|Home | About | Journals | Submit | Contact Us | Français|
To estimate the association between number of parents in the childhood home and childhood sexual abuse (CSA) with adjustment for childhood socioeconomic status (CSES).
Probability sample of 298, 18–49‐year‐old men from Philadelphia County, number of parents living in childhood home, socioeconomic data and CSA histories were obtained.
197 (66%) men participated. 186 (94%) of these lived with at least one parent; 76 (39%) and 110 (56%) lived with one parent versus two parents, respectively. 22 (29%) of 76 and 18 (16%) of 110 reported CSA histories, respectively (OR 2.08, p=0.04). Two approaches to adjustment for CSES indicated continued association between parent number and CSA (OR 2.38–2.39, p=0.05–0.07). Parent number was associated with numerous differences in CSA perpetrator characteristics and abuse experiences. Men from one‐parent versus two‐parent families reported significantly more non‐family and female perpetrators (p=0.03 and 0.01, respectively) and fondling experiences (p=0.04).
Findings provide additional support for the association between parent number and CSA in boys, suggesting that parent number is not just a proxy for CSES. CSA experiences also differed between one‐parent and two‐parent homes. Findings generate numerous hypotheses for future study.
Sexually abused males often come from childhood homes without two married biological parents present.1,2,3,4 Two studies have documented that this high one‐parent rate is significantly higher for males with childhood sexual abuse (CSA) histories than for non‐abused males, suggesting that one‐parent households may confer an increased risk for CSA in boys.5,6 One‐parent homes may confer risk by increasing total parental absence (ie, given only one parent and the likelihood that parent must work and do household errands, it is likely that children from such households are more often in a home situation where the single parent is absent).
Parental absence has been conjectured to increase CSA risk because it leads to less protection of children from exploitative adults.7 Some have suggested that children with absent parents, in particular, need more rather than less protection. These children may have higher emotional needs that lead them to seek out attention and affection from non‐parent adults, which they do in environments that are conducive to being exposed to exploitative adults who can take advantage of the children's attention‐seeking and affection‐seeking needs.7,8,9
Authors in areas other than sexual abuse suggest that socioeconomic disadvantage may explain much of the non‐CSA risk seen for children from one‐parent households.10,11 Studies indicating an increased CSA risk in boys from one‐parent homes have not adjusted for this socioeconomic disadvantage in their analyses. This is concerning, as several studies in males have reported an association between CSA and socioeconomic status (SES).12,13,14 Thus, the question of whether the effect of parent number on the likelihood of boys experiencing CSA is independent of socioeconomic variables has not been established.
The aim of this study was to confirm in a probability sample of men that having grown up in a one‐parent household was associated with a higher rate of having experienced sexual abuse, using adjusted analyses that included assessments of SES. A secondary aim was to determine whether the potential influence of parent number on likelihood of boys experiencing sexual abuse could be informed by assessing differences in the characteristics of sexual abuse experiences by whether a participant had come from a one‐parent versus two‐parent home.
We identified potential study participants by random digit dialling using telephone exchanges for Philadelphia County areas with high incidences of reported HIV seroconversion.15 In all, 100 men were to be recruited into each of three age groups: 18–29, 30–39 and 40–49 years. A research firm with expertise in telephone surveys on sensitive subjects completed all screening and interviewing. Interviewers recorded contact information for men who agreed to participate in a study of how “childhood experiences have affected adult men's health and well‐being”.
Of 13116 phone numbers called, screening was initiated with 2418 households. Of these, 1975 did not include a qualified potential participant and 145 did not complete the screening, resulting in 298 men who were eligible and expressed interest in participation. An information packet describing the study and principal investigator, offering a $15 incentive for study participation, and including a consent form (originally approved on 24 July 2002 by the University of Pennsylvania Institutional Review Board's Committee on Studies Involving Human Beings) was posted to these potential participants. These men were called back 2 weeks later and a full telephone interview was administered to those agreeing to participate.
After routine sociodemographic questioning (ie, age, race), men were asked, “Did you live with both of your parents, your mother only, your father only or someone else most of your childhood (up until you were 18 years old)?”. Possible response options were “Both (at the same time)”, “Both (serially, such as in shared custody arrangement)”, “Mother only”, “Father only” or “Someone else”. Those who chose “Both (serially, such as in shared custody arrangement)”, “Mother only” and “Father only” were placed in the “one parent” subgroup. Rerunning analyses with participants in shared custody removed from this subgroup did not change results.
