Prevalence of child welfare services involvement among homeless and low-income mothers: a five-year birth cohort study.

This paper investigates the five-year prevalence of child welfare services involvement and foster care placement among a population-based cohort of births in a large US city, by housing status of the mothers (mothers who have been homeless at least once, other low-income neighborhood residents, and all others), and by number of children. Children of mothers with at least one homeless episode have the greatest rate of involvement with child welfare services (37%), followed by other low-income residents (9.2%), and all others (4.0%). Involvement rates increase with number of children for all housing categories, with rates highest among women with four or more births (33%), particularly for those mothers who have been homeless at least once (54%). Among families involved with child welfare services, the rate of placement in foster care is highest for the index children of women with at least one episode of homelessness (62%), followed by other low-income mothers (39%) and all others (39%). Half of the birth cohort eventually involved with child welfare services was among the group of women who have ever used the shelter system, as were 60% of the cohort placed in foster care. Multivariate logistic regression analyses reveal that mothers with one or more homeless episodes and mothers living in low-income neighborhoods have significantly greater risk of child welfare service involvement (OR = 5.67 and OR = 1.51, respectively) and foster care placement (OR = 8.82 and OR = 1.59, respectively). The implications for further research, and for child welfare risk assessment and prevention are discussed. Specifically, the salience of housing instability/homelessness to risk of child welfare service involvement is highlighted.

Introduction

In the US, 1.4 to 2.15 million children are estimated to have been homeless at some point in 1996 (Burt & Aron 2000) and 547,000 children were in formal out-of-home placements and under child welfare agency supervision in the beginning of 1999 (Department of Health & Human Services 2000). Both phenomena reflect major family disruptions linked to residential instability, poverty, and psychosocial problems such as substance abuse and domestic violence. Yet little research exists on the relationship between homelessness and receiving protective supervision by a child welfare agency, including out-of-home placement. This article addresses this topic through a prospective, 5-year examination of a comprehensive one-year birth cohort in Philadelphia and records of involvement with the municipal shelter and child welfare systems.

Background

Homelessness, after its dramatic increase during the 1980's, has remained a significant social problem throughout the 1990s (Choi & Snyder, 1999; Children's Defense Fund, 1998; Susser, Moore, & Link, 1993; Link et al., 1994). Homelessness among women with children has generated particular concern, with the number of homeless women with children rising at a disproportionate rate in comparison to single homeless adults during the 1980's and comprising one-third of the national homeless population by 1986 (Burt & Cohen, 1989). The proportion of homeless persons in families has since remained at that level, although the number of homeless families has continued to increase as the number of homeless persons in general has increased (Burt et al., 1999).

Homelessness now appears to be a relatively common phenomenon, especially among the urban poor. A national telephone survey conducted in the fall of 1992 revealed that 13.5 million people, or 7.4% of the national population, had experienced "literal homelessness" at some point in their lives, including as many as 7.5 million people, or 3.2% of the US population, in the previous five years (Link et al., 1994, Interagency Council on the Homeless 1994). Closer examination reveals the incidence of homelessness to be unequally distributed by race and age, with as many as 16% of poor African American children under the age of 5 becoming homeless each year in large US cities (Culhane & Metraux, 1999).

In a manner similar to homelessness, children who experience protective supervision from the child welfare system are disproportionately poor (Vondra 1993; Ammerman & Hersen 1990) and African American (Lawrence-Webb 1999). The ecological and psychological effects of poverty, especially with the presence of interacting racial inequalities, negatively affect a family's abilities both to maintain housing stability (Hopper & Milburn 1996) and to care for and parent children (Harden 1998). Despite these shared characteristics between families who are homeless and who are involved with child welfare services, there has been little research looking into the relationship between these two phenomena. Homeless families appear more likely to have children in foster care placements than other poor families, but this body of evidence is far from conclusive. Bassuk et al. (1997b), in a survey of 77 sheltered homeless families and 90 low-income housed families in Worcester MA, reports that 19% of preschool-aged children in homeless families had been placed in foster care, as compared to 8% of the low-income children. Nunez (1994), in surveying 398 homeless families in New York City, finds that 35% of these families have an open child welfare services supervision case and that 20% have one or more children in foster care. Zlotnick, Kronstadt & Klee (1998) find that, in a sample of 195 children in foster care, half of their birth parents had histories of homelessness. While they all offer dramatic findings, the limited study groups and exclusively univariate descriptions of the relationship between homelessness and foster care prevent general conclusions to be drawn from these studies about this relationship.

