Of the 14 eligible studies, four used the same primary data set for analysis (Mapp 2006; Pazdera et al. 2013; Renner et al. 2015; Schuetze and Eiden 2005). This means that there are 11 separate samples in this review with size of samples ranging from 44 to 483; a total of 1545 participants (see Table 1 for a summary of demographic characteristics). Of the 11 different samples, five employed a cross-sectional research design and two further studies included a case–control comparison group. The remaining four samples used a prospective design, measuring CSA at time point one and parenting variables at time point two. All studies recruited only mothers with six of the 11 samples from the USA, two from Canada, two from Australia and one from Scotland. Most studies recruited mothers from a non-clinical population (eight out of the 11 samples; n = 1391) mostly using a response to advert procedure and only three of the 11 samples were recruited from a clinical population (n = 154), including a mother and baby unit, a mental health outpatient clinic and a therapeutic community. Reporting on the ethnicity of participants varied: three studies did not report the ethnicity of participants; five of the 11 samples were mostly Caucasian participants and three samples reported a majority of African-American participants.
Eleven of the 14 studies (79%) used the Parenting Stress Index (PSI; Abidin 1995) or the Parenting Stress Index – Short Form (PSI-SF; Abidin 1995), see Table 2. One further study used several subscales of the PSI (Renner et al. 2015) and another study used a measure which included some items from the PSI (Barrett 2009). Only one study used an alternative measure, the Everyday Stress Index (Lutenbacher 2000). The frequent use of the PSI and the PSI-SF makes comparison between studies more viable.
In contrast, there was little homogeneity among studies regarding measurement of CSA (see Table 2). Two studies used the Childhood Trauma Questionnaire (CTQ; Bernstein and Fink 1998), but the remainder of the studies used either a different measurement tool such as the Child Abuse and Trauma Scale (CATS; Sanders and Becker-Lausen 1995) used in Harmer et al. (1999), or questions designed by the researchers.
Only six studies explicitly stated their definition of CSA (Alexander et al. 2000; Douglas 2000; Mapp 2006; Pazdera et al. 2013; Renner et al. 2015; Schuetze and Eiden 2005). Within these six studies, two limited their definition of CSA to contact abuse only (Alexander et al. 2000; Douglas 2000) and the remaining four, which used the same primary data set, included both contact and non-contact abuse. Additionally, the majority of studies used measures that simply measured the presence or absence of CSA. The exception to this is Wright et al. (2005) who initially asked mothers who had experienced CSA to respond to an advert for participants. Responses to the anonymised mailed questionnaire were then coded for severity by the researchers. In summary, the lack of consensual definitions and measurement of CSA makes comparison between studies difficult.
No studies were rated as strong in quality overall using the EPHPP tool (see Table 2). Eight were rated as moderate and six were rated as weak in quality, though several studies contained components that were rated as strong. Most of the studies were rated as moderate in the data collection section with three studies being rated as strong, mainly due to robust reporting of the reliability and validity of the measures used. Ten studies were rated as moderate on selection bias with the study sample considered to be at least somewhat likely to be representative of the target populations. However, four studies were rated weak mostly because participants self-referred into the study. A notable limitation in the majority of studies (n = 11) was the lack of description of possible confounding variables in either the methodological design or analysis of the studies. Most studies highlighted this issue later in the discussion section when suggesting possible explanations of their results, but very few address potential confounders earlier on.
Direct associations between CSA and parenting stress
Seven of the 14 studies indicated a degree of direct association between experiencing CSA and later parenting stress, with six presenting statistically significant results (correlations ranged between r = .13 to .33; Cohen’s d ranged between .22 to .65) and one indicating the mean scores of the CSA group were markedly higher than the norms provided by Abidin (1995). Two of the 14 studies did not find any association between CSA and parenting stress and the remaining five studies suggest other variables may affect the relationship between CSA and parenting stress, such as locus of control and current stressors.
Two of the seven studies which found an association between CSA and parenting stress found a significant positive association between mothers who reported CSA and higher scores on the PSI-SF (Douglas 2000; Pereira et al. 2012). These two studies were from different samples. The remaining five studies reported significant associations between CSA and one subscale of the PSI (Buist and Janson 2001; Ethier et al. 1995; Renner et al. 2015; Schuetze and Eiden 2005; Wright et al. 2005), including the parenting domain (n = 4) and the optional life stress scale (n = 1).
Both the Douglas (2000) and Pereira et al.’s (2012) study were rated as moderate in quality. Douglas was only rated as weak on controlling for confounds as the study reported that the index group in this study were significantly more likely to be younger, live in a more deprived area and have experienced parental separation, divorce or death than the control group, yet these variables were not discussed in the method or controlled for in the analyses. The significant results in this study may therefore be accounted for by confounding variables such as these, with elevated stress reported by the index group possibly being associated with variables other than CSA per se. Alternatively, the significant results found in this study may be due to the very clear limits on the definition of CSA which was contact abuse only before the age of 16, whereas several other studies that report non-significant effects included non-contact sexual abuse (e.g. Mapp 2006). Arguably, lasting effects of CSA may be more likely following contact rather than non-contact sexual abuse, possibly accounting for the significant results in this study.
