The Association Between Reported Childhood Sexual Abuse and Breastfeeding Initiation

Abstract

This study examined the association between self-identified childhood sexual abuse and breastfeeding initiation. A nationally representative sample of 2017 parents with children younger than 3 years was surveyed by telephone about child-rearing needs. Respondents were asked to report childhood sexual abuse and breastfeeding practices. Responses of 1220 biological mothers were analyzed. A possible association between self-reported childhood sexual abuse and breastfeeding initiation was investigated through multivariate logistic regression. Seven percent of the respondents reported experiencing childhood sexual abuse. Women who reported childhood sexual abuse were more than twice (adjusted odds ratio = 2.58; 95% confidence interval = 1.14, 5.85; P = .02) as likely to initiate breastfeeding compared with women who did not report childhood sexual abuse. Parenting attitudes and behaviors were compared to consider whether greater concern with parenting is an explanation for this association. In this nationally representative sample, self-identified childhood sexual abuse is associated with an increased likelihood of breastfeeding initiation.

It is estimated that as many as 20% of US women were sexually abused as children.¹ Many of the long-term effects of childhood sexual abuse have been investigated,² including associations with a range of adult health outcomes such as depression, low self-esteem, substance abuse, and interpersonal problems.^2–6 However, only 1 study has examined the effect of childhood sexual abuse on breastfeeding.⁷

Childhood sexual abuse is associated with several adult characteristics that may decrease the likelihood of breastfeeding. Women who were abused as a child are more likely to experience abuse in their adult relationships^5,8 or during pregnancy.⁹ Their pregnancies are more likely to be unintended.^10–12 Their educational attainment and socioeconomic status also tend to be lower than nonabused women.^4,13 Partner abuse,¹⁴ unintended pregnancy,¹⁵ and low socioeconomic status¹⁶ are associated with a decreased likelihood to breastfeed. In addition, childhood sexual abuse increases a woman’s risk for mental health difficulties later in life, including depression and lower self-esteem levels,^4,6,7,13,17 which may also affect nursing intentions and duration.^18,19 They may be more likely to feel uncomfortable with various intimate parenting tasks, including bathing a child²⁰ and possibly breastfeeding.²¹

In the only previous study that has examined the relationship between childhood sexual abuse and breastfeeding, Benedict et al surveyed 360 women in Baltimore seeking prenatal care between 1990 and 1993.⁷ They found that breastfeeding initiation tended to be higher among women who experienced childhood sexual abuse relative to women who did not; however, the finding did not reach statistical significance.

To further explore the relationship between childhood sexual abuse and breastfeeding, we used data from the Commonwealth Survey of Parents with Young Children.²² We added childhood sexual abuse to a model that we had previously developed to predict breastfeeding initiation and duration.²³ In our previous study,²³ demographic characteristics such as marital status, education, race-ethnicity, maternal age, parity, and census region of country were found to be significant predictors of breastfeeding initiation. In addition, attendance of childbirth education classes and provider encouragement were also found to be associated with an increased likelihood that a woman will initiate breastfeeding. We also included pregnancy intention²⁴ and social support²⁵ as covariates, given their association with breastfeeding in other studies. We included parenting attitudes in the model because they may mediate the relationship between childhood sexual abuse and breastfeeding. Lastly, we included measures of childhood physical and emotional abuse to examine the independent effect of each type of abuse on breastfeeding initiation.

Materials and Methods

Description of Sample and Survey Procedure

We conducted secondary data analyses using data from the Commonwealth Survey of Parents with Young Children, which was designed by the Commonwealth Fund of New York to assess parenting needs and behaviors of a nationally representative sample of parents with children younger than 3 years in 1995.²² A series of 26-min, in-depth, structured telephone interviews were conducted by trained staff at DataStat, Inc. (Ann Arbor, MI), using a stratified random-digit-dial sample design that was specified by Princeton Survey Research Associates. No fewer than 6 attempts were made at different times to establish contact. Using census tract data on race-ethnicity for stratification, the survey oversampled African American and Hispanic parents so their experiences could be analyzed in detail; however, appropriate weights were used to correct for this oversampling in the final analysis to provide a nationally representative sample. Male respondents to the Commonwealth Survey were excluded from our secondary analysis so we could examine the relationship between the mother’s experience of childhood sexual abuse and breastfeeding initiation. Asian Americans, Native Americans, and women of other or mixed race and ethnicity were excluded from our analysis because of small sample sizes.

