Childhood Predictors of Daily Substance Use among Homeless Women of Different Ethnicities

The objective of this study was to describe associations between childhood factors and adult daily substance use profiles among homeless women and to identify independent predictors of daily substance use for the overall sample and for ethnic/acculturation subgroups. Structured surveys were administered to 1,331 homeless women in Los Angeles who were either daily drug and alcohol users for the past 6 months or nondaily substance users. Physical abuse and parental drug abuse predicted daily drug use in the whole sample and selected subgroups, whereas parental alcohol abuse predicted daily alcohol use in the whole sample. Teen self-esteem was also found to have a protective effect on daily alcohol use for the sample and for African American women. Negative peer influence in adolescence predicted daily drug use among high-acculturated Latinas. In summary, childhood abuse, parental substance use, and negative peer influence affect important roles in homeless women's daily substance use.

The impact of childhood abuse on subsequent drug and alcohol use among homeless women has received particularly little attention. Moreover, little is known about other childhood factors that may lead to substance use in this population or how the impact of these factors differs as a function of ethnicity and acculturation. Daily drug and/or alcohol use in the past 6 months strongly suggests physiological or psychological dependence and is likely to be associated with a range of functional problems, for example, crime involvement, impaired social relationships, and employment difficulties (Anglin & Hser, 1990;Longshore, Hsieh, & Anglin, 1993). Therefore, problematic substance use was defined in terms of daily drug or alcohol use. This article addresses four main research questions: (a) What are the associations between childhood factors and daily substance use among homeless women? (b) Do these childhood factors differ by ethnicity and acculturation? (c) Are childhood abuse measures independent predictors of daily substance use when other childhood factors are accounted for? and (d) Do independent predictors of daily substance use by homeless women differ by ethnicity and acculturation?

METHOD PARTICIPANTS AND SETTING
The participants in this study were obtained from a purposive sample of 1,344 homeless women residing in 47 traditional and sober living shelters or obtained through outreach from these sites in the downtown and surrounding areas of Los Angeles. Women were considered eligible if they were (a) age 18 or older and (b) homeless. A homeless woman was defined as one who spent the previous night in a shelter, hotel, motel, or home of a relative or friend and was uncertain as to her residence in the next 60 days or stated that she did not have a home or house of her own in which to reside (Gelberg & Linn, 1989). Among the original sample of 1,344 women, 1,331 had complete information on drug and alcohol use in the past 6 months. Twelve women who did not identify themselves as either African American, Caucasian, or Latina were excluded due to their small sample size.

PROCEDURE
Data were collected between 1994 and 1996 to examine initiation and continuation of substance use among homeless women. Research staff consisted of female African American, Latina, and Caucasian nurses and outreach workers extensively trained in working with homeless and drug-addicted women. Initial contacts in shelters were made through letters to site directors. Homeless women residing within the participating shelters or obtained through outreach were then recruited through presentations provided by research staff to groups of women or on a one-on-one basis. All women interested in participating in the study notified the project nurses and outreach workers. Those who met the inclusion criteria were informed of the study and were required to read and sign a written informed consent. Only 4% of women who met the eligibility criteria declined to participate.
After consent was obtained, appointments were made for interviews to be conducted in a variety of places considered convenient by the participants. A trained nurse or outreach worker of the participant's ethnicity administered a face-to-face structured interview, which took approximately 60 minutes to complete. The questionnaires used were available in both English and Spanish. Women received $10 for their time. The study was approved by the human subjects protection committee of the University of California, Los Angeles.

