Abstract
Objective. Previous studies of homeless children have described more health problems and service use than in housed children, but failed to control for potential confounding factors that may differ between these children. This observational study examines the relationship of homelessness and other determinants to health status and service use patterns in 627 homeless and low-income housed children.
Methods. Case-control study of 293 homeless and 334 low-income housed children aged 3 months to 17 years and their mothers conducted in Worcester, Massachusetts. Information was collected about mothers' housing history, income, education, emotional distress, and victimization history. Standardized instruments were administered to assess children's health. Health service use questions were adapted from national surveys. Main outcome measures included health status, acute illness morbidity, emergency department and outpatient medical visits. Multivariable regression analyses were used to examine the association of family and environmental determinants, including homelessness, with health status and service use outcomes.
Results. Mothers of homeless children were more likely to report their children as being in fair or poor health compared with their housed counterparts. Homeless children were reported to experience a higher number of acute illness symptoms, including fever, ear infection, diarrhea, and asthma. Emergency department and outpatient medical visits were higher among the homeless group. After controlling for potential explanatory factors, homeless children remained more likely to experience fair or poor health status (adjusted odds ratio [OR] = 2.83; 95% confidence interval [CI], 1.16, 4.87), and a higher frequency of outpatient (OR = 1.71; 95% CI, 1.18, 2.48) and emergency department visits (OR = 1.21; 95% CI, 0.83, 1.74). Mothers' emotional distress was independently associated with acute illness symptoms and frequent use of outpatient and emergency department settings.
Conclusions. Homelessness is an independent predictor of poor health status and high service use among children. The present findings highlight the importance of preventive interventions and efforts to increase access to primary care among homeless children.
Homelessness, with its associated adverse health consequences, continues to be an important national problem.1 Families with children now comprise more than one third of the overall homeless population, with estimates of >500 000 children on the streets each year.2,,3 Researchers have documented the health needs of homeless children, including delays in routine screening and immunizations,4–9 high rates of acute and chronic illness,4,,6,9,10 nutritional problems,4,,6,1014–16 impaired access to primary health care,4,,7,8,11 and high rates of emergency room use and hospitalization.4,,6,8,11
In general, homeless children have been shown to suffer from more health problems and use medical care services to a greater extent than their housed counterparts. These studies have primarily been descriptive in nature and limited by small sample sizes,8,,9nonrepresentative samples,5,,7,8,13 or lack of appropriate comparison groups.4,,7,9,11,13,14 With the exception of a study that focused on growth delay in homeless children,15previous studies of homeless children have failed to control for potentially confounding sociodemographic, environmental, or maternal factors that may differ between homeless and housed children, which could explain poorer health outcomes or high service use patterns between these groups. The extent to which homelessness independently predicts adverse health outcomes or high health care service use patterns is primarily unknown.
As part of a comprehensive epidemiologic study of homeless families and children in Worcester, Massachusetts, we attempted to move beyond previous descriptive reports of homeless children's health. The objectives of the present study were to compare the health status and service use patterns of 293 homeless and 334 low-income children, aged 3 months to 17 years, and determine the extent to which homelessness and other potential explanatory factors contribute to poor health outcomes and high service use rates.
METHODS
Sample
The Worcester Family Research Project is a case-control study of 220 homeless and 216 low-income, housed families and their 627 dependent children aged 3 months to 17 years residing in Worcester, Massachusetts17 between August 1992 and July 1995. Homeless mothers were recruited from 9 shelters and 2 welfare hotels serving homeless families.17 During the course of the study, 361 families were approached for enrollment, 102 refused to participate and another 39 failed to complete all four interviews. No significant differences were found between women who completed the study and those who refused with respect to race and number of children. Homeless women who refused to participate were slightly younger than those who completed the study (24.2 years vs 26.1 years). The homeless women who dropped out of the study before completing all interviews were similar to study participants in terms of race, age, and number of children.
Housed families who had never been homeless were recruited from the Worcester Department of Public Welfare office. A screening form was used to assess the housing history of all women meeting with their caseworkers. Women without a history of previous homelessness were asked to participate in the study. Of the women approached, 141 were disqualified for previous homelessness, 148 refused to participate, and 31 dropped out before completing the study. Housed women who refused to participate were similar to the study sample in terms of age and number of children. They were slightly more likely to be Puerto Rican (49% vs 36%). Compared with the study sample, women who failed to complete all interview sessions were similar to study participants in age, race, and number of children.
