1 min (mean)	8.3	8.4	0.250
Apgars at 5 min (mean)	8.7	8.7	0.640
1 min Apgar ≤ 5	4	4	1.000
5 min Apgar ≤ 5	2	3	0.649
Neonate intubated	2 (3.0)	1 (1.5)	0.559
Neonatal death	1 (1.5)	2 (3.0)	0.559
Admission to NICU	16 (23.9)	15 (22.3)	0.838
Average days in NICU	27.3	7.5	0.007

^*Not significant when controlled for gestational age

Though not more likely to be admitted to the neonatal intensive care unit (NICU), neonates born to homeless parents who did go to the NICU had a longer stay, at a mean of 27.3 vs 7.5 days (p = 0.007). We found no association between housing status and Apgar scores at 1 and 5 min, neonatal intubation, or neonatal death.

Prenatal, postpartum, and social outcomes are documented in Table 3. Homeless patients had fewer prenatal visits and presented later to prenatal care. They were more likely to have fragmented prenatal care, which was defined as prenatal care partially or fully received at an outside facility or no prenatal care prior to presentation for delivery (58.2% vs 25.3%, p < 0.001). They were more likely to endorse substance use prior to pregnancy (34.3% vs 9.0%, p < 0.001) and during pregnancy (23.8% vs 1.5%, p < 0.001). They were more likely to carry a psychiatric diagnosis (49.2% vs 11.9%, p < 0.001); most common psychiatric diagnoses were depression (32.8% vs 5.9%, p < 0.001), anxiety (14.9% vs 0, p = 0.001), and bipolar disorder I/II (16.4% vs 1.5%, p = 0.002). Homeless patients were more likely to use mental health services during care (17.9% vs 5.9%, p = 0.03) and to have a psychiatry consult called during their delivery admission (11.9% vs 1.5%, p = 0.016).

Table 3.

Prenatal, social, and postpartum outcomes

Outcome	Homeless n (%) (N = 67)	Housed n (%) (N = 67)	p
Prenatal
Method of dating			0.274
LMP and 1st trimester ultrasound	19 (28.3)	22 (32.8)
1st trimester ultrasound	11 (16.4)	11 (16.4)
Ultrasound after 14 weeks	11 (16.4)	6 (9.0)
Dated on presentation for delivery	6 (9.0)	1 (1.5)
Number of prenatal visits (mean) (SD)	5.8 (3.1)	9.5 (3.1)	< 0.001
Fractured prenatal care*	39 (58.2)	17 (25.3)	< 0.001
Number of triage visits (mean)	0.98	0.8	0.790
Gestational age at 1st prenatal visit (weeks) (SD)	20.0 (SD 11)	12.6 (SD 8.5)	0.0012
Diagnosis of hypertensive disorder	21 (31.3)	17 (25.3)	0.427
Gestational diabetes	1 (1.5)	5 (7.5)	0.095
Pregestational diabetes	1 (1.5)	0	0.315
History of asthma	15. (22.3)	9 (13.4)	0.176
Diagnosis of HIV	2 (3.0)	0	0.315
Substance abuse during pregnancy	16 (23.8)	1 (1.5)	< 0.001
Substance abuse prior to pregnancy	23 (34.3)	6 (9.0)	< 0.001
Psychiatric diagnosis	33 (49.2)	8 (11.9)	< 0.001
Depression	22 (32.8)	4 (5.9)	< 0.001
Anxiety	10 (14.9)	0	0.001
PTSD	1 (1.5)	1 (1.5)	1.000
OCD	0	2 (3.0)	0.154
Personality disorder	0	1 (1.5)	0.315
Bipolar disorder (I/II)	11 (16.4)	1 (1.5)	0.002
Schizophrenia	1 (1.5)	0	0.315
Schizoaffective disorder	1 (1.5)	0	0.315
Psychosis NOS	1 (1.5)	0	0.315
Used mental health services during prenatal care	12 (17.9)	4 (5.9)	0.030
Psychiatry team called during admission	8 ( 11.9)	1 (1.5)	0.016
Admission to psychiatric unit during pregnancy	1 (1.5)	0	0.315
Social and postpartum
Involvement by Administration for Children’s Services	10 (14.9)	2 (3.0)	0.016
Hospital readmission within 6 weeks	4 (6.0)	0	0.248
Received immediate postpartum contraception	38 (56.7)	32 (47.8)	0.102
Attended postpartum visit	28 (41.7)	44 (65.6)	0.006
Received contraception at postpartum visit	7 (10.4)	20 (29.8)	0.030
Received postpartum contraception overall	45 (67.1)	52 (77.6)	0.176
Breastfeeding (exclusive)	19 (28.3)	36 (53.7)	0.004
History of emotional, physical, or sexual abuse	32 (47.7)	9 (13.4)	< 0.001
Custody
Likely to have other children	49 (73.1)	37 (55.2)	0.031
Does not have custody of other children	9 (13.4)	2 (3.0)	0.036
Children live with others (custody not specified)	9 (13.4)	3 (4.5)	0.082