The measurement of SES of participants' childhood homes (childhood socioeconomic status CSES) began with participants' responses to the question “What was your father's/mother's primary or usual occupation when you were growing up (ie, what occupation did she/he work at the longest during your childhood)?”. Responses were used to compute a gross estimate of likely earnings per year per parent. As each participant had lived their childhood during different time frames, most occupations' earnings were estimated by forward‐discounting them to median yearly earnings that were obtained from the sex‐specific and state‐specific 2000 Census “Earnings by Detailed Occupation: 1999” tables.16 For these estimates, parents were assumed to have worked in the occupation full‐time, year‐round and in Pennsylvania. Earnings for variations on “military service” were set at an E‐7 enlistee level who had 14 years of service (income was available for Fiscal year (FY) 2005 and discounted to FY 1999 by 0.03/year).17 Earnings for variations on “housewife” were set to zero. Variations on “civil servant” were set to missing, as was no stated occupation at all. Both parents' earnings were summed to achieve total parental earnings in two‐parent families. Only the earnings for the parent with whom the participant stated he had lived were used to arrive at total parental earnings in one‐parent families.
Participants also had been asked, “How many brothers/sisters did you have while you were growing up?”. Responses to this question were used to arrive at a total number of individuals living in the household (the parent/parents + the siblings + the participant).
Total earnings by total number in household were used to characterise CSES as being 200% versus >200% of the federal poverty line (200% FPL) according to the 1999 Federal Poverty Guidelines (because 1999 is the year for which 2000 Census earnings refer).18 The CSES variable was dichotomised to reflect the bluntness of its measurement. The 200% FPL multiple was used to reflect its widespread use to capture the “working poor”.19,20
Finkelhor's four “funnelling” questions were asked to screen for CSA.5 Any “yes” response led to additional questions that detailed who and how old the potential perpetrator was at the time of the experience(s) that had led to a “yes” response, what occurred at the initial experience and any subsequent experience(s) that might have occurred with the potential perpetrator, how old the respondent was at the initial experience (and at the final experience if more than one event had occurred), how many experiences with this potential perpetrator had occurred altogether (if there was more than one event), whether there were other people than the potential perpetrator involved in the experience(s), whether the potential perpetrator(s) ever used force (such as restraint or a weapon), or any other form of coercion or threat, and whether the respondent defined the experience(s) to be sexual abuse.
Later in the interview, the respondent also was asked, “How old were you the first time you had any type of sexual experience with another person willingly?”. If earlier in the interview they had answered the sexual abuse questions affirmatively, this question was introduced with the phrase, “Not counting the experience(s) you mentioned in the last section”. After this question they were asked with whom they had had their first willing sexual experience, who/how old this person was and what occurred.
CSA was defined as any sexual experience before 18 years of age (even those characterised as “willing”) in which: (1) a power differential existed between a victim and perpetrator—that is, where the perpetrator was 5 years older than a victim of <13 years, was 10 years older than a victim of 13–17 years or was an authority figure (ie, teacher); (2) coercion was self‐reported to have occurred; or (3) penetration (ie, oral, anal or vaginal) of victim by perpetrator or perpetrator by victim occurred when the victim was prepubertal (11 years) and perpetrator was postpubertal (>12 years) and at least 2 years older than the victim.21,22 Three participants were considered to have met the above CSA definition although they did not report the age of their perpetrator; a power differential based on age difference was presumably met given that they reported experiences (at ages 6, 15 and 15 years) with a stranger who they described as an “adult” (but whose exact age they did not know). Two other participants did not meet any of the definition's criteria but were included: the first was 16 years old and his perpetrator was 21 years old at the time of the experience and he defined his experience to have been sexual abuse “[b]ecause someone was touching my genitals without my permission”; and the second was 16 years old, his female perpetrator 22 years old, and he defined his experiences to have been sexual abuse “because this person knew that I was 16 years old” (presumably implying that she knew he was too young to consent or refuse). Results were not changed when analyses were rerun with the latter two participants included in the “not abused” subgroup.