MEASURES
Many of the instruments presented below have been previously assessed for content validity. Race/ethnicity was measured as part of a structured instrument that elicited information about respondents' sociodemographic characteristics. Assimilation to the U.S. culture among Latinas was determined by a 20-item scale developed by Cuellar, Harris, and Jasso (1980). Twelve of the 20 items that made up this scale were related to language use and preference. Burnam, Hough, Karno, Escobar, and Telles (1987) found this measure to have sound psychometric properties and that women who were more acculturated had a higher prevalence of drug and alcohol dependence than those less acculturated. A cutoff point of 60 corresponds to being equally comfortable with English and Spanish. Latina respondents who scored below this cutoff point were considered to be low acculturated.
Substance use was assessed by the slightly revised Drug History Form (Simpson, 1991). Substances on the form were alcohol, inhalants, marijuana, hallucinogens, crack/freebase, other cocaine, heroin, street methadone, other opiates, amphetamines and methamphetamine, librium, barbiturates, and other sedatives. Items included lifetime use and frequency of use in the past month and the past 6 and 12 months. For purposes of this study, problematic substance use was operationalized as daily use of drugs or alcohol for the past 6 months.
Predictive validity for the drug use measure was well established by Anglin et al. (1996). Furthermore, test-retest reliability for daily drug use and abstinence is in an acceptable range of .61 to .71 (Anglin, Hser, & Chou, 1993).
History of childhood victimization was assessed by three items using a yes-no response format: (a) "Were you ever sexually abused as a child (i.e., before age 18 or before you started living on your own)?" (b) "Were you ever physically abused as a child?" and (c) "Were you ever verbally or emotionally abused as a child?" History of parental substance abuse was assessed by two items with a yes/no/ don't know response format: (a)"Did either of your parents have a drinking problem?" and (b) "Did either of your parents use illegal drugs when you were a child?" Childhood family characteristics were assessed with single items. These characteristics included assessing family socioeconomic status as measured on a 5-point scale ranging from 1 (poor) to 5 (upper class); relationship with parents, measured on a 6-point scale from 1 (excellent) to 5 (very poor) and 6 (no relationship to evaluate); closeness of the family, measured on a 5-point scale from 1 (very close) to 5 (not close at all); and maternal and paternal support, measured on a 5-point scale from 1 (very supportive) to 5 (not around).
Childhood alternate living situation was assessed by an item asking respondents whether they had spent any time living in a juvenile hall, residential treatment facility, orphanage, or other institution or group care facility during the first 18 years of their life.
Self-esteem as a teenager was measured by a single item asking how the women had felt about themselves as adolescents. Responses ranged from 1 (liked yourself a great deal) to 5 (disliked yourself a great deal).
Childhood support was assessed by asking the women who had most negatively influenced their attitudes and behavior when they were adolescents and whether there were any adults they trusted enough to speak to about their personal problems and feelings.

DATA ANALYSIS
Two dichotomous outcome measures were created. The first represented daily drug use (daily use of drugs except tobacco during the past 6 months), and the second represented daily alcohol use (daily use of alcohol during the past 6 months). Chi-square tests were used to examine associations between these two substance use measures and categorical childhood characteristics and between the same childhood characteristics and ethnicity/acculturation (African Americans, Caucasians, and high-and low-acculturated Latinas). Analysis of variance and Student-Newman-Keuls tests were used to contrast the four ethnic/acculturation groups on ordinal measures, such as family socioeconomic status and family closeness; t tests were used to contrast those who did and did not use drugs and alcohol on a daily basis. Some measures were reverse coded for analysis so that higher scores reflected more of the given construct.
Multiple logistic regression analysis was used to identify independent predictors of daily drug and alcohol use for the total sample and for women in each of the ethnic/acculturation subgroups. Because of the large number of potential predictors and the relatively small sizes of some ethnic subgroups, variable selection techniques were used. All childhood characteristics were entered into stepwise backward logistic regression analyses, and findings were confirmed with stepwise forward analyses. Variables significant at the .20 level in the stepwise runs were included in full model runs and constituted the final models; however, we use the .05 level for statistical significance in reporting.