Thus it is unclear whether homeless families represent a particularly at-risk group for involvement with child welfare services, or whether they share a similar level of risk for such involvement with other low-income families. A further variable that potentially mediates this relationship involves parity, the previous number of children borne by the mother. Higher parity levels indicate larger households and may increase the difficulty and stress associated with locating and maintaining housing, as well as providing for and parenting children. This study evaluates the relationship between homelessness and the risk of children becoming involved with child welfare services (including foster care placement) as a function of the housing and socioeconomic status of the birthmothers (ever-homeless, other low-income, and other families) and their level of parity.

Methodology

Data

Data sources for the study included information from vital statistics records maintained by state authorities, information from records of in-home child welfare service maintained by the City of Philadelphia's Department of Human Services, and information from records of official requests for public shelter maintained by City of Philadelphia's Office of Emergency Services and Shelter (OESS). The basic strategy was to identify and tabulate all occurrences of OESS shelter requests and/or CWS involvement pertaining to a study cohort of Philadelphia resident women who delivered live infants between September 1, 1993 and August 31, 1994 (N = 23,227). For women with multiple births within the one year study period, we randomly selected only one record for use in this study. OESS shelter request data was available for the 9-year period ending in May of 1999; CWS involvement regarding either foster care placement or in-home services was available for a 15 year period ending in May of 1999. Shelter requests and CWS system involvement pertaining to each of these respective time periods was determined for every women in the study cohort, using an electronic data file with merged OESS, CWS and vital statistics record information. (The merged file was originally created by City of Philadelphia's Children and Families Cabinet, for the purpose of examining the extent of cross-system involvement of children from low-income, high risk families; computerized matching algorithms involving names, dates of birth, and social security numbers were used when the file was created to match and merge the OESS, DHS, and vital records data. All identifiers were then stripped from the file for the purpose of this analysis.) Preliminary analyses of the data revealed that of the 23,227 mothers in the study cohort, 2,019 (8.7%) had some type of involvement with CWS, concurrent with or subsequent to the index birth. In addition, a total of 2,703 of the mothers (11.6%) in the study cohort had made an official request for public shelter placement; of those 1,651 (61%) were actually placed and spent at least one-night in a public shelter. The discrepancy between shelter request and actual entry into the shelter system is attributable to the system's inability to track placement dates prior to 1993, incomplete reporting and lack of placement among shelter requestors.

Finally, mothers' addresses available from the birth record were geocoded and matched to 1990 census block groups. Census block groups from the lowest quintile of median household income were designated as "low-income neighborhoods." Women residing in low-income neighborhoods at the time of the cohort children's birth were so tagged. In order to create mutually exclusive housing categories, mothers tagged to the public shelter files were separated from the group of low-income residents, and from the mothers with birth record addresses geocoded to other block groups.

Statistical analyses

Categorical variables were derived to separate birth mothers into three mutually exclusive housing conditions: mothers who requested shelter, low-income residents, and all others. Parity was assessed using the number of previous live births to the birth mother, noted on cohort birth records. Parity was categorized into 5 levels: no previous live births, two live births, three live births, and four or more previous live births. Demographic variables for each housing group were evaluated for significant differences using one-way anova designs and chi-squared analyses. The initial date of CWS involvement with a mother was used to determine the length of time after the index birth that the CWS involvement began. For cases in which mothers had open CWS files from previous live births, initial CWS involvement was noted to be "at birth." The number and percentage of newly opened CWS cases in one-year intervals was assessed for each housing condition by mothers' parity status. Placement rates within the CWS involved population were investigated for each housing group by mothers' parity status. Chi-squared analyses were conducted to detect significant differences of service utilization and placement rates across groups.

Separate logistic regression models were constructed to evaluate the risk of CWS involvement at cohort birth and during the 5-year period post birth. As previously mentioned, mothers who were involved with CWS "at birth" were not primiparous women, that is they were women with one or more previous live births. A third logistic regression model was created to investigate the specific risk associated with foster care placement over the same 5-year period. The regression models were created to determine risk of CWS involvement for each housing category, controlling for mothers' age, parity, race, and educational level. Risk was assessed through evaluating the odds ratios for each independent variable in the logistic equation.