Group comparisons were performed using two‐tailed t tests for continuous variables and χ2 methods (or Fisher's exact test when expected frequency for one subgroup was <5) for categorical variables. Two forced‐entry multivariable logistic regressions were completed to assess the adjusted association between parent number and CSA, where all variables that either were related to the adjustment for the potentially confounding nature of SES on the relationship between parent number and CSA or were associated with parent number or with CSA (p<0.10) were included (dummy variables were used for ordinal data). Two regressions were completed to allow different approaches to adjusting for CSES: the first, indirect approach used participants' race/ethnicity, educational attainment and income as possible proxy indicators for CSES, and the second, more direct approach used the CSES variable described above. Both approaches were used because both had flaws. The first approach was flawed because educational attainment and income could be less a proxy for CSES than they are outcomes of CSA itself. The second approach was flawed because of recall bias and numerous assumptions (ie, full‐time work), and many missing data. The goal for using both approaches to adjustment was to ascertain whether there was consistency across different methods. Given concerns about overfitting a model, and given that race/ethnicity, educational attainment and income are often associated with CSES (which was the case with data from this study, results not shown), race/ethnicity, educational attainment and income were not entered into the second regression with CSES. SPSS V.12.0 was used to manage and analyse data.
Of the 298 men who were recruited, 99 were either subsequently unwilling to participate or could not be recontacted, resulting in a sample of 197 men (66% participation rate of eligibles). Of the three variables assessed at screening in all 298 potential participants, two variables (age, p=0.78 and race, p=0.88) did not differ by participation status. There was a nearly significant (p=0.06) difference in educational attainment, however, with more participants reporting higher attainment than non‐participants (20% vs 22% with <12th grade/ Graduate Equivalency Diploma, 24% vs 38% high school graduates, 20% vs 18% with some college education and 35% vs 23% with a college and/or graduate degree).
Of the participants, 66, 62 and 69 were in the 18–29, 30–39 and 40–49‐year‐old age groups, respectively. The race/ethnicity distribution was African‐American (54%), Caucasian (32%), Hispanic (9%), Asian/Pacific Islander (2%), “Mixed” (2%) and “Arab” (1%). About 8% self‐identified as gay/bisexual. Education ranged from <12th grade/Graduate Equivalency Diploma (20%), high school graduate (24%), some college education (20%) to college/graduate degree (35%). Income ranged from $20000 (29%), $20001–40000 (27%), $40001–75000 (27%) to >$75000 (18%).
A total of 76 (39%) and 110 (56%) men lived with one parent versus two parents, respectively; 11 (6%) men had lived with someone other than parents (no results are reported hereafter for those who did not live with a parent). Of the men who had lived with one parent only, 64 (84%) had lived with their mother, 7 (9%) with their father and 5 (7%) with both, albeit serially (ie, custody arrangement). Table 11 shows that variables for age (p=0.02), race/ethnicity (p<0.001), educational attainment (p=0.005) and childhood socioeconomic status (CSES; p<0.001) were associated with one‐parent versus two‐parent status.
Of the 186 participants 40 (22%) met the criteria for CSA. Table 22 shows that variables for age (p=0.03), race/ethnicity (p=0.04) and sexual identity (p=0.001) were associated with CSA history.
In all, 22 (29%) of the 76 and 18 (16%) of the 110 men from one‐parent versus two‐parent homes reported CSA histories, respectively (OR = 2.08, p=0.04). CSA histories of men who lived with one parent were compared according to whether they lived with their mother, their father or both parents serially. As rates were not significantly different (18/64 vs 2/7 vs 2/5, respectively; p=0.85), these three one‐parent categories remained combined for all subsequent analyses.
Table 33 shows that after adjustment both for CSES using the indirect approach of race/ethnicity, education and income, and for variables from fromtablestables 1 and 22 with p values 0.10, the association between parent number and CSA was strengthened (OR=2.39, p=0.05) compared with the bivariate results. When direct adjustment for CSES was used (table 33),), the association between parent number and CSA was unchanged from the indirect approach, although the p value was somewhat attenuated (OR=2.38, p=0.07).
Age, overall, was not significantly associated with CSA in either model (p=0.17–0.20). Sexual identity was substantially associated with CSA in both models (OR=4.8–6.85, p=0.03–0.003): 6 of 7 (86%) gay/bisexual men and 12 of 30 (40%) heterosexual men reported a male perpetrator (p=0.04).
