|Home | About | Journals | Submit | Contact Us | Français|
Child maltreatment has been linked to negative adult health outcomes; however, much past research includes only clinical samples of women, focuses exclusively on sexual abuse and/or fails to control for family background and childhood characteristics, both potential confounders. Further research is needed to obtain accurate, generalizable estimates and to educate clinicians who are generally unaware of the link between childhood abuse and adult health. The purpose of this project is to examine how childhood physical abuse by parents impacts mid-life mental and physical health and to explore the attenuating effect of family background and childhood adversities.
We analyzed population-based survey data from 2,800 middle-aged men and women in the Wisconsin Longitudinal Study using self-reported measures of parental childhood physical abuse, mental health (depression, anxiety, anger), physical health (physical symptoms and medical diagnoses), family background, and childhood adversities.
Parental physical abuse was reported by 11.4% of respondents (10.6% of males and 12.1% of females). In multivariate models controlling for age, sex, childhood adversities, and family background, childhood physical abuse predicted a graded increase in depression, anxiety, anger, physical symptoms, and medical diagnoses. Childhood physical abuse also predicted severe ill health and an array of specific medical diagnoses and physical symptoms. Family background and childhood adversities attenuated but did not eliminate the childhood abuse/adult health relationship.
In a population-based cohort of middle-aged men and women, childhood physical abuse predicted worse mental and physical health decades after the abuse. These effects were attenuated by age, sex, family background, and childhood adversities, but not eliminated.
Childhood maltreatment has been associated with an array of psychological and somatic symptoms in adult life (McCauley et al., 1997; Moeller & Bachman, 1993; Mullen, Romans-Clarkson, Walton, & Herbison, 1988; Walker et al., 1999) as well as psychiatric diagnoses including depression (Wise, Zierier, Krieger, & Harlow, 2001), anxiety disorders (MacMillan et al., 2001), eating disorders (Kendler et al., 2000), and posttraumatic stress disorder (Widom, 1999). In addition, research has documented that the combination of multiple types of maltreatment are associated with increased health problems (Felitti et al., 1998). Medical diagnoses associated with childhood maltreatment are primarily syndromes characterized by pain and disability in the absence of firm physical or laboratory abnormalities (Katon, Sullivan, & Walker, 2001) including chronic pain syndromes (Walling et al., 1994), headaches (Goodwin, Hoven, Murison, & Hotopf, 2003), fibromyalgia (Boisset-Pioro, Esdaile, & Fitzcharles, 1995), chronic fatigue syndrome (Taylor & Janson, 2002), and irritable bowel syndrome (Drossman et al., 1990). Despite this strong evidence, many recent reviews in the medical literature have failed to acknowledge childhood maltreatment as an important predictor of adult health problems (Barsky & Borus, 1999; Leventhal, 1999; Mehler, 2001; Whooley & Simon, 2000).
Research has established relationships between multiple forms of childhood maltreatment and adult health. However most of the existing research emphasizes childhood sexual abuse and/or childhood abuse among girls (Irving, 2006; Springer, Sheridan, Kuo, & Carnes, 2003; Widom, 1999) despite the fact that childhood physical abuse is more prevalent than sexual abuse and childhood physical abuse is more likely than sexual abuse to occur in both men and women (US Department of Health and Human Services, Administration on Children, Youth and Families, 2002; US Department of Health and Human Services, Administration on Children, Youth and Families, 2006). In addition, much of the previous research has been conducted on small and/or clinical samples. While these studies are informative, sample selection issues limit the generalizability of the findings.
Furthermore, not all previous studies of childhood physical abuse and adult health have controlled for family background and early childhood adversity variables – potentially important confounders. Childhood physical abuse is seldom an isolated event and only when these other factors are accounted for can one confidently argue about a health effect of childhood physical abuse above and beyond other adversities in family environment. Most of the research that includes multiple childhood adversities and/or family background use the ACE study and examine the combined effect of multiple adversities rather than analyzing the effect of each adversity individually (Anda et al., 1999; Dong, Dube, Felitti, Giles, & Anda, 2003; Felitti et al., 1998; Hillis, Anda, Felitti, & Marchbanks, 2001). While these projects are important in emphasizing the cumulative effect of abuse, it is also essential to understand how specific types of abuse impact adult health net of other childhood adversities and family background.
Therefore, the purpose of the current project is to explore how one specific type of child maltreatment---childhood physical abuse---affects different aspects of adult health in a population-based sample of men and women. To isolate the effects of childhood physical abuse, we account for a variety of family background and childhood adversity measures to rule out the possibility of a spurious association between childhood physical abuse and adult health problems.