Results

Table 1 displays descriptive demographic statistics for cohort mothers. Using a one-way ANOVA, a main effect was found for maternal age, F (2, 23191) = 283.4, p < .0001. Tukey's HSD analyses revealed that age differences were significant between all housing comparisons, except between the homeless and housed low-income groups. A one-way ANOVA indicated that a main effect was also present for previous live births, F (2, 23191) = 513.4, p < .0001. All univariate comparisons were shown to be significant by Tukey's HSD analyses, parity increasing with housing instability. Chi-squared statistics for race and ethnicity ([chi square] = 4560.1, df = 6), educational level ([chi square] = 1919.2, df = 4), and marital status ([chi square] = 2602.7, df = 4), were all significant at the .0001 level.

Table 2 presents the frequency of CWS involvement for each housing group by parity level. Out of the 23,227 cohort births, 2,019 (8.7%) became involved with CWS within a 5-year span post birth. Significant differences between housing groups were revealed using chi-squared analyses ([chi square] = 3484.1, df = 12). Whereas 37% of the homeless population were involved with CWS over the 5 years, only 9.2% of low-income neighborhood residents and 4.0% of other families were involved. For all housing groups, the percentage of involvement increased with mothers' parity. More than half of homeless mothers (54.0%) with 4 or more previous live births were involved with CWS over 5 years. At the same parity level, 28.0% of low-income mothers and 17.6% of other mothers demonstrated involvement.

Not all families involved with CWS have children placed in foster care. It was found that 1029 children across all housing groups were placed. This represents 4.4% of the original cohort, and 51.0% of the CWS involved population. Chi-square analyses revealed that placement rates were significantly different across housing categories ([chi square] = 2547.0, df = 10). Regardless of parity, children born into families with a history of homelessness were placed at a significantly higher rate (61.8%) than both low-income residents (39.2%) ([chi square] = 59.6, df = 1) and the other housing group (38.6) ([chi square] = 67.8, df = 1). No significant difference in placement rate was found between the low-income neighborhood residents and the other housing group. Among the homeless population the rates of foster care placement decreased as parity level increased. This trend did not hold for the other two groups.

Table 3 displays the odds ratios for CWS involvement for housing categories and covariates. The risk of CWS involvement at birth is 6.87 times greater between ever-homeless families as compared to families neither homeless nor in the lowest quintile of income. The risk for homeless families after a 5-year span drops somewhat to 5.73 times that of the reference group. In addition, among the CWS involved population, ever-homeless families are at the greatest risk for foster care placement (OR = 8.82). Although still a risk, low-income housed families were only 1.52 times as likely to be involved with CWS than families living above the lowest quintile of income at the time of birth, and 1.51 times more likely over 5 years. The risk of foster care placement among low-income residents is 1.59 times greater than the reference group.

The logistic regression analyses demonstrated that other risk factors for CWS involvement at the time of birth include, being of African-American race, having greater parity, and attaining less than a high school education. Similar demographic risk factors are evident for CWS involvement and foster care placement over a 5-year period. However, teen pregnancy is an additional risk for both CWS involvement and foster care placement, whereas maternal age less than 35 years is only a risk factor for CWS involvement.

Implications for Policy and Practice

This study tracked a one-year Philadelphia birth cohort for a period of five years. The investigation revealed that for this cohort, more than one third (37%) of the children of ever-homeless mothers became involved with CWS, as compared to fewer than one-tenth (9.2%) among low-income residents. Indeed, half of all children in the birth cohort who eventually became involved with CWS were the children of ever-homeless mothers.