Given the significant bivariate association between parent number and CSES (table 11),), consideration had been given to the possibility that an interaction term was necessary in the second model of table 33.. This question was explored by analysing the bivariate associations between parent number and CSA for separate CSES strata. This indicated that the association between parent number and CSA was substantial for the low‐SES subgroup (OR=4.41, 95% CI=1.16 to 16.72; p=0.001), but not for the high‐SES subgroup (OR=0.95, 95% CI=0.23 to 3.82; p=0.99). However, the Breslow–Day test for homogeneity of the odds ratios (ORs) did not indicate effect modification (p=0.11), so an interaction term was not included in the reported model.
For exploratory purposes (and given that the p value was tending towards significant), however, the second model from table 33 was reanalysed with an interaction term to allow assessment of stratum‐specific ORs using adjusted beta coefficients. These indicated that associations remained substantial for the low‐SES subgroup (OR=4.97, 95% CI=1.215 to 20.35; p=0.03), but not for the high‐SES subgroup (OR=0.99, 95% CI=0.22 to 4.48; p=0.99).
Table 44 shows the characteristics of reported CSA experiences according to whether participants lived with one or both parents. Variables for perpetrator sex (p=0.01), perpetrator a family member (p=0.03), fondling (p=0.04) and oral sex (p=0.02) were associated with one‐parent versus two‐parent status. These findings suggested that men who lived with one parent were more likely to have been sexually abused by a female, and by someone not from their family; abuse more often involved fondling of the victim and/or perpetrator. Men who lived with two parents were more likely to have been sexually abused by a male, and by someone from their family (although most perpetrators still were not from their family); abuse more often involved oral sex of the victim and/or perpetrator.
Perpetrators reported by men from one‐parent homes included one babysitter's 16‐year‐old daughter, one babysitter's 26‐year‐old son and two female babysitters (15 and 16 years old). Perpetrators reported by men from two‐parent homes included two female babysitters (16 and 17 years old). Although there was a trend for more babysitter‐related perpetrators in men from one‐parent versus two‐parent homes (19% vs 11%, respectively), these differences did not reach statistical significance.
Findings reported for this probability sample of men recruited from a high‐risk urban locale further support a new literature indicating that males from one‐parent families are at increased risk for CSA. Results indicate that having been raised in a one‐parent family significantly increased, to more than double, the odds of being sexually abused as a boy. The fact that this association comports with a larger literature that indicates that children from one‐parent homes have significantly higher rates of poor outcomes, more unstable metabolic control in diabetics, higher morbidity and mortality, more injuries, more frequent child abuse reports, more psychiatric problems and suicide attempts, greater social impairment and lower educational attainment, among other outcomes, adds plausibility to the finding.23,24,25,26,27 Furthermore, the association between parent number and CSA remained significant even after adjustment (using two different approaches) for CSES, a finding not previously reported.
While results indicate that parent number is not a proxy for CSES, it remains challenging to consider it an independent risk factor. It may act as a proxy for parental absence instead, as has been hypothesised previously.7,8,9 It could also act as an indicator for other risk factors and/or dynamics, such as likelihood of family violence (ie, more violence in one‐parent families).28 Higgins and McCabe29 have indicated the importance of identifying family violence in studies of CSA sequelae, but this importance is likely to be true for risk factor studies as well. Alternatively, or in addition, parent number could be an indicator of more alcohol and/or drug (mis)use in the home (ie, more alcohol drinking in one‐parent homes).30 Parental alcoholism and drug misuse have been reported to be significantly higher in males with versus those without CSA histories.31,32,33,34 Parental absence, other family violence and alcohol/drug use were not assessed in this study and were not included in adjusted models. Future studies should do that.
Future studies should not, however, dismiss the potential importance of CSES. There was a strong suggestion that the association between parent number and CSA might only be occurring in lower socioeconomic strata, a finding consistent with prior reports on physical violence in one‐parent families.28 While the study's analyses did not confirm this stratum‐specific finding, the statistical power may have been too limited to do so. Future studies of parent number should be powered to investigate this question of CSES effect‐modification on CSA risk more conclusively.