Data for this project are from the sibling respondents of the Wisconsin Longitudinal Study (WLS). The WLS is a long-term population-based study of men and women who graduated in 1957 from Wisconsin high schools and their siblings (Sewell, Hauser, Springer, & Hauser, 2003). The sibling respondents (hereafter respondents) were interviewed in 1977 and 1994. The 1994 round of data collection consisted of a one hour telephone interview followed by a 20 page mail survey. Data collection in 1977 focused on education, occupation, and family demographics whereas data collection in 1994 focused on health, well-being and aging, in addition to education, occupation and family. The variables for these current analyses are from the 1994 survey with the exception of family background characteristics from the 1977 survey.
Reflecting the 1957 Wisconsin population, the WLS cohort is almost entirely white of non-Hispanic origin. Although the sample originated in Wisconsin, the 1977 and 1994 interviews were conducted with respondents across the globe. The WLS is approved by the University of Wisconsin IRB; these secondary analyses were exempted from additional human subjects approval.
The 1993–94 mail survey of siblings included questions on childhood physical abuse based on the Conflict Tactics Scale (Straus, Gelles, & Steinmetz, 1981). Respondents were asked separate questions about whether their mothers or fathers “slapped, shoved, or threw things at them” before age 17. The response categories were “a lot,” “some,” “a little,” or “not at all” scaled from 3 (“a lot”) to 0 (“not at all”). Those who reported “some” or “a lot” of abuse by either or both parents as opposed to “a little” or “not at all” were classified as having experienced childhood physical abuse.
Depression was assessed with the Center for Epidemiologic Studies Depression Scale (CES-D) (Radloff, 1977); anxiety and anger were assessed with the Spielberger’s Anxiety and Anger scales (Spielberger, 1996; Spielberger & Gorsuch, 1983; Spielberger, Sydeman, Owen, & Marsh, 1999). The anxiety and anger scales included 10 items each whereas the CES-D included 20 items. Each of the items assessed how many of the days in the past week respondents experienced a specific symptom. Respondents who completed all anxiety and anger items were included in the analyses whereas respondents were included in the depression analyses if they answered at least 10 of the 20 items. To construct each of these scales we divided the total score by the number of items answered, multiplied this value by the number of items in the scale, and rounded this score.
Using a modified version of the Duke Older Adults Research Survey (OARS) (Duke University Center for the Study of Aging and Human Development, 1978) respondents were asked whether “a medical professional has ever said you have…,” followed by the following list of 17 conditions: allergies, anemia, arthritis/rheumatism, asthma, serious back trouble, bronchitis/emphysema, cancer, circulation problems, colitis, diabetes, heart trouble, high blood pressure, high cholesterol, kidney/bladder problems, chronic liver trouble, multiple sclerosis, and ulcer. From a symptoms list created for the WLS, respondents reported whether or not they have experienced any of a list of 27 physical symptoms in the past six months and how often. For this analysis, a respondent was coded as having a symptom if s/he reported experiencing it once a week or more often in the last 6 months.
Six indicators of family background were used: mother’s education (years), father’s education (years), parental income (log transformed), father’s occupational standing (occupational education scores [Hauser & Warren, 1997]), number of siblings, and grew up on a farm. When direct reports were not available because of the sampling design or missing data, reports from the graduate sibling were used if they lived together most of their childhood.
Four indicators of childhood adversity were used: whether the respondents grew up with a problem drinker (parental drinking problem), whether their parents had serious marital problems when they were growing up (parental martial problem), whether their parents slapped, shoved or threw things at each other some or a lot of the time (parental violence), and whether they did not grow up with both parents until 16 years old (broken family). For this cohort, growing up without both parents was most often the result of widowhood.
There were 6,260 siblings approached to participate in the 1994 telephone interview and 4,804 (77%) completed the interview. Of these respondents, 3,501 also completed the mail survey, yielding a 73% response rate for the mail instrument. This project utilizes two samples---one sample controls for a vast array of family background and childhood adversity characteristics and the second sample employs a more streamlined set of controls, as described in the statistical analysis section. The sample containing all controls contains approximately 2,050 cases and the second sample contains approximately 2,800 cases. The decrease from 3,501 to 2,800 cases is largely because many of the family background questions were only asked of a smaller sample of siblings interviewed in 1977. The decrease from 2,800 to 2,051 cases is largely because the question about parental drinking problems was asked of an 80% random subsample of the phone interviews. Almost all of the missing data are due to random subsampling and therefore pose no threat to the validity of the analyses.