RESULTS
The sample was predominantly African American (49%); 15% were highacculturated Latina, 15% were low-acculturated Latina, and 21% were White. The mean age of the women was 33 years (SD = 8.9), and they had a mean educational level of 11 years (median = 12). Almost half (46%) of the women reported being verbally or emotionally abused as children; childhood sexual and physical abuse were reported by 30% and 27%, respectively. Of the women, 51% said at least one parent had a drinking problem, 23% reported a parent with a drug problem, and 46% cited peers as most negatively influencing them as teenagers. The average woman came from a working-class, moderately close family and regarded her mother as being midway between somewhat supportive and neutral toward her when she was a child. The average father was seen as being neutral in terms of support.
In terms of substance use, 27% of the women reported daily alcohol use. Almost half (48%) of the women reported daily drug use. The most prevalent drugs used by the sample on a daily basis were crack (28%), marijuana (16%), cocaine (9%), heroin (8%), and methamphetamine (7%). Daily use of all remaining drugs was reported by 3% or less of this sample.

PROFILES OF DAILY DRUG AND ALCOHOL USERS
As depicted in Table 1, homeless women who reported childhood sexual, physical, and emotional abuse and having a parent who abused drugs or alcohol were more likely to be involved in daily drug use than women who did not report these characteristics. No significant associations were found between daily drug use and having had a trusted adult during childhood, the citing of peers as their most negative influence, or having lived in a juvenile institution.
Homeless women who reported childhood physical and emotional abuse and having a parent with an alcohol or drug problem were also more likely to report daily alcohol use. Those who had lived in a juvenile institution for some period of time were more likely to report daily use of alcohol as compared to their counterparts who had not. Having had a trusted adult in childhood and having a peer as their most negative adolescent influence were not related to daily alcohol use by the women.
Daily alcohol use by these women was associated with less close-knit families, less maternal and paternal support, and poorer self-esteem in adolescence (p < .001; data not tabled). A similar childhood profile was found for daily drug use, except for self-esteem, which was significant at the .05 level. Family socioeconomic status was not related to daily substance use.

PROFILES BY ETHNICITY/ACCULTURATION
As shown in Table 2, significant differences in daily drug and alcohol use status and childhood characteristics were observed among homeless women of 40 AMERICAN BEHAVIORAL SCIENTIST different ethnicity and level of acculturation. About one third of African American (30%) and Caucasian (33%) women reported daily alcohol use (as compared to 18% for both high-and low-acculturated Latinas). Approximately half of African American (49%), Caucasian (55%), and high-acculturated Latinas (50%) reported daily drug use as compared to 35% of low-acculturated Latinas. Low-acculturated Latinas were least likely to report a history of childhood physical, sexual, or emotional abuse and to feel their parent(s) had an alcohol or drug problem. They were also least likely to report having had a trusted adult in childhood or living in a juvenile institution. However, they were most likely to cite a same-or opposite-sex peer as their most negative influence in adolescence. Caucasians were most likely to report childhood physical and emotional abuse, whereas African Americans were most likely to have had an adult they trusted as children.  Table 3 summarizes childhood psychosocial measures and socioeconomic status for women in the four groups. Caucasians came from the highest socioeconomic status families; however, they received the least parental support, came from the least close-knit families, and reported the lowest self-esteem as teenagers. Although low-acculturated Latinas came from the poorest families, they reported relatively high levels of family cohesiveness and parental support. African Americans reported relatively high levels of maternal support and relatively low levels of paternal support. Table 4, physical abuse in childhood and parental drug abuse positively predicted daily drug use in the total sample and among low-acculturated Latinas. Among African Americans, parental drug abuse was also related to daily drug use, whereas maternal support had a protective effect. Among Caucasians, childhood physical abuse and parental drug use positively predicted daily drug use, whereas living in a juvenile institution was negatively related to daily 42 AMERICAN BEHAVIORAL SCIENTIST  drug use. For high-acculturated Latinas, having a peer as their most negative adolescent influence positively predicted daily drug use. Among low-acculturated Latinas, women from more close-knit families were less likely to report daily drug use, but those from higher socioeconomic backgrounds were more likely to report daily use. Turning to daily alcohol use (see Table 5), parental alcohol abuse positively predicted daily alcohol use for the entire sample, whereas teen feelings of self-esteem had a protective effect. Family closeness and teen self-esteem had a protective effect on daily alcohol use for African Americans. In addition, among high-acculturated Latinas, women with more close-knit families and higher family socioeconomic status were least likely to report daily alcohol use.