There are many reasons why homeless families may be particularly susceptible to involvement with child welfare agencies, and why families involved with child welfare services might likewise be more likely to be admitted to homeless shelters. First, homeless parents and their children must confront the detrimental conditions differentially associated with the incidence of homelessness, including more severe poverty, housing crowding, substandard housing conditions, and domestic violence (Bassuk et al., 1997a; Culhane, Lee & Wachter 1997; Shinn et al. 1998). Each of these factors may increase the likelihood of child neglect and abuse, and the likelihood that these families will come to the attention of child welfare workers. Second, families involved with child welfare services may be referred or placed in shelter by social workers as part of their service plans, as a result of unsafe housing conditions coming to the attention of case workers. The benefits of shelter placement, including being made eligible for subsidized housing programs, may also increase the perceived value of shelter admission by families and/or case workers. Furthermore, once homeless, children's development may be compromised by the experience of being homeless or living in a public shelter, including as a result of increased social isolation of the family, disruptions to schooling, shelter crowding, a loss of parental autonomy, and substandard living conditions. Children and families in shelter are also likely to be under greater scrutiny by child protective services workers by virtue of their homeless status, and residence in a supervised facility. Long-term housing instability and homelessness, or factors associated with them, may also infringe upon parents' ability to parent effectively, including having fewer supports from extended family (often resulting from homeless parents' childhood disruption or abuse), having substance use or mental health problems, and larger family size (Zlotnick, Kronstadt, & Klee, 1998). Given the broad range of such possible associations, care should be taken in drawing conclusions about the reasons for high rates of child welfare involvement among homeless families. Further research in this area is needed.

The high degree of overlap in agency caseloads also suggests important policy and programmatic implications for child welfare and public shelter services. Although more research is needed to determine the appropriateness of various interventions, it is clear that child welfare agencies should have a vested interest in working closely with public shelter programs. If housing problems are so common among CWS families, perhaps CWS systems should play a more explicit role in the financing, development and management of transitional and permanent, supportive housing programs for cross-system involved families. Likewise, the delivery of child abuse prevention and family support services to homeless families may significantly benefit children and CWS systems, in that they may reduce abuse and neglect (50% of the CWS caseload in this birth cohort were ever-homeless) and out-of-home placements (60% of the foster care placements in this cohort were the children of ever-homeless mothers). Moreover, the provision of housing assistance to families involved with child welfare services would likely reduce their homelessness, and the possible negative consequences of homelessness for children and their families. And finally, a more comprehensive system of housing assistance, that assured poor families of stable, affordable, and adequate housing, could both reduce the incidence of homelessness and abuse and neglect. Again, future research could help identify which of these strategies will be most beneficial for families.

This study has also shown that parity cannot be ignored when considering the risk of CWS involvement. Mothers with more children in their care are faced with greater parenting responsibilities, greater income demands, and more difficulty finding housing of adequate size and affordability. The added stress associated with these factors may place women with greater parity at a higher risk of CWS involvement, and homelessness. The results of this study support this hypothesis. The rate of CWS involvement over 5-years increased from 3.7% for primiparous women to 32.7% for women with four or more previous live births. This trend holds across all housing groups. The rate increase was most dramatic for homeless women. Among this population of homeless families, 23.6% of primiparous mothers became involved with CWS within five years of the cohort birth. This rate rises sharply at the highest level of parity, with 54.0% of homeless women with 4 or more previous live births becoming involved with CWS.

Foster care placement is also affected by parity level. Once again, homeless families appear to experience the highest rate of involvement. Over the 5-year period, it was found that 61.8% of CWS involved homeless families had children placed in foster care, as compared to 39.2% of CWS involved low-income residents, and 38.6% of the other housing group with CWS involvement. Interestingly, parity level had no clear relationship with foster care placement rates for the low-income and other housing group. In CWS involved homeless families, the rate of foster care placement decreased as the number of children per family increased.

The cause of the peculiar pattern of foster care placement across levels of parity in CWS involved homeless families is unknown. It could be that CWS is reluctant to place children in foster homes that cannot accept multiple siblings at one time. Another possibility is that the severity of problems drawn to the attention of CWS may be influenced by family size. For example, women with several children may be referred to CWS for neglect due to limited resources, whereas primiparous women may be referred more frequently for abusive parenting. The results of this investigation suggest that more research is needed to determine how housing instability influences the practices of family social service agencies.