Children between the ages of 3 months and 17 years who were living with their mother were eligible for assessment. Depending on family size, 1 child was randomly selected from each of three age groups: infants and toddlers (0–2.4 years; n = 218); preschool children (2.5–5.9 years; n = 180); and school-aged children (6 years and older; n = 228). Because fewer families had school-aged children, this group was oversampled, with up to 2 children enrolled from each family.
Worcester is a mid-sized city with a population of ∼169 000.18 Unlike most cities this size, Worcester has a large Puerto Rican population and a small black population, which was reflected in the study sample.18 Like most cities in the Northeast, the overwhelming majority of homeless families are headed by women.19 As a result, only female-headed families were enrolled into the study; two-parent homeless families were excluded because of very low numbers in the community.
Data Collection
Informed consent was obtained from each enrolled mother to interview all family members. Data were collected from both mothers and children using structured interviews conducted by trained interviewers. All interviews were conducted separately for mothers and children in the respondent's choice of Spanish or English. For mothers, interviews took place during 3 to 4 sessions and covered information such as childhood and adult life events, support networks, histories of violent victimization, and mental and physical health status. In addition, mothers completed an interview about the child's background, health, life events, support network, and service use. Information about children's health conditions and acute illness symptoms was gathered from approximately equal proportions of the homeless and housed during the winter months of December through March (30% vs 33%, respectively). As an incentive to participate, mothers received vouchers redeemable for merchandise at local stores and children received an age-appropriate toy or voucher. Child interviews consisted of developmentally appropriate assessments of cognitive and emotional well-being.
Instruments
All instruments were chosen on the basis of their reliability, validity, and past use with low-income populations. Additional consideration was given to ease of administration in shelters and previous use with Latino populations. All interview protocols were translated into Spanish by bilingual and bicultural translators. Wherever possible, preexisting Spanish versions of both mother and child instruments were used.
Assessment of Mothers
Demographic information, including housing history, income, education, jobs, family size and structure, and service use was gathered using a modified version of the Personal History Form.20 This instrument was designed for use with homeless and low-income housed persons.
Current mental health symptomatology and distress were assessed using the Symptom Checklist 90 (SCL-90).21 This self-report instrument provides a current profile across nine symptom dimensions: somatization, obsessive-compulsive, interpersonal sensitivity, depression, anxiety, hostility, phobic anxiety, paranoid ideation, and psychoticism. The Global Severity Index is a summary score that combines a symptom count with a measure of distress to produce an overall rating.
Physical and sexual abuse across the lifespan was assessed using the Conflict Tactics Scale.22 Violent acts were contextualized by the addition of questions on severity and duration of abuse. Support networks were assessed using the Personal Assessment of Social Support Scale.23
Assessment of Child Health and Life Events
Dimensions of child health status were measured including health perception, accidents, acute illness morbidity, medical conditions, and disability. Mothers' perceptions of the child's health status and self-reports of accidents, medical conditions, and disability days were measured by questions adapted from the National Health Interview Survey, Child Health Supplement.24 Acute illness symptoms were adapted from the Rand Health Insurance Experiments' Child Health Questionnaire.25 We asked about the occurrence of the following symptoms: fever; ear infection or earache; sore throat with fever; diarrhea or cough with fever of at least 3 days duration; broken bone; accidental poisoning; head injury; seizure; poor eating habits; vomiting; bronchitis, wheezing, or asthma attack; abdominal pain; headache; and trouble sleeping. Health service use questions were drawn from the National Health Interview Survey.24
Stressors experienced in the past year by each child >2.5 years old were assessed with the Masten's Life Event questionnaire.26 Mothers completed this 39-item instrument that consists of discrete negative events (eg, “During the past year, at least one parent became seriously ill or was injured”) or chronic strains (eg, “Frequent arguments between adults in household”). We created a count of 12 severe negative discrete events that were out of the child's ability to control. Positive responses are summed to give a cumulative stress score. Mothers were also asked about the child's history of foster care placement, physical or sexual abuse, and number of moves in the past year and during the child's lifetime.