Bivariate comparisons of demographic and health measures for abused and non-abused participants were performed individually by one-sided t-tests because we hypothesized increased problems associated with childhood physical abuse. For multivariate models, three types of dependent variables were constructed to assess different dimensions of the childhood abuse/adult health relationship. First, natural log transformed counts of the medical diagnoses and physical symptoms and the total scores of the depression, anxiety, and anger scales were analyzed. Second, non-transformed outcomes cut at the 90th percentile were used to examine extreme ill health. For example, we created a dummy variable indicating whether someone reported more symptoms than 90% of the sample. Third, each medical condition and cluster of physical symptoms by organ system was assessed. Here each organ system grouping was coded as 1 if a person reported any of the symptoms in the group.
Logistic regression was used for dichotomous outcomes with coefficients indicating the increased or decreased odds of the outcome for each independent variable. Ordinary least square (OLS) regression was used to analyze scales of depression, anxiety, anger, number of symptoms and number of diagnoses. Outcome variables in these OLS analyses were transformed with a natural log to reduce the skew of the dependent variable and to correct for violations of the normal residual distribution assumption. Because the outcomes were log transformed, the effect of the independent variables can be understood best by exponentiating the coefficients [%change in y = 100*(exp[B]−1)] to obtain a value interpretable as a percent change in the outcome.
OLS models began with reduced-form regressions of health measures on childhood abuse to obtain the total effects of abuse on each health outcome. Controls for age and sex, family background, and childhood adversity were then sequentially added resulting in a sample of 2,051 cases with complete data medical diagnoses and all controls.
Because most of the family background and childhood adversity measures were correlated and were not significant, a set of the most efficient measures was chosen for models in the Table 3 and and44 using Multiple Indicator and Multiple Cause (MIMIC) models. Examining only these efficient measures (father’s education, farm background and parental marital problems in childhood) resulted in an increase of approximately 750 cases for a total sample size of 2,836 (results available on request).
Descriptive statistics for the sample with complete data on all independent variables and number of illnesses (N = 2,051) are shown in Table 1. Childhood physical abuse was reported by 11.4% of respondents. Over half of the respondents were women, and the mean age of 55 years was similar for abused and non-abused respondents. Abused respondents had fathers with significantly less education and significantly lower occupational education. In addition, abused respondents had significantly more siblings, were more likely to grow up with problem drinkers, more likely to report having parents with serious marital problems, and more likely to report violence between parents than non-abused respondents. Abused respondents reported worse health than non-abused respondents for all five outcomes examined.
The first three columns of Table 2 contain the zero-order OLS models, which show the total effect of abuse on all five health outcome scales. The second three columns contain models controlling only for sex and age, and the last three columns control for sex, age, and all childhood adversities. Only significant effects are presented in Table 2.
In all of the models childhood physical abuse significantly predicts an increase in negative health. In the zero-order model, physical abuse was associated with a .19 increase in the natural log of the number of medical diagnoses. The coefficient .19 can be transformed into a percent change by exponentiation: %change in y = 100*[exp(B)-1] = 100*[exp(.19)−1] = 21%. Therefore, the results in Table 2 demonstrate that physical abuse is associated with a 21% increase in medical diagnoses, 22% increase in the number of physical symptoms, 24% increase in depression, 27% increase in anger, and a 23% increase in anxiety. After controlling for sex and age, the effect of abuse is slightly, but not substantially, attenuated.
Controlling for all family background and childhood adversity variables (the last three columns), the effect of physical abuse on all of the outcomes is attenuated but remains significant at the .05 level and significant at the .01 level for medical diagnoses and anger. Net of sex, age, all family background variables and all childhood adversity variables, childhood physical abuse is significantly associated with a 15% increase in medical diagnoses, a 16% increase in medical symptoms, a 19% increase in depression, a 22% increase in anger, and a 21% increase in anxiety. It should be noted that most of the family background variables in Table 2 are not significant.
OLS results using the larger sample with the trimmed set of family background and childhood adversity variables were very similar to results presented in Table 2. Specifically, these analyses demonstrate that respondents who were physically abused in childhood are significantly less healthy for all outcomes regardless of sex, age, family background and childhood adversities (results available upon request).
Table 3 shows the logistic regression models for each health outcome scale assessed at the 90th percentile. These results indicate that childhood physical abuse is associated with extreme ill health for medical diagnoses, depression, anger and anxiety, yet not physical symptoms after controlling for sex, age, and the trimmed set of family background and childhood adversity variables. Specifically, childhood physical abuse is associated with an 84% increase in the odds of having more medical diagnoses than 90% of the sample, a 61% increase in the odds of reporting more depression than 90% of the sample, a 102% increase in the odds of reporting more anger than 90% of the sample, and 78% increase in the odds of reporting more anxiety than 90% of the sample.