^*This applies to any patient who received prenatal care partially or fully at an outside institution, or received no prenatal care

After delivery, homeless patients were more likely to have a case opened with the Administration for Children’s Services, 14.9% vs 3.0% (p = 0.015). They were less likely to attend their postpartum visit (41.7% vs 65.6%, p = 0.006) and were less likely to exclusively breastfeed within the first 6 weeks after delivery (28.3% vs 53.7%, p = 0.005). They were more likely to endorse a history of emotional, physical, or sexual abuse (47.7% vs 13.4%, p < 0.001). Both groups experienced similar rates of immediate postpartum contraception uptake and overall postpartum contraception uptake. Homeless individuals were more likely to have other children (73.1% vs 55.2%, p = 0.031) but less likely to have custody of their children (13.4% vs 3.0%, p = 0.036). No association was found between housing status and readmission within 6 weeks after delivery.

Discussion

Our study found that homelessness in pregnancy was associated with numerous adverse perinatal outcomes. Our results build on a limited number of pre-existing studies, providing necessary information on outcomes for homeless pregnant individuals while using standardized definitions of homelessness.

In our study, homelessness was linked to more interventions at or around time of delivery and higher risk elements of obstetric care, including increased rates of cesarean section and preterm delivery. These outcomes have long-term ramifications. Cesarean sections impact both postpartum healing and subsequent pregnancies, and surgery increases the likelihood of complications for homeless patients [16]. Preterm delivery is known to be associated with subsequent adverse outcomes for children [17, 18], and this is exacerbated for families without stable housing. Traveling to visit their hospitalized neonates places an additional burden on parents already overwhelmed by postpartum recovery. Furthermore, once discharged, these babies may need additional medical services, leading to logistical and financial stressors for economically disadvantaged families [18]. The increased likelihood of an evaluation by the Administration for Children’s Services places these families at further disadvantage, raising the possibility of family separation and the amplification of long-term inequities [19]. Overall, the neonates born to homeless individuals face outcomes that may place them at risk for lifelong negative health and psychosocial ramifications.

Participation in pre- and post-natal care provides continuity, education, and infrastructure for pregnant patients. Our homeless cohort presented later and less frequently to prenatal care and were more likely to miss their postpartum visit, which is consistent with prior studies and has been linked to poor obstetric and neonatal outcomes [5, 20]. Difficulty with postpartum visit adherence is a recognized issue in obstetrics, particularly for underserved groups who would especially benefit from the support and services provided by postpartum care [21]. Our cohort was also less likely to breastfeed. Lower breastfeeding rates reflect potential lost opportunities: the practice confers neonatal and maternal benefits, and it also has economic utility, particularly for new parents who lack safe food storage [22].

Homeless pregnant individuals suffer from a complex, interconnected web of psychosocial burdens including increased substance use, depression, and personal history of abuse, perpetuating a cycle of disenfranchisement [10]. They were four times more likely to engage in substance abuse prior to pregnancy and, staggeringly, sixteen times more likely to use illicit substances during pregnancy. Drug use in pregnancy places patients at risk for numerous poor health outcomes including preterm birth and low birthweight infants [23]. Our study population was more likely to be grappling with a psychiatric diagnosis or history of abuse: untreated psychiatric disease during pregnancy leads to difficulty with adherence to care along with increased substance use and risk of self-harm, and history of abuse is linked to myriad adverse health outcomes [23–25]. It is clear that homeless pregnant patients are facing enormous challenges based on psychosocial risk factors.