Data Analysis
Differences between homeless and housed children were assessed using the t test for continuous variables and the χ2 test for categorical variables. Two-tailed tests of significance were used for all analyses. Logistic regression analyses were used to examine the relationship between housing status and child health outcomes, while controlling for potentially confounding variables. The outcomes under study included emergency department visits in the past year (0–1 vs 2 or more); ambulatory care visits in the past year (0–2 vs 3 or more, including well-child and sick visits); mother's perception of their child's health status as fair or poor versus good to excellent; and number of acute illness symptoms reported in the past month (0–1 vs 2 or more). The multivariate models developed for each outcome included important background factors. Each model contained covariates for age, ethnicity, family size, and housing status. Other predictors were included on the basis of a priori postulated associations with the outcomes under study as well as on the basis of univariate associations with the principal study endpoints (P < .15). Variables included in these analyses were: history of physical or sexual abuse; foster care placement; number of moves in the past year; mother's age, education, income; size of mother's support network; history of battering and emotional distress. All nominated predictors competed for entry using a stepwise algorithm. Variables that were eliminated from the generated regression models were individually reentered to check for negative confounding. Additionally, health service outcomes controlled for children's health status and mother's social support, whereas symptom and health status outcomes included the medical condition count as an additional controlling variable. Adjusted odds ratios and accompanying 95% confidence intervals were calculated. All analyses were initially run in SAS (version 6.11) statistical software27 and then rerun in SUDAAN (version 7.0) statistical software28 to account for correlated data as a result of sibling pairs within the sample.
RESULTS
In total, 293 homeless and 334 low-income housed children were assessed. Table 1 compares the characteristics of homeless and housed children and their mothers. Homeless children were significantly younger and were more likely to have moved in the past year. A significantly greater proportion of homeless children had ever been in foster care or been investigated by social service agencies for potential neglect. Although not significant, the number of serious stressful life events experienced in the past year was also higher among homeless children (1.4 vs 0.6). The median duration of homelessness was 8.7 weeks.
Characteristics of Sheltered Homeless and Low-income Housed Children and Their Mothers in Worcester, Massachusetts
Although the majority of both homeless and housed mothers were receiving Aid to Families With Dependent Children (72.3% vs 93%;P < .001), homeless mothers had significantly lower annual income and lower educational attainment. Less than 5% of both groups were currently working. Homeless mothers were more likely to report using alcohol or drugs during their pregnancy. Although similar proportions (20%) of both homeless and housed mothers began their prenatal care after the first trimester, more than half of the women in both groups reported current cigarette smoking. Homeless mothers reported having significantly fewer nonprofessional sources of social support than housed mothers. Although homeless mothers, compared with the housed, were more likely to report that they had experienced violence by an intimate partner during their lifetime, rates of abuse were high in both groups. Homeless mothers were more emotionally distressed on the SCL-90 than their housed counterparts.
In examining health characteristics and service use patterns of homeless and housed children (Table 2), the mothers of homeless children were significantly more likely to report their children as being in fair or poor health. Almost half of the homeless mothers compared with one quarter of the housed reported that they had significant worries about their child's health in the past 3 months. The number of accidents requiring medical care during the past year was equivalent between the groups.
Health Characteristics and Service Use Patterns of Children by Housing Status in Worcester, Massachusetts
Approximately equal numbers of medical conditions during the past year were observed between the homeless and housed children with one-third of mothers in both groups reporting that their children had two or more conditions (Table 2 and Fig 1). The mothers of homeless children were significantly more likely to report that their children had ear infections than housed children. Despite the lack of differences observed according to housing status, high absolute rates of childhood asthma were reported. Of the homeless and housed children with asthma, more than half had mothers who reported current smoking.
Prevalence of selected conditions in the past year according to housing status.
Half of the homeless children compared with approximately one third of the housed children were reported to have two or more acute illness symptoms during the past month (Table 2). The mothers of homeless children were more likely to report that their children had experienced fevers, ear infections, diarrhea, and bronchitis or asthma during the past month (Fig 2). A comparable proportion of homeless and housed children (15% vs 13%, respectively) had stayed in bed more than a half day on at least one occasion during the last 2 weeks because of illness or injury.
Prevalence of selected symptoms in the past month according to housing status.
Health service use was higher among the homeless (Table 2). More than one third of the homeless children reported two or more emergency department visits during the past year compared with only one fifth of the housed children. Homeless children also had higher outpatient visit rates for well and sick care. Homeless children were more likely to have been hospitalized in the past year compared with the housed children.