There were clear differences in CSA experiences by parent number, and these findings are another new contribution of this study. Prior reports of high CSA rates in one‐parent families (families that are often headed by single mothers) led to conjecture that stepfathers and boyfriends were the perpetrators of this abuse.1,4 Men with CSA histories from one‐parent families in this sample, however, reported significantly more female perpetrators who were not from their family than did men with CSA histories from two‐parent families. Although these rates of female perpetration are not substantially different than those reported in prior studies of adolescent male samples, they are in the higher range of female‐perpetration rates reported for adult male samples.35 This could suggest that abused men from one‐parent homes are not as likely to redefine sexual abuse perpetrated by women as sexual initiation as many males are conjectured to do as they mature from adolescents to adults.35
The predominance of female perpetrators in men from one‐parent families may suggest that findings are related to childcare needs in one‐parent homes. Perpetrators of men from one‐parent homes tended towards being younger and more often babysitters or family members of babysitters. These trends, however, did not meet statistical significance. Future studies should better characterise the time spent in the care of non‐parents, the type and quality of these non‐parent arrangements, and their relationship with CSA history. Non‐significant trends could be meaningful in future studies in which babysitter‐related information is specifically obtained (it was only volunteered by respondents in the current study). If such a hypothesis was confirmed in future studies, it would support the possibility that functional support in one‐parent versus two‐parent families may be met by non‐family members, which in turn may expose boys in one‐parent families to an environment that has a higher risk for abuse. Future confirmation of CSES stratum‐specific differences, using more participants and a CSES assessment approach that allows a less blunt measure of CSES, also might suggest that wealthier one‐parent families are able to pay for better functional support and/or require less of it. Confirmation of this series of hypotheses would suggest the need for: future interventions that improve single parents' skills in identifying low‐risk individuals to provide functional support and that improve single parents' oversight of these individuals' interactions with children in the home, and public policies that provide better options and more support for child care, particularly for those near and under the FPL.36
The findings indicate an association between sexual identity and CSA that is independent of parent number and CSES. It is not clear whether this association points to a risk factor, an outcome, both or something else. As has been suggested previously, some young men may explore their emerging gay/bisexual identity in venues where abuse may happen more frequently and/or may be targeted by some perpetrators because of actual or presumed sexual identities.35 Alternatively, and/or in addition, the association between CSA and sexual identity may be causal in nature, although no longitudinal studies have confirmed let alone explored this possibility. Gay/bisexual men also are likely to be more experienced in sexual marginalisation than heterosexual men are, and thus may experience CSA disclosure as less of a marginalising event than do heterosexual men, particularly if CSA was male perpetrated.
Two study limitations warrant specific notice. First, the sample was non‐affluent, urban and largely minority. Thus, results cannot be extended beyond this subpopulation. Furthermore, participants were found to have nearly significant higher educational attainment than non‐participants and, thus, findings may also be biased towards those with higher educational attainment. Second, information about parent number was obtained from a general question asking about with whom participants lived before 18 years of age. As responses were likely to be a summary assessment of participants' entire childhood, and not specifically a characteristic of the household at the time of the sexual abuse, there is the potential for misclassification. Such misclassification would appear to be non‐differential in nature, so reported results are likely to be biased towards the null. Studies of any adult sample, however, will lead to recall bias‐related misclassification, even if participants are specifically asked about parental characteristics at the time of abuse, as was performed by Benedict and Zautra37 in a mixed‐sex sample with few males (in which case misclassification may have been differential in nature if abused participants had given more thought to specific aspects of their childhood histories than did non‐abused participants).
Despite the caution that generalisability of findings may be narrow, an association between number of parents in the home and CSA was found in boys again, this time after adjustment for potential confounders and/or effect modifiers including SES. The twofold increase in CSA related to having lived with one parent versus two parents was probably underestimated due to misclassification. With recent US Census reports indicating that nearly 20 million children live with a single mother or father, that one‐parent families increased from 3.5 million to 12 million over the past 30 years, and that nearly one‐third of one‐parent families were living below the poverty line, the findings reported from the current study highlight the need to pursue the numerous questions that remain about the association between parent number and CSA in boys.38,39
I thank Mary D, Sammel, ScD, Assistant Professor of Biostatistics, University of Pennsylvania School of Medicine, for her statistical input.
CSA - childhood sexual abuse
CSES - childhood socioeconomic status
FPL - federal poverty line
SES - socioeconomic status
Funding: WCH was funded by a grant from the National Institute of Drug Abuse (DA015635) and is a recipient of a Research Career Development award from the Health Services Research & Development Service of the Department of Veterans Affairs.
Competing interests: None declared.