Table 4 shows the influence of childhood physical abuse on individual diagnoses and physical symptoms grouped by organ system controlling for sex, age, and the trimmed set of family background and childhood adversity measures. For respondents who were abused in childhood, the odds of having allergies, arthritis/rheumatism, asthma, bronchitis/emphysema, circulation problems, high blood pressure, heart troubles, liver troubles, and ulcers was 34 to 167 percent higher than for those who were not physically abused in childhood after controlling for sex, age, family background and childhood adversity variables. Self-reported physical abuse in childhood was also associated with cardiopulmonary, constitutional and musculoskeletal physical symptoms at the .05 level.
Self-reported physical abuse in childhood increased the likelihood of reporting more diagnosed illnesses, physical symptoms, anxiety, anger and depression nearly 40 years after the abuse took place. Childhood physical abuse was associated with decreases in mental and physical health, having worse health than 90% of the sample, and an array of specific diagnoses and symptoms by system after adjusting for sex, age, family background, and childhood adversities. The results showed that controlling for childhood adversities and family background was necessary to attain an accurate estimate of the impact of abuse on mid-life health.
Much of the past research examining the effects of childhood physical abuse on adult health collapses its effects with childhood sexual abuse (Dong et al., 2003; Felitti et al., 1998; McCauley et al., 1997; McNutt, Carlson, Persaud, & Postmus, 2002; Walker et al., 1999). These projects highlight that childhood physical abuse often does not occur alone and that experiencing multiple abuses increases adult health risks. However, it is also essential to understand how specific types of abuse impact adult health---taking into account other childhood adversities.
A small body of literature has examined the medical impact of physical abuse in childhood; however, most of this literature has not accounted for other childhood adversities. For example, Shaw and Krause (2002) found that physical abuse predicted more chronic medical conditions and more depression in adulthood, but they did not control for family background or other childhood adversities. Thompson et al. (2002), however is one example of research that did control for childhood adversities. Specifically, they examined data from the National Violence Against Women Survey and found that childhood physical victimization was associated with lower perceived general health, chronic mental health conditions, the occurrence of injury and alcohol and drug use after controlling childhood sexual abuse and several adult-aged mediators (Thompson, Arias, Basile, & Desai, 2002).
Evidence such as those in the current study linking abuse in childhood with the long-term development of specific medical diagnoses is essential because it can begin to show the mechanisms through which childhood maltreatment impacts adult health (Felitti et al., 1998; Goodwin et al., 2003; Shaw & Krause, 2002; Stein & Barrett-Connor, 2000; Walker et al., 1999). Specific medical diagnoses results are particularly compelling when multiple studies with diverse samples and different abuse questions produce similar findings. For example, both the current project and Shaw and Krause (2002) found that heart trouble but not diabetes or cancer was significantly greater in survivors of childhood physical abuse. Goodwin et al. (2003) examined the same cohort studied by Shaw and Krause (2002) and found childhood physical abuse was associated with recurring stomach problems and ulcers which is consistent with the current findings of ulcer diagnoses. However, unlike Shaw and Krause (2002), the current results indicate that hypertension was greater among those who reported childhood physical abuse.
The association of childhood physical abuse with adult psychiatric morbidity in this study is also consistent with the findings of others, although this is the first project we know of to provide an assessment of how anger is affected by childhood maltreatment in a population-based sample (MacMillan et al., 2001). The novel finding of increased anger in adults reporting childhood physical abuse could point to an additional mechanism through which childhood abuse might have an impact on adult physical health outcomes.
After controlling for an extensive array of family background and childhood adversity measures the effect of abuse on adult health outcomes generally was attenuated but not eliminated. This finding suggests that the strength of the relationship between abuse and health outcomes in previous studies that did not control for these confounders might not be as strong as reported. Future research aiming to obtain accurate estimates of the relationship between childhood physical abuse and adult health must control for additional childhood characteristics. Results suggest that controlling for one or few childhood adversity variables (such as parental marital problems) might be more efficient than including multiple variables.
Even though childhood abuse meets a number of epidemiological criteria for a causal relationship between an exposure and an outcome (Springer et al., 2003), acknowledgment of childhood abuse as a risk factor for conditions commonly managed by internists is surprisingly rare in the medical literature (Barsky & Borus, 1999; Carnes, Sarto, & Springer, 2001; Leventhal, 1999; Mehler, 2001; Sharpe & Carson, 2001; US Department of Health and Human Services, Public Health Services, & Agency for Health Care Policy and Research [AHCPR], 1993; Whooley & Simon, 2000). Such omissions underscore the urgency of reporting about childhood abuse in the medical literature as well as the importance of continuing research on the relationship between maltreatment and adult health conditions.