Table 1
Descriptive Statistics for Mothers across Housing Groups

                                 Requested
                                  Shelter     Low Income      Other
                                (n = 2703)    (n = 4342)   (n = 16182)

Age, mean (SD) *                24.7 (6.0)    24.4 (6.6)    26.6 (6.4)
Live Births, mean (SD)*          2.23 (1.9)    1.4 (2.1)     1.1 (1.7)
  Race, % **
  African-American (non-hisp)   91.9          65.9          42.8
  Caucasian (non-hisp)           3.7           7.9          44.5
  Asian                           .4           3.2           5.4
  Hispanic Origin                3.6          22.8           7.1
  Other                           .2            .1            .1
Educational Level, % **
  Less than 12 years            51.9          42.0          20.4
  High School Grad              38.4          40.6          43.2
  Beyond High School             9.7          17.4          36.5
Marital Status, % **
  Married                        6.2          18.8          48.6
  Not Married                   93.7          81.1          51.4

* p < .0001 based on a chi-squared test statistic.

** p < .0001 based on an F-test statistic from an analysis of variance.

Table 2
Frequency of CWS Involvement and Foster Care Placement By Parity and
Housing Group

                  Housing          Total         No CWS
                 Stability         Number      Involvement
Parity            Measure         of Births     (n and %)

0            Requested Shelter    n = 518       396 (76.4)
Previous     Lowest Quintile      n = 1533     1462 (95.4)
  Live         of Income
Births       All Others           n = 6671     6538 (98.0)
             Total                n = 8722     8396 (96.3)

1            Requested Shelter    n = 594       428 (72.1)
Previous     Lowest Quintile      n = 1210     1129 (93.3)
  Live         of Income
  Birth
             All Others           n = 5073     4923 (97.0)
             Total                n = 6877     6480 (94.2)

2            Requested Shelter    n = 581       360 (62.0)
Previous     Lowest Quintile      n = 776       698 (89.9)
  Live         of Income
  Birth
             All Others           n = 2644     2506 (94.8)
             Total                n = 4001     3564 (89.1)

3            Requested Shelter    n = 421       247 (58.7)
Previous     Lowest Quintile      n = 434       372 (85.7)
  Live         of Income
  Birth
             All Others           n = 1079      971 (90.0)
             Total                n = 1934     1590 (82.2)

4+           Requested Shelter    n = 589       271 (46.0)
Previous     Lowest Quintile      n = 389       280 (72.0)
  Live         of Income
  Birth
             All Others           n = 715       589 (82.4)
             Total                n = 1693     1140 (67.3)

All Births   Requested Shelter    N = 2703     1702 (62.3)
             Lowest Quintile of   N = 4342     3941 (90.8)
               Income
             All Others           N = 16182   15527 (96.0)
             Total                N = 23227   21170 (91.1)

                                     At Birth         Birth to 5

                  Housing         CWS involvement   CWS involvement
                 Stability         (cumulative n     (cumulative n
Parity            Measure             and %)            and %)

0            Requested Shelter        0 (0)            122 (23.6)
Previous     Lowest Quintile          0 (0)             71 (4.6)
  Live         of Income
Births       All Others               0 (0)            133 (2.0)
             Total                    0 (0)            326 (3.7)

1            Requested Shelter       32 (5.4)          166 (27.9)
Previous     Lowest Quintile          9 (0.7)           81 (6.7)
  Live         of Income
  Birth
             All Others              23 (0.5)          150 (3.0)
             Total                   64 (.09)          397 (5.8)

2            Requested Shelter       73 (12.6)         221 (38.0)
Previous     Lowest Quintile          9 (1.2)           78 (10.1)
  Live         of Income
  Birth
             All Others              29 (1.2)          138 (5.2)
             Total                  111 (2.8)          437 (10.9)

3            Requested Shelter       87 (20.7)         174 (41.3)
Previous     Lowest Quintile         19 (4.4)           62 (14.3)
  Live         of Income
  Birth
             All Others              32 (3.0)          108 (10.0)
             Total                  138 (7.1)          344 (17.8)

4+           Requested Shelter      227 (38.5)         318 (54.0)
Previous     Lowest Quintile         60 (15.4)         109 (28.0)
  Live         of Income
  Birth
             All Others              57 (8.0)          126 (17.6)
             Total                  344 (20.3)         553 (32.7)

All Births   Requested Shelter      419 (15.5)        1001 (37.0)
             Lowest Quintile of      97 (2.2)          401 (9.2)
               Income
             All Others             141 (.9)           655 (4.0)
             Total                  657 (2.8)         2019 (8.7)