In examining access to health services and the use of well-child care, only small differences were seen between the respective comparison groups. More than 99% of both the homeless and housed children had medical insurance coverage, primarily through Medicaid. Although the homeless children, compared with the housed, were significantly less likely to have a regular health provider (93% vs 98%;P < .05), or to have up to date immunizations as reported by their mother (95% vs 98%; P < .05), absolute rates were high for both groups. Among children under the age of 2, significantly more of the homeless compared with the housed had failed to receive any well-child care (8.3% vs 1.2%;P < .05) in the first 2 months of life. Among children aged 9 months to 6 years, only two-thirds of the homeless compared with four-fifths of the housed, had ever been screened for lead poisoning (67.2% vs 80.6%; P < .001). A comparable proportion of both groups of children aged 3 and older had never seen a dentist (homeless vs housed, 15.1% vs 11.2%).
A multivariable regression analysis was performed to examine the association between homeless status (Table 3), as well as other key variables identified in univariate analyses, and service use patterns and health outcomes (Table 4). For the two health service use outcomes examined, homeless as compared with housed children were significantly more likely to be seen in the emergency department or outpatient setting on multiple occasions after controlling for previously described covariates (Table 3). Homeless mothers were more than twofold more likely to report that their children were in fair or poor health. Albeit not statistically significant, the mothers of homeless children were more likely to report the presence of multiple acute illness symptoms during the past month than were mothers of housed children (Table 3).
Crude and Multivariable Adjusted Odds Ratios of Selected Service Use and Health Characteristics According to Housing Status*
Crude and Multivariable Adjusted Odds Ratios of Service Use and Health Outcomes According to Selected Characteristics
With regard to other factors associated with service use patterns (Table 4), the child's age (younger children more likely to use emergency department and ambulatory care clinics on a more frequent basis), and mothers' emotional distress as measured by the SCL-90 were significantly associated with more frequent use of the emergency department and ambulatory care clinics. Puerto Rican children and children reported by their mothers to be in fair or poor health were also significantly more likely to use the emergency department two or more times in the past year after adjusting for additional covariates.
Other factors in addition to previously noted homeless status were also associated with health outcomes (Table 4). An increasing number of medical conditions was significantly associated with fair or poor health status as was larger family size. Mothers' emotional distress was independently associated with more frequent acute illness symptoms in children as was the presence of a greater number of reported conditions.
CONCLUSION
In summary, our findings highlight the adverse impact of homelessness on children's health. Homeless children must struggle not only with the health consequences of being poor, but the added burden of residential instability. Recent changes in welfare legislation threaten to increase the number of children who will experience homelessness.47 Although prevention of homelessness represents the only appropriate long-term solution, community health and outreach efforts must target homeless children as a group with high vulnerability for health problems, particularly, asthma and infectious diseases. Aggressive screening for and effective treatment of health problems, along with preventive measures, offer the possibility of improving homeless children's health status and well-being.
The data about high emergency department use among homeless children suggests that barriers to primary health care services exist despite extensive outreach efforts in communities nationwide. Further efforts to improve homeless families' access to ongoing primary care through provision of transportation and increased flexibility of health service delivery agencies will be important in this era of cost containment and managed care. Additional studies that look more closely at the precipitants and nature of homeless children's health care visits are needed to more clearly understand the observed high usage patterns.
ACKNOWLEDGMENTS
This work was supported by Grant MH47312 and Grant MH51479 from the National Institute of Mental Health and Grant MCJ250809 from the Maternal and Child Health Bureau.
We would like to acknowledge the important contributions of the following persons in the conduct of this research: project interviewer staff, Nancy Popp, EdD; Meg Brooks; Angela Browne, PhD; John Buckner, PhD; and Amy Salomon, PhD.
Footnotes
- Received August 25, 1997.
- Accepted February 27, 1998.
Reprint requests to (L.W.) Department of Family and Community Medicine, University of Massachusetts Medical Center, 55 Lake Ave North, Worcester, MA 01655-0309.
- SCL-90 =
- Symptom Checklist 90
REFERENCES
- Copyright © 1998 American Academy of Pediatrics
In this issue
Jump to section
Related Articles
Cited By...