Eighty-one percent of the households contacted had a respondent complete the screening questions, and 68% of the respondents deemed eligible by the screening questions completed the entire interview. Thus, the overall response rate was 55%, which generated a sample of 2017 parents with children younger than 3 years. The primary study did not obtain information on nonrespondents. Following exclusion of 697 male respondents and 100 female respondents because of race-ethnicity or missing data, the responses of 1220 women were included in this analysis. The weighted sample consisted of 73% white women, 14% African American women, and 13% Hispanic women. Eighty-one percent of the women were married, 6% were divorced, and 14% were single. Fifty-six percent of the respondents were between ages 25 and 34, 12% were over 35, and 32% were younger than 25. Thirty-three percent of respondents received only a high school education, and 15% did not attend high school; 52% had received some college or graduate education. Twenty-seven percent of women reported an annual household income of less than $20,000; 43% reported an income of greater than $40,000. The proportions of respondents who resided in the Northeast, North Central, South, and West regions were 20%, 25%, 34%, and 21%, respectively.

Primary Variables

The primary independent variable for our analysis was whether or not the mother reported having experienced sexual abuse as a child. Respondents were asked, “When you were growing up, did you feel you were ever sexually abused, or not?” The survey question relied upon the respondents to self-identify their experience of abuse. It did not gather details regarding the experience of the abuse such as duration, severity, age at time of incident, or type of perpetrator. Nor did the questionnaire inquire about current abuse. This type of measure for sexual abuse probably underestimates the prevalence of sexual abuse because respondents are asked to identify their experience as sexual abuse and are given only a single opportunity to do so.¹

The primary outcome variable for our study was breastfeeding initiation. Respondents were asked “Was [your child] breastfed more than a month, less than a month, or not at all?” For breastfeeding initiation, the responses were dichotomized to whether or not the respondent breastfed, irrespective of duration. We also examined breastfeeding duration of greater or less than 1 month among those who initiated breastfeeding. Demographic characteristics and obstetric histories were also collected.

Bivariate and Multivariate Analyses

Data were analyzed using SAS (SAS Institue, Inc, North Carolina) and STATA (Stata Corporation, College Station, TX). Bivariate comparisons were made using the Pearson chi-square statistic. In cases of small cell sizes (< 5), the validity of the Pearson chi-square results was verified using a weighted logistic regression analysis in STATA as an analogous test of independence.

Multivariate analysis was performed to assess the independent effect of childhood sexual abuse on breastfeeding initiation.²⁴ We adjusted for variables known to be associated with breastfeeding.^23–25 Covariates included in the final model include self-identified childhood sexual abuse, marital status, education, race-ethnicity, maternal age, parity, having prenatal care insurance coverage, having childbirth delivery insurance coverage, census region of country, attendance at a childbirth class, encouragement to breastfeed by doctors or nurses, intent of pregnancy, proxy measures of social support and parenting attitude, and self-identified childhood physical and emotional abuse. Only variables associated with breastfeeding initiation which had a P value of < .25 in the bivariate analyses were considered for the multivariate analysis, and therefore prenatal care, mode of delivery, and medical problems in the first 2 weeks after delivery were excluded from the final model.²⁶

Household income was omitted from the model to avoid problems associated with missing data and multicollinearity with several other variables. Education, government assistance, and insurance coverage were used as proxy estimates for income. Government assistance included receiving assistance from any type of government programs such as Aid to Families with Dependent Children, Medicaid, and the Special Supplemental Food Program for Women, Infants, and Children. Maternal age at the time of breastfeeding initiation was calculated by subtracting the child’s age from maternal age at the time of the interview. Marital status, region of the country, and education were collected on respondents at the time of the interviews, and not during breastfeeding. A proxy measure of social support (asking how many people lived nearby to whom the respondent could turn for help if a problem arose) was included in the model. A proxy measure of parenting attitude (asking whether or not the respondents wanted to raise their children differently) was also included. Finally, our tolerance calculations did not find multicollinearity among the different types of child abuse, and so childhood emotional and physical abuse were included to assess if other types of abuse have an independent effect on breastfeeding initiation. Childhood emotional and physical abuse were measured in the same manner as childhood sexual abuse.

Covariates were considered to be significant predictors of breastfeeding initiation if their P values were less than or equal to .05. All covariates were entered as a single block into the logistic regression model.²⁶ Multivariable analysis using the same logistic regression model was also performed to assess the relationship between childhood sexual abuse and breastfeeding duration.