                                     Birth to 5

                                    Cumulative n
                  Housing          Placed and % of
                 Stability        CWS cases Placed
Parity            Measure          in Foster care

0            Requested Shelter        93 (76.2)
Previous     Lowest Quintile          33 (46.5)
  Live         of Income
Births       All Others               74 (55.6)
             Total                   200 (61.3)

1            Requested Shelter       112 (67.5)
Previous     Lowest Quintile          31 (38.3)
  Live         of Income
  Birth
             All Others               50 (33.3)
             Total                   193 (48.6)

2            Requested Shelter       132 (59.7)
Previous     Lowest Quintile          29 (37.2)
  Live         of Income
  Birth
             All Others               52 (37.7)
             Total                   213 (48.7)

3            Requested Shelter       102 (58.6)
Previous     Lowest Quintile          26 (41.9)
  Live         of Income
  Birth
             All Others               39 (36.1)
             Total                   167 (48.5)

4+           Requested Shelter       180 (56.6)
Previous     Lowest Quintile          38 (34.9)
  Live         of Income
  Birth
             All Others               38 (30.2)
             Total                   256 (48.0)

All Births   Requested Shelter       619 (61.8)
             Lowest Quintile of      157 (39.2)
               Income
             All Others              253 (38.6)
             Total                  1029 (51.0)

Table 3
Relative Risk of CWS Involvement and Foster Care Placement

                          Odds of CWS              Odds of CWS
                     Involvement At Birth    Involvement over 5 years

Request Shelter       6.87 ** (5.39-8.76)      5.73 ** (5.02-6.55)
Lowest Quintile       1.52 * (1.13-2.05)       1.51 ** (1.30-1.75)
  of Income
Teen                  0.80 (.49-1.32)          2.13 ** (1.65-2.77)
20-24 years           0.64 * (.46-.90)         1.42 * (1.14-1.78)
25-34 years           1.25 (.95-1.65)          1.50 ** (1.22-1.83)
Caucasian             0.57 ** (.41-.79)        0.53 ** (.45-.63)
Hispanic              0.41 * (.03-.57)         0.60 ** (.49-.73)
Asian                 0.14 ** (.26-.63)        0.19 ** (.11-.36)
One Live Birth       Not Applicable            1.75 ** (1.48-2.08)
Two Live Births       2.12 ** (1.52-2.96)      2.93 ** (2.43-3.52)
Three Live Births     4.71 ** (3.37-6.59)      4.62 ** (3.76-5.68)
Four or More Live    10.91 ** (7.90-15.07)     8.27 ** (6.71-10.18)
  Births
Less than High        1.61 ** (1.32-1.97)      1.64 ** (1.46-1.86)
  School
Beyond High School    0.60 * (.43-.85)        .60 ** (.50-.72)
  Education

                      Odds of Foster Care
                     Placement over 5 years

Request Shelter       8.82 ** (7.32-10.62)
Lowest Quintile       1.59 ** (1.27-2.00)
  of Income
Teen                  1.66 * (1.17-2.34)
20-24 years           1.07 (1.0-1.58)
25-34 years           1.25 (.96-1.65)
Caucasian             0.46 ** (.35-.59)
Hispanic              0.45 ** (.07-.48)
Asian                  .18 ** (.32-.61)
One Live Birth        1.26 * (1.0-1.58)
Two Live Births       1.88 ** (1.47-2.40)
Three Live Births     2.60 ** (1.98-3.42)
Four or More Live     3.48 ** (2.64-4.58)
  Births
Less than High        1.54 ** (1.31-1.81)
  School
Beyond High School     .76 * (.60-.97)
  Education

Note. * p <.01, ** p <.001, *** Reference group for Odds at Birth = one
previous birth and for Odds Over 5 Years = No previous births.

References

Ammerman, R. T. & Hersen, M. (1990). Research in child abuse and neglect: Current status and an agenda for the future. In R. T. Ammerman and M. Hersen, eds. Children at Risk: An Evaluation of Factors Contributing to Child Abuse and Neglect. New York: Plenum Press. 3-22.

Bassuk, E. L., Buckner, J. C., Weinreb, L. F., Browne, A., Bassuk, S. S., Dawson, R., & Perloff, J.N. (1997b). Homelessness in Female Headed Families: Childhood and Adult Risk and Protective Factors. American Journal of Public Health, 87, (2), 241-248.