- Early Assessment and Intervention for Families Experiencing Homelessness: A randomized trial comparing two parenting programs
- Timing and Duration of Pre- and Postnatal Homelessness and the Health of Young Children
- Unstable Housing and Caregiver and Child Health in Renter Families
- Providing Care for Children and Adolescents Facing Homelessness and Housing Insecurity
- Providing Care For Immigrant, Homeless, and Migrant Children
- Homeless Children: What Every Clinician Should Know
- Health Services Use by Children of Migratory Agricultural Workers: Exploring the Role of Need for Care
- Hunger: Its Impact on Children's Health and Mental Health
- Predictors of Change in the Functional Status of Children With Human Immunodeficiency Virus Infection
COMMENTS
In a relatively limited number of studies performed to date, homeless children have been shown to have high rates of acute and chronic illness, and of emergency department use and hospitalization, compared with the general population4,,11 and low-income samples.5,,8,10 Using data from a comprehensive epidemiologic study, the present observational study explores the extent to which homelessness, as well as other potential explanatory factors, are associated with health and service use outcomes in a large sample of homeless and low-income housed children. To our knowledge, this is the first study to determine the extent to which homelessness independently predicts poor health outcomes or high service use in these children.
Our results highlight the independent relationship between homelessness and poor health status. Homelessness likely operates as a risk factor for children's poor health through several means, including exposure to the specific conditions of life in shelters, the instability and distress in the period leading up to the shelter stay, and the greater exposure to stress and adverse environmental factors among homeless children compared with their housed peers. As others have described,9,,10,29 shelters are often overcrowded and have shared food preparation practices, thereby, increasing the risk of transmitting contagious illnesses. Homeless mothers, with less control of their physical environment, have fewer resources to cope effectively with a child's illness. Additionally, the pressure for mothers to deal with their survival needs may interfere with the timely use of health services for their children.
For many children, the period leading up to homelessness is often filled with fear and family distress as well as disconnection from neighborhoods, schools, and supportive relationships. Compared with their housed peers, homeless children were more stressed in the present study, as reflected by their higher rate of moves in the past year, and greater likelihood of experiencing a care and protection investigation or foster care placement in the past. Several studies have described the relation between stressful life experiences and children's higher rate of injuries and illness.30,,31
Mother's perception of their child's health status is related to how they use services and possibly to future health and development.32 Although we controlled for mothers' emotional distress in examining the relationship of homelessness to health status and services, mothers' perceptions may reflect the stress of homelessness, which we were unable to measure directly, and therefore, may have led to biased reports of their children's health status. However, studies of urban low-income women with high levels of environmental stress32,,33 suggest that lower maternal ratings of child health are, in fact, associated with more child illness and hospitalization, independent of mothers' mental health, social support, and stress levels.32,,33 Any reporting bias in our study is, therefore, likely to be small if even operative.
A previous study comparing homeless and housed children (primarily African-American) in Los Angeles10 reported equally high rates of disability days and acute illness symptoms in both groups. In contrast, in our study homeless children were more likely to have acute illness symptoms, and higher prevalence rates of selected symptoms, including diarrhea, ear infections, and asthma, than their housed counterparts. The difference in results between these two observational studies may be attributed to the higher percentage of Puerto Rican children in our sample, the shorter duration of homelessness, and geographic variation in shelter eligibility criteria. When adjusting for potentially confounding factors in the regression analyses, homelessness is no longer associated with acute illness symptom count although it remains predictive of poor health status. Because we did not assess functional limitations attributable to physical illness, it is possible that homeless children may have more severe acute illness symptoms or that differences in symptoms, or perception of such, may contribute to their reported poor health status. In addition to homelessness, other factors may also be important in predicting illness symptoms such as dietary intake and environmental conditions.
A growing literature has described the association between poverty and asthma morbidity.34,,35 The absolute rates for asthma reported during the past year in both the homeless and housed children were very high compared with the general population as well as urban poor samples.34,,36 One-month prevalence rates were twice as high in homeless compared with housed children. Although we were unable to measure the severity of asthma, possible explanations for higher than expected prevalence rates in both groups may relate to dilapidated and overcrowded housing conditions, high levels of maternal stress and social isolation, exposure to passive smoke or noxious pathogens, and elevated levels of emotional and behavioral problems in both groups of children compared with normative samples.35,,37,38 Previous studies have noted that emotional problems can exacerbate asthma symptoms.39 The particularly high rate of reported asthma in homeless children in our sample may in part relate to crowded shelter spaces that facilitate spread of viral infections, mother's need to address survival needs, and higher rates of stressful life events. Additionally, Puerto Rican children, who comprise one third of our sample, have been found to have high rates of asthma.40Strategies for managing asthma in this high risk population will need to consider the unique stresses facing homeless mothers and their children, in addition to current education and self-management approaches.