DISCUSSION
The rates of 30% for childhood sexual abuse and 27% for childhood physical abuse reported by the women in this study confirm findings from previous studies indicating that homeless women have frequently been victimized early in their lives (Bassuk, 1993;Bassuk et al., 1996Bassuk et al., , 1997D'Ercole & Struening, 1990). The lower prevalence of childhood physical and sexual abuse found in this study as compared to some other studies of homeless women (Bassuk et al., 1996(Bassuk et al., , 1997 may be due in part to the more restrictive definitions that were used and the fact that women were forced to label their experiences. However, in a 1997 survey of a probability sample of 974 homeless women in Los Angeles, Wenzel, Anderson, Gifford, and Gelberg (in press) found a rate of childhood sexual abuse (32%) similar to that in our sample using a question that did not involve labeling but did elicit information about forced acts of vaginal, oral, or anal sex. Thus, our victimization figures are likely to be conservative estimates.
Almost half of this study's sample reported being emotionally abused in childhood. Emotional abuse has not been examined thoroughly in previous research, although many clinicians believe the impact of this type of abuse is similar to that of physical abuse (Hart & Brassard, 1987). Consistent with our expectations, women who did not report problematic substance use had markedly fewer experiences with childhood abuse in any form.
Study findings also revealed that a number of childhood factors were associated with daily drug and alcohol use in this sample of homeless women. A family history of drug and alcohol problems was common in this sample and was associated with subsequent use of drugs and alcohol. Moreover, daily alcohol users were more likely to have lived in a juvenile institution. In a study of childhood risk factors for homelessness, Koegel, Melamid, and Burnam (1995) reported that about half of the respondents had lived apart from their families, suggesting high levels of personal disturbance or familial disruption. Moreover, homeless women were more likely to have lived in foster care as compared to homeless men. These authors contend that the current problems of residential and economic instability faced by homeless adults are often extensions of patterns and risk factors that were present in their childhood and may contribute to predispositions to substance use, mental illness, and disruptions in social relationships. Childhood placements away from home were similarly found to be associated with adult psychiatric disorders among a sample of homeless men (Susser, Struening, & Conover, 1987). Childhood factors were also notably different by ethnicity and acculturation. Confirming a previous study with homeless women (Nyamathi, Bennett, Leake, Lewis, & Flaskerud, 1993), low-acculturated Latinas were least likely to abuse drugs, whereas about half of African Americans, Caucasians, and high-acculturated Latinas did so. Low-acculturated Latinas were also less likely than the other three groups of women to report negative family characteristics and a history of abuse. On the other hand, Caucasians were most likely to disclose physical and emotional abuse histories and placement in a juvenile institution. They also reported the lowest level of adolescent self-esteem.
It is interesting that the childhood physical abuse measure was predictive of daily drug users, whereas no childhood abuse measures were predictive of daily alcohol users. Instead, other measures such as parental alcohol abuse, teen self-esteem, and family closeness appeared more important. Furthermore, whereas daily drug users reported living with parents who abused drugs, daily alcohol users reported living with parents who abused alcohol. These findings support the strong role modeling that occurs by parents and its negative impact on subsequent problematic substance abuse on children later in life (Roye & Balk, 1996).
Adolescent self-esteem proved to be the most protective predictor against daily alcohol use for the entire sample and is consistent with the literature that reports that higher self-esteem has generally been associated with health-promotion behaviors (Muhlenkamp & Sayles, 1986;Nyamathi, 1991). Family closeness was also a significant protective factor against daily alcohol use for African Americans and high-acculturated Latinas and for daily drug use among low-acculturated Latinas. This result supports the evidence that shows the generally positive association between parental involvement and positive outcomes for children, such as better outcomes for education and employment, less depression, and less incidence of pregnancy during adolescence (Roye & Balk, 1996). Homeless women with more recent contact with family were similarly less likely to have a history of substance abuse or criminal activity (Roll, Toro, & Ortola, 1999).