Bassuk E. L., Weinreb L. E, Dawson R., Perloff J. N., & Buckner J. C. (1997a). Determinants of behavior in homeless and low-income housed preschool children. Pediatrics. 100: (1) 92-100.

Burt, M. R. & Aron, L. Y. (2000). America's Homeless II: Population and Services. Washington DC: The Urban Institute.

Burt, M. R., Aron, L. Y., Douglas T., Valente, J., Lee, E., Iwen, B. (1999). Homelessness: Programs and the People They Serve, Findings of the National Survey of Homeless Assistance Providers and Clients. Washington DC: Urban Institute.

Burt, M. R. & Cohen, B. E. (1989). America's Homeless: Numbers, Characteristics, and the Programs that Serve Them. Washington DC: Urban Institute.

Children's Defense Fund. (1998). The State of America's Children: Yearbook 1998. Washington, D.C.: author.

Choi, N. G. & Snyder, L. J. (1999). Homeless families with children: Barriers to finding decent housing. Journal of Poverty, 3, (2), 43-66.

Culhane, D. P. Lee, C. M. & Wachter, S. M. (1996). Where the homeless come from: A study of the prior address distribution of families admitted to public shelters in New York City and Philadelphia. Housing Policy Debate, 7,(2): 327-365.

Culhane, D. P. & Metraux S. (1999). Assessing relative risk for homeless shelter usage in New York City and Philadelphia." Population Research and Policy Review, 18,(3): 219-236.

Department of Health & Human Services (2000). The AFCARS Report. Washington DC: author. (http://www.acf.dhhs.gov/programs/cb/afcars/stats.html).

Harden, B. J. (1998). Building bridges for children: Addressing the consequences of exposure to drugs and to the child welfare system. In R. L. Hampton, V. Senatore, & T. P. Gullota, eds. Substance Abuse, Family Violence, and Child Welfare. Thousand Oaks CA, Sage Publications. 18-61.

Hopper, Kim and Milburn, Norweeta G. 1996. Homelessness among African-Americans: A historical and contemporary perspective. In Jim Baumohl, ed. Homelessness in America. Phoenix, AZ.: The Oryx Press. 123-131.

Interagency Council on the Homeless (1994). Priority Home: The Federal Plan to Break the Cycle of Homelessness. Washington DC: US Department of Housing and Urban Development.

Lawrence-Webb, C. (1999). African American children in the modern child welfare system: A legacy of the Flemming Rule. In S. Jackson & S. Brisset-Chapman, eds. Serving African American Children. New Brunswick NJ: Transaction Publishers. 9-30.

Link, B. L., Susser, E. S., Stueve, A., Phelan, J., Moore, R. E., & Struening, E. L., (1994). Lifetime and five-year prevalence of homelessness in the United States. American Journal of Public Health, 84, (12), 1907-1912.

Nunez, R.D. (1994). Hopes, Dreams and Promise: The Future of Homeless Children in America. New York, NY: Institute for Children and Poverty, Homes for the Homeless Inc.

Shinn M., Weitzman B. C., Stojanovic D., Knickman J. R., Jimenez L., Duchon L., James S., & Krantz D. H. (1998). Predictors of homelessness among families in New York City: From shelter request to housing stability. American Journal of Public Health, 88, (11): 1651-1657.

Susser, E., Moore, R., & Link, B. (1993). Risk factors for homelessness. American Journal of Epidemiology, 15, (2), 546-556.

Vondra, J. I. (1993). Childhood Poverty and Child Maltreatment. In J.A. Chafel, ed. Child Poverty and Public Policy. Washington DC: The Urban Institute. 127-166.

Zlotnick, C, Kronstadt, D., & Klee, L., (1998). Foster care children and family homelessness. American Journal of Public Health, 88, (9), 1368-1371.

DAVID WEBB

The Philadelphia Department of Public Health

SUSAN GRIM

STEPHEN METRAUX

DENNIS CULHANE

The University of Pennsylvania

Jennifer F. Culhane

Thomas Jefferson University

834 Chestnut Street, Suite 320

Philadelphia, PA 19107

[e-mail]