Results

Of the 1220 women (weighted n = 1049) included in this analysis, 66% initiated breastfeeding and 34% never initiated breastfeeding. Of those who initiated breastfeeding, 19% breastfed for less than 1 month and 81% breastfed for more than 1 month. While approximately 70% of white and Hispanic women initiated breastfeeding, only about 40% of African American women did so. Women were significantly less likely to breastfeed if they were under the age of 25, single or divorced, had a household income of less than $20,000, resided in the South, received government assistance, or did not attend college or childbirth classes (P < .0001 for all demographic factors).

Overall, 7% (weighted n = 72) of the biological mothers reported they experienced sexual abuse during childhood. This is a lower prevalence rate than would be expected if multiple questions were used to assess sexual abuse or if women were not relied upon to identify their experiences as “abuse.”¹ However, the prevalence rate is similar to other studies that have used an analogous methodology to measure sexual abuse.^27,28 In addition, 11% reported childhood physical abuse and 26% reported childhood emotional abuse. Women who reported sexual abuse as a child were significantly more likely to be divorced, separated, or never married, have lower incomes and receive government assistance. They were also more likely to have an unwanted pregnancy and to start prenatal care late. There was a significant correlation between childhood sexual abuse and other types of child abuse (Table 1).

Table 1.

Demographics, Breastfeeding Influences, and Abuse Histories of Women Who Did and Did Not Self-Identify Experiencing Childhood Sexual Abuse

Independent Variable	Self-Identified Childhood Sexual Abuse	Did Not Self-Identify Childhood Sexual Abuse
	n (%)	n (%)
Marital status (n = 1051)*
Married	50 (70)	796 (81)
Divorced/separated or never married	21 (30)	183 (19)
Age (n = 1053)
< 25	28 (39)	314 (32)
25–29	24 (34)	303 (31)
30–34	16 (22)	241 (25)
35+	4 (5)	123 (13)
Race-ethnicity (n = 1056)
Caucasian (non-Hispanic)	54 (76)	717 (73)
African American	7 (10)	136 (14)
Hispanic	11 (15)	131 (13)
Region (n = 1056)
Northeast	15 (21)	195 (20)
North central	15 (21)	246 (25)
South	24 (33)	334 (34)
West	18 (25)	210 (21)
Education (n = 1054)
< High school	10 (14)	150 (15)
High school grad	28 (39)	314 (32)
Some college	25 (35)	285 (29)
College grad+	8 (11)	232 (24)
Household income (n = 961)**
< 20,000	29 (41)	235 (26)
20,000–39,999	23 (33)	260 (29)
≥40,000	18 (26)	397 (45)
Government assist (n = 1050)**
Yes	46 (64)	432 (44)
No	26 (36)	546 (56)
Initiated breastfeeding (n = 1049)*
Yes	56 (77)	633 (65)
No	16 (23)	344 (35)
Breastfeeding duration (n = 688; includes only women who initiated)
< 1 month	15 (27)	116 (18)
> = 1 month	41 (73)	517 (82)
Parity (n = 1056)
Primiparous	29 (40)	417 (42)
Multiparous	43 (60)	567 (58)
Wanted pregnancy (n = 1056)†
Yes	33 (45)	661 (67)
No	39 (55)	323 (33)
Prenatal care (n = 1044)*
1st trimester	62 (86)	909 (94)
After 1st trimester or did not receive prenatal care	10 (14)	63 (6)
Attended childbirth classes (n = 1054)
Yes	46 (64)	705 (72)
No	26 (36)	277 (28)
Childhood emotional abuse (n = 1046)†
Yes	54 (76)	219 (22)
No	17 (24)	756 (78)
Childhood physical abuse (n = 1050)†
Yes	44 (61)	70 (7)
No	28 (39)	908 (93)
Want to raise† children differently (n = 1050)
Same way I was raised	4 (19)	468 (48)
Raise differently	58 (80)	464 (47)
Same in some ways; different in other ways	1 (1)	47 (5)

All numbers and percentages reported in Table 1 are weighted. Cell numbers and percentages are rounded to the nearest integer.

^*P < .05, chi-square test.

^**P < .01, chi square test.

^†P < .001, chi-square test.