Though the current study has many strengths, it shares several weaknesses common to studies of childhood abuse and adult health. First, despite the longitudinal nature of the WLS, the abuse and health data are cross-sectional. We hypothesize that abuse increases the number of health problems, but we cannot rule out the possibility that the presence of ill health may have enhanced recall of abuse. Most studies suggest an under-reporting of abuse, however, so those who do report abuse in childhood are likely to actually have had this experience (Fergusson, Horwood, & Woodward, 2000).
Second, mood congruency bias may be a potential problem; this may occur, for example, if depressed individuals “selectively recall negative experiences and hence may exaggerate or misrepresent the presence of childhood adversity” (Brewin, Andrews, & Gotlib, 1993). Though some research suggests this is a problem (Lewinsohn & Rosenbaum, 1987) many others have found no evidence of this bias (Brewin et al., 1993; Fergusson et al., 2000; Maughan, Pickles, & Quinton, 1995; Robins et al., 1985). Sensitivity tests controlling for depression in the current models predicting medical diagnoses suggest that mood congruency bias is not a serious concern (results available upon request).
A third limitation of these analyses is that we did not control for other specific types of childhood abuse including sexual abuse and emotional abuse. However, the WLS contains a variety of childhood adversities that often co-occur with childhood sexual, emotional, and physical abuse. By controlling for these childhood adversities we may have statistically accounted for much of the common negative environment associated with childhood sexual and emotional abuse.
Fourth, we do not account for adult violence. Research has robustly demonstrated that violence in childhood is associated with violence in adulthood---especially for women (Kendall-Tackett, 2003). It is, therefore, possible that our findings of increased adverse health problems are due to domestic violence, rather than childhood violence. However, sensitivity tests do not suggest this is a strong possibility (results available up request).
Finally, there are limitations with the measure of childhood physical abuse used. The current measure draws from only two items, and, therefore, its reliability may be weak. Though only two items were used, they have performed well in previous studies (Shaw & Krause, 2002) and are from a well-validated and often-used scale (Straus et al., 1981). However, future research should include additional measures from the conflict tactics scale.
The results in this project point to several areas for future research. Other projects would do well to examine how the long-term health effect of abuse may differ for childhood sexual abuse and childhood emotional abuse. Scholars could also examine what factors in childhood are protective and serve to enhance survivors’ resiliency. Finally, future research should examine potential mechanisms linking childhood physical abuse with adult health. Exploring health behaviors as a potential pathway might be particularly fruitful given extant literature demonstrating that childhood abuse is associated with negative health behaviors. For example, clinical studies examining health behaviors as outcomes suggest a relationship between childhood abuse and smoking (Anda et al., 1999), drinking (Dong et al., 2003), high-risk sexual activities (Hillis et al., 2001), and obesity (Williamson, Thompson, Anda, Dietz, & Felitti, 2002). It would also be useful to examine mental health as one potential mechanism through which childhood abuse adversely impacts adult physical health.
In conclusion, this study demonstrates that self-reported physical abuse in childhood is associated with poor mental and physical health, including chronic medical conditions, even decades after the abuse. This association was attenuated but generally persisted when controlling for family background and childhood adversity variables. Given the personal and health care burden associated with the long-term effects of childhood abuse and the availability of effective interventions, recognition of this link appears to be underemphasized in medical teaching and practice.
The authors would like to acknowledge Robert M. Hauser, Director of the Wisconsin Longitudinal Study, for his encouragement, insightful suggestions, collaborative spirit, and technical expertise, Jeremy Freese for his statistical consultation and support, Editor Leventhal for his constructive criticism, and the two anonymous reviewers for their sage comments.
This work was supported by National Institute on Aging grants No. K07 AG0074 and R01 AG9775, the DHHS OWH National Centers of Excellence in Women’s Health Program, the National Science Foundation grant No. 0123666, the Vilas Estate Trust, and the Jean Manchester Biddick-Bascom endowed professorship. Dr. Carnes is partially supported by the GRECC at the Madison VA Hospital. Earlier versions of this paper were presented at the Life Course Inequality: Adverse Childhood Experiences and Adult Health Symposium at Purdue University, the Population Association of America Annual Meeting, and the Gerontological Society of America Annual Meeting.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.