In the case of Caucasians, in particular, having spent time in a juvenile institution was a protective factor against drug abuse and may have filled a void created by dysfunctional families. Similarly, high-acculturated Latinas who were protected by close-knit families were found to report less problematic alcohol use than did the African American and Caucasian women. In contrast, high-acculturated Latinas who were most likely to cite their peers as their most negative influence when they were adolescents reported higher levels of drug use than Latinos not reporting this negative influence. This finding may reflect the powerful role of family in Hispanic cultures and the need for more detailed consideration of the impact of negative peer influence on high-acculturated Latinas.
In this study, higher family socioeconomic status independently predicted daily drug use for low-acculturated Latinas especially, whereas for high-acculturated Latinas it had a protective effect with alcohol abuse. Although we have no straightforward explanation for these findings, there are clearly a variety of pathways that may lead women to drug addiction and homelessness. For example, wealthier high-acculturated Latinas may have benefited from a better educational exposure to the dangers of substance use, whereas lower acculturated Latinas may have been less likely to be exposed to the drug media blasts. However, it is always plausible that individuals with money may simply be better able to purchase drugs or alcohol. Prospective studies are needed to more thoroughly investigate why some individuals develop problems with alcohol and drugs when others with apparently similar backgrounds do not and to tease out why some factors may affect various ethnic groups differently. Research to elucidate substance abuse among high-acculturated Latinas is particularly important, because negative peer influence was the sole predictor of daily drug use in this sample.
Our sample of homeless women compares similarly with that of other samples of homeless adults from Los Angeles. For example, in Koegel, Burnam, and Farr's (1988) study, the mean age and predominant ethnicity of the homeless women participants paralleled the demographics characteristics of our homeless women. Both samples reported the same amount of alcohol use in the past 6 months (27%). In another study comparing substance abuse status of homeless and domiciled low-income users of a medical clinic, Linn, Gelberg, and Leake (1990) reported similar findings in terms of the mean age, substance use patterns, and childhood rates of abuse, thus lending support to the generalizability of our sample.
There are several limitations to this study. First, data on childhood abuse are based on self-report and, due to their highly sensitive nature, may well have been underreported. Although the interviewers were from the community and were trained to develop a good rapport with the participants, some women may have blocked out memories of childhood abuse, especially sexual abuse, and others may have underreported abusive experiences they did remember due to their own feelings of guilt and shame. Second, it is important to note that women who did not report daily drug use during the past 6 months may nevertheless have had problems related to drug use during the same time frame or earlier in their lives. Thus, we may have underestimated the lifetime prevalence of problem drug use in our sample. We also acknowledge multiple comparisons as a limitation. Finally, although we cannot be certain that daily drug or alcohol use unequivocally implies problem use or abuse/dependence on a formal diagnostic basis, we feel confident in our assumption that problem use or abuse/dependence is very likely among persons using substances daily over a period as long as 6 months.
Despite these limitations, this study provides valuable insights into relationships between childhood factors and daily substance use among homeless women, highlights cultural differences, and has implications for health care providers and intervention programs dealing with homeless women and families. In recent years, families have come to comprise more than 36% of the homeless population (U.S. Conference of Mayors, 1995), and their numbers continue to grow in the wake of welfare reform. The majority of these families are headed by women with young children. Most homeless families are able to avail themselves of emergency shelter services; however, they live in a chaotic environment with endemic substance use and violence. The importance of parental drug and alcohol abuse found in this study underscores the need for access to treatment programs for substance abuse for homeless parents and homeless women who will likely give birth in the future even if they do not currently have children living with them. Although the need for substance abuse prevention may be less urgent for the future of children in low-income domiciled populations, these overall considerations apply to them as well, because most women in the sample came from this background.
Finally, clinicians caring for homeless women who abuse drugs and alcohol need to have time to screen them for childhood histories of abuse and dysfunction to help them to identify underlying reasons for their addictions and encourage them to seek appropriate substance abuse and mental health services. Likewise, intervention programs designed to help these women need to incorporate strategies that help them to confront the diverse childhood factors that may have led them to initiate substance use.