A significantly greater proportion of women who reported childhood sexual abuse initiated breastfeeding, compared with women who did not identify such abuse history (77% vs. 65%). The adjusted odds ratio for initiating breastfeeding among women who reported experiencing sexual abuse as a child was 2.58 (95% CI = 1.14, 5.85; P = .02) (Table 2). Marital status (married), education (having some college or more), race/ethnicity (Caucasian, Hispanic), region of the country (West), attending a childbirth class, being encouraged to breastfeed by a doctor or nurse, and reporting having no one to turn to if a problem arises were associated with a significantly increased likelihood of breastfeeding initiation. All variables except for the proxy measure for social support (having no one to turn to) follow trends found in the literature.^{15,23–25,29–30} However, because the information measuring social support was collected at the time of the interview, not at the time of breastfeeding, this result should be interpreted with caution. Variables that are not significantly associated with breastfeeding initiation include maternal age, parity, receipt of government assistance, prenatal care or childbirth insurance coverage, wantedness of the pregnancy, intent to raise children differently, and self-identified childhood emotional or physical abuse.

Table 2.

Adjusted Odds Ratio (AOR) of Significant Predictors of Breastfeeding Initiation in Multivariable Logistic Regression

Independent Variable (Weighted n = 974)*	AOR (95% Confidence Interval)
Childhood sexual abuse (c.f. no sexual abuse)†	2.58 (1.14, 5.85)
Marital status (c.f. never married)
Married	1.71 (1.03, 2.83)
Education (c.f. < high school)
Some college education	2.39 (1.28, 4.49)
College graduate or beyond	6.23 (2.90, 13.40)
Race/Ethnicity (c.f. African American)
Caucasian (non-Hispanic)	1.79 (1.06, 3.01)
Hispanic	2.54 (1.34, 4.80)
Region of country (c.f. South)
West	2.57 (1.48, 4.47)
Attended childbirth class (c.f. did not attend)	1.79 (1.17, 2.72)
Encouraged to breastfeed (c.f. was not encouraged)	4.52 (3.01, 6.78)
Social support (c.f. many people to turn to if problem arises)
None	2.74 (1.19, 6.33)

Nonsignificant predictors that are also in the model include being divorced or separated, having a high school education, maternal age (25–29, 30–34, and 35+ years old versus < 25 years old), not receiving government assistance versus receiving government assistance, having insurance coverage that paid for prenatal care in part or full versus not having insurance coverage, having childbirth delivery insurance coverage that paid for delivery in part or in full versus not having insurance coverage, living in the Northeast or North central census regions, wanting the pregnancy versus not wanting the pregnancy, being primiparous versus multiparious, having few people to turn to when a problem arises, wanting to raise children differently than you were raised or wanting to raise children the same in some ways but different in other ways than you were raised versus raising your children the same way you were raised, reporting experiencing childhood emotional abuse versus not, and reporting experiencing physical abuse versus not.

Among women who initiated breastfeeding, proportionately more women who reported no history of childhood sexual abuse (82%) breastfed for more than 1 month, compared with women who identified a history of childhood sexual abuse (73%). This difference, however, was not statistically significant in our bivariate (Table 1) and multivariate analyses (data not shown).

A better understanding of how women who reported childhood sexual abuse differed from women who did not report childhood sexual abuse may help explain the association between childhood sexual abuse and breastfeeding initiation. One hypothesis is that women who have self-identified their childhood experience as sexual abuse may be more concerned about parenting. If this hypothesis is valid, self-identified sexually abused women may be more likely to choose to breastfeed since breastfeeding is often perceived by women as a healthier way to feed infants.¹⁹ Therefore, we compared the 2 groups on a number of parenting attitudes and behaviors through Pearson chi-squares. There were only a few significant differences in parenting behaviors between these groups. Women who reported sexual abuse as a child were significantly more likely to have attended a class or discussion about parenting and were also more likely to comfort their child immediately when he or she starts crying (P = .02, P = .01). Other behaviors bordered on significance, including using the media to get information about parenting (P = .10) and discussing a child’s crying with a health professional (P = .10). However, numerous parenting attitudes and behaviors were not significantly different between women who reported childhood sexual abuse and those who did not, such as the frequency with which they read to, played with, sang to, played music with, or hugged their child in the past week; their interest in learning more about various aspects of child development and behavior; and their confidence in and their ability to cope with parenting.

Discussion

Our study suggests that self-identified childhood sexual abuse was associated with a significantly increased likelihood of breastfeeding initiation. Specifically, women who reported experiencing sexual abuse as a child were 2.6 times as likely to initiate breastfeeding as women who reported no such history of abuse. The direction of the association between breastfeeding and self-identified childhood sexual abuse is contrary to what may be expected in light of the correlation between child abuse and risk factors for noninitiation of breastfeeding. The association between self-reported childhood sexual abuse and breastfeeding initiation in this study is in the same direction as that reported in the only previous study⁷ to consider this relationship. Their finding, however, did not reach statistical significance. Despite the relatively small number of women who reported sexual abuse in our study, the reasonable narrow confidence interval indicates that the adjusted odds ratio is fairly precise.