Homelessness in pregnancy should be recognized as an independent risk factor, as it compounds other, well-established challenges for struggling patients. It may be tempting to attribute our findings to other issues, such as poor prenatal care, mental illness, race/ethnicity, or comorbid medical conditions, but we propose that these risk factors are along the causal pathway and are not confounders [12]. The instability of homelessness itself makes it difficult to follow up with prenatal visits, antenatal testing, and other recommendations. Homeless patients may be labeled as “noncompliant” due to their challenges with follow-up or struggles with psychiatric disorders or substance abuse, introducing provider bias and eroding the patient-provider relationship.

Our study had several limitations. Our small sample size may have reduced the power of our study to detect differences between subjects and controls. We focused on a small, urban patient population that was predominantly Black and Hispanic/Latino and shelter-housed. In other parts of the United States, the homeless population may feature different demographics, which could limit the generalizability of our results. Information on long-term follow-up, level of education, and type of insurance were not available to us. We were not able to see records for hospitals other than our own, though many patients visited hospitals outside this network, which may have contributed to missing data in certain areas. However, this mimics the clinical reality of providers caring for patients in this setting. Some variables may have been unreliable and subjective, such as race, which was described by patients to social work staff. Patients’ exact words were not documented, which may have led to inaccurate collection of demographic information.

A strength of our study was the breadth of documentation available to us. All postpartum patients are screened by social workers at our institution using consistent documentation, allowing for multiple variables to be assessed across the entire sample in a standardized manner. Data was collected systematically using a reliable electronic medical record. Our database was tailored specifically to our study and was checked for accuracy and consistency multiple times throughout analysis. Our definition of homelessness included a lack of shelter along with individuals with unstable housing, in accordance with the HUD definition. This ensured that we were able to incorporate data for any patients within our sample who experienced this condition.

There has been very little research on perinatal outcomes for homeless pregnant individuals, highlighting their invisibility within healthcare. Exploring outcomes for vulnerable groups allows us to evaluate population health and redirect funding, thereby reducing the societal costs of healthcare disparities. By understanding the challenges experienced by homeless families, policymakers and public health initiatives can focus on their specific needs.

The current prenatal care delivery model requires patients to present at frequent, scheduled intervals, which is rendered difficult by a lack of housing. Patients may be struggling with getting their children to school (sometimes far from the location of their shelter), following up with requirements of social services (such as welfare, Administration for Children’s Services, and family court), and grappling with their own psychiatric or medical needs. Given these challenges, homeless pregnant patients may find it easier to present to emergency settings, such as labor and delivery triage. These visits are often limited to addressing immediate concerns and not providing the prenatal care services and counseling that were missed in the outpatient setting. If homeless pregnant patients also suffer from mental illness or comorbidities, this heightens the urgency of establishing longitudinal follow-up. Pregnancy presents a critical moment to change patients’ perceptions of the healthcare system and form important bonds to assist in postpartum challenges [5]. Thus, our findings should prompt a larger discussion of how to best provide care for individuals with housing instability for whom the current prenatal care structure may be insufficient.

Some potential solutions have been documented, such as targeted outreach programs [10]. In 2018, Ake et al. performed a qualitative needs assessment of shelter-housed pregnant women in Milwaukee, WI [7]. Their team found that patients strongly desired education on topics such as pre-eclampsia and postpartum depression, but needed support accessing care. Prenatal healthcare providers have a unique role to play as they intersect with pregnant patients at a key moment for intervention and education.

Future research should examine more effective ways to deliver care to this population and potentially integrate this care with other services, such as social support, mental health services, and substance abuse treatment. Without targeting the challenges faced by homeless pregnant people, we cannot make progress in advancing patient-centered care for the most vulnerable, thereby improving societal health outcomes for this generation and those to come. In a medical landscape plagued by inequality, the issues facing homeless pregnant individuals and their children are urgent and must be addressed.