The regression model we used for identifying children with more frequent outpatient and emergency department visits included health needs and other predisposing sociodemographic and psychosocial variables that have previously been related to health service use patterns.39,41–43 We did not include emotional and behavioral health measures because they were not collected on children under 30 months. Because almost all families had medical coverage, we were unable to look at the effects of health insurance on health care usage rates.
In the present study, homelessness was significantly associated with more frequent emergency department and outpatient medical visits in both crude as well as multivariable adjusted analyses. The higher use rates may reflect the fact that homeless children seem to have more acute illness symptoms that require urgent as well as regular medical care (eg, asthma, ear infections, repeat diarrhea). Delay in receiving timely services because of competing family pressures may also lead to increased illness severity and potentially, more need for care. Furthermore, homeless children are living in shelters in which staff strongly encourage and facilitate well-child care and treatment of symptoms, which may contribute to the higher usage rates observed. Additionally, shelter staff may be sensitized to the high rates of infectious diseases common among homeless children and therefore encourage higher use of medical resources than may be necessary. We cannot comment, however, on the nature of medical visits because of our methods of data collection. It may be, in fact, that homeless children are using appropriate care and that low-income housed children require additional efforts to improve their access to needed services.
Despite the fact that >90% of homeless children in our sample had health insurance and a regular health care provider, emergency department use was high. Although we cannot comment on the appropriateness of the emergency department visits, the fact that homelessness is associated with greater use suggests gaps in the health care delivery system as well as unique barriers to the receipt of primary care services. Because mothers play the central role in defining children's need for and use of services, their high stress levels and survival demands may contribute to emergency department use, which offers more flexibility and availability. Our data demonstrate that homeless mothers are very worried about their children's health, another aspect that may encourage emergency department use. Additionally, programmatic and bureaucratic obstacles, such as inadequate transportation, long waits for appointments, and inadequate capacity to respond to acute care needs may seriously limit access to primary care. Although some communities have developed effective outreach linked to primary care for homeless families, additional efforts must be made to decrease barriers to less costly health services.
The observed association between mother's emotional distress and children's acute illness symptoms, and use of emergency department and outpatient settings, emphasizes the importance of addressing the mother's needs as well. Although mothers who are distressed may be more likely to report illness symptoms in their children, an opposite scenario may also exist. It is also likely that mother's distress may lead to more symptoms in their children, particularly somatic symptoms that may be expressions for emotional distress such as eating and sleep problems, headaches, and abdominal pain.
Previous studies have reported an inconsistent association between mothers' emotional status and high service use for their children.42,44–46 In contrast, our findings indicate a strong relationship between mothers' emotional distress and children's high outpatient and emergency room use. This might be explained by our sampling of extremely poor families as well by the inclusion of a high proportion of ethnic minorities. Severe poverty and stress, social isolation, and emotional distress can influence a mother's perception of her child's health needs. These findings further highlight the need for clinicians to attend to both the mother's and child's needs.
Several limitations of the study must be considered. Failure to gather information about functional status or symptom severity limit our ability to fully assess children's morbidity or the association of these factors to our primary study endpoints. We also did not validate the illness and service utilization self-reports by reviewing medical records or other sources. Although unknown, differential recall bias might be operative in homeless as compared with housed mothers, potentially distorting the extent or direction of observed associations. Our interpretation of high service use rates is further limited because we did not gather information about the appropriateness of medical care usage or were able to distinguish between routine and sick care visits. Additionally, we were unable to identify the acute sequence of events that led to increased use of emergency department and outpatient visits. Prospective collection of data may allow for a clearer delineation of usage patterns as well as illness symptoms and additional indicators of poor health status. Lastly, because Worcester has a well-developed service delivery system for poor and homeless families, the data may not accurately reflect access barriers in other communities.