One possible explanation for the association between childhood sexual abuse and breastfeeding initiation is that childhood sexual abuse survivors may be more concerned about parenting than non–sexually abused women. That 80% of sexually abused women want to raise their children differently than how they were raised, as compared with less than half of non–sexually abused women, suggests that they may be more attentive to parenting their child. However, overall there were very few differences in parenting attitudes and behaviors between women who reported childhood sexual abuse and those who did not. It remains unclear whether or not heightened concern about parenting among childhood sexual abuse survivors may account for their increased likelihood to breastfeed.

It is intriguing that self-identified childhood sexual abuse was significantly associated with breastfeeding while childhood emotional and physical abuse were not. Although studies have noted similarities in the long-term effects of the 3 types of abuses, childhood sexual abuse may be uniquely associated with sexual problems such as lower sex drives or fear of sex, sexual mal-adaptive behavior, and sexual disturbance in adult relationships.^3,4,31 Douglas also found women with a history of childhood sexual abuse were significantly more anxious about performing intimate parenting tasks, such as bathing their children, than women without such a history.²⁰ One might therefore expect breastfeeding to be more closely associated with sexual abuse than other types of abuse, given the intimacy of breastfeeding, but the positive association in our study is unexpected. It is also interesting to note that self-identified childhood sexual abuse was not significantly related to breastfeeding duration among women who initiated breastfeeding. This nonsignificant finding may be attributable to smaller sample size, or it may suggest that factors mediating the relationship between childhood sexual abuse and breastfeeding initiation may have less influence on breastfeeding duration.

Because children who are victims of sexual abuse are often abused physically and emotionally as well, we attempted to examine the synergistic and cumulative effects of experiencing multiple types of childhood abuse using interaction terms. However, due to the small number of women reporting multiple types of abuse in this study, we were unable to investigate these effects. It remains to be shown by future research how different types of child abuse affect breastfeeding independently and synergistically.

There are several limitations to our study. First, the primary survey question assessing childhood sexual abuse is limiting because women were required to self-identify their experience as sexual abuse. The lower prevalence of childhood sexual abuse reported in this study (7%), as compared with that reported in other studies that specified actions that constituted sexual abuse or used multiple questions to encourage disclosure of sexual abuse (20%), may reflect in part the imprecision of the measure of childhood sexual abuse.^1,32 Thus, the control group may be contaminated by women with a hidden history of childhood sexual abuse. Women who are unaware of or conceal their child abuse history may behave differently from those who self-identify such history in a survey. The psychological coping mechanisms may also differ between the 2 groups of childhood sexual abuse survivors. Women who are unaware of or conceal their history may be more likely to employ denial³³ or dissociation³⁴ to cope with their childhood traumas; it is not known how these coping mechanisms influence breastfeeding practices. Our primary data set does not allow us to distinguish between those who are unaware of or conceal their history from those who are able to self-identify their history. If women who are unaware of or conceal their history of childhood sexual abuse are less likely to breastfeed, our finding of a positive association may be a spurious artifact.

Second, women were being asked to recall breastfeeding up to 3 years before the survey. Recall error may occur, though studies have shown that long-term maternal recall of perinatal events, including breastfeeding, is fairly accurate and reproducible.³⁵

Third, the survey question did not clearly define breastfeeding initiation. Getting the baby to the nipple is different from one or more successful breastfeedings a day. Neither did the study define sexual abuse in terms of duration and severity. Since duration and severity may influence the long-term impact of childhood sexual abuse,^5,36 it is not unreasonable to speculate that abuse of different duration and severity may affect breastfeeding differently.

Fourth, although we controlled for confounding using several factors, we were not able to control for other important risk factors such as current experiences of abuse, nor were we able to control for childhood factors other than emotional and physical abuse. Breastfeeding may be associated with characteristics in the family of origin, or other factors along the life course, which are also associated with childhood sexual abuse.

Lastly, the primary study did not ask why women chose to breastfeed, nor did it explore the psychological aspects of breastfeeding among women who were abused as a child. Further studies are needed to investigate the mechanism for the association between childhood sexual abuse and breastfeeding.

In conclusion, we found a positive significant association between women who self-identified themselves as sexually abused during childhood and breastfeeding initiation. To our knowledge, this is the first report of a significant association between childhood abuse and breastfeeding behavior based on a national sample. Further research is needed to gain a better understanding of the association.