The significant results found in the Pereira et al. (2012) study may in part be due to the large sample size (N = 291) which may have been sufficient to detect subtle associations between CSA and parenting stress in the community sample and protect against type II errors. The study was rated as moderate in quality, only scoring one weak rating due to the cross-sectional study design. However, this study was rated as strong on data collection as it used measurement tools that have been shown to be both valid and reliable, the CTQ and the PSI-SF. The CTQ does include non-contact CSA, but the use of a standardised measure of childhood trauma which reports robust reliability (α = .91 for the whole scale, .94 for the CSA subscale in a community sample; Scher et al. 2001) may have enabled consistent reporting of experiences across participants.
Of the five studies that report associations between CSA and a subscale of the PSI, three were rated as moderate in quality and two were rated as weak. Buist and Janson’s (2001) study is of moderate quality overall, with a weak rating for the lack of description regarding control of confounding variables. They reported that the CSA group in their sample scored significantly higher on the optional life stress scale on the PSI than the comparison group (d = .65). As this is the only study to report the optional life stress subscale of the PSI it is difficult to make any assumptions about the significance of this finding. No significant difference was reported between the CSA group and the comparison group on either the parent or child domain of the PSI which may be due to a lack of power as the sample size was relatively small (N = 45; CSA group n = 23, comparison group n = 22) which increases the possibility of type II error.
Renner et al. (2015) found that women reporting CSA had slightly higher mean scores on all five subscales of the PSI parenting domain they included in their study when compared to women not reporting CSA. Effect sizes were calculated for these subscales and three were found to show a small effect (see Table 2). Additionally, Schuetze and Eiden (2005) reported that CSA was significantly associated with parenting stress on the parent domain of the PSI, but not significantly associated with the child domain. Both these studies used the same primary data set and are of moderate quality, which suggests the results reported may reliably indicate that there is a degree of association between CSA and later parenting stress on the parent domain of the PSI for the participants in this study, which were drawn from a community sample.
Both Ethier et al. (1995) and Wright et al. (2005) were rated as weak on the quality assessment tool, though both reported associations between CSA and scores on the parenting domain of the PSI. Ethier et al. explored issues pertaining to motherhood for negligent mothers, with parental negligence defined as “a serious omission from the parent who endangers the child’s development” (p. 622). All mothers in this group had been implicated in severe maltreatment and were found to experience significantly higher levels of stress than the control group. Both the index and comparison groups contained mothers with histories of CSA and Ethier et al. found that total sexual abuse was significantly associated with stress on the parent domain of the PSI for both the index and comparison group. However, only the mothers in the control group were found to have significant associations with CSA on the total stress score. One possible explanation for this is that the index group may have more current daily stresses than the control group, given their alleged maltreatment of their children. The effects of CSA therefore appear more salient for the control group who may not have such difficult situations to contend with.
Finally, with regard to direct associations between CSA and later parenting stress, Wright et al. (2005) found that the mean scores for mother’s reporting CSA on the parent subscales of the PSI were markedly higher on six out of seven subscales compared to the normative sample from Abidin (1995). Again, this provides further support for an association between the parent domain of the PSI in particular and historical CSA. However, this study was predominantly weak in quality, particularly with regard to selection bias and research design, as participants had responded to an advert asking for mothers who had experienced CSA. This self-selection bias may have skewed the results making the sample in the study not representative of the population of people who have experienced CSA.
Two studies reported no association between CSA and later parenting stress. Alexander et al. (2000) did not find a significant main effect of CSA on parenting stress. However this study was rated as weak in quality with a cross-sectional design, possible selection bias with recruitment relying on response to advert and lack of control for confounders. The second study, Barrett (2009), was rated as moderate in quality and had the largest sample in this review (N = 483). Barrett reported the mean of the CSA group was not significantly different from the control group on the measure of parenting stress used and CSA did not reach significance in the regression analysis (see Table 2). It is possible that the use of non-formal measurement tools affected the results obtained and this component was rated as weak on the EPHPP. For example, the CSA measure was: “has a stranger, acquaintance, date or relative ever tried or succeeded in doing something sexual to you against your wishes?” (p. 496) with affirmative responses followed up with a question regarding age of occurrence. This may also mean that the abuse group included participants for whom the abuse may not have been as severe as other studies which used a more stringent measure of CSA such as Douglas (2000) who defined CSA as women with a history of contact child sexual abuse before the age of 16. Idiosyncratic measurement of CSA is not unusual throughout the studies in this review, but for parenting stress other studies used a validated measure whereas Barrett did not, opting instead for a scale from a women’s employment study which was conducted in the USA, that “included items from the PSI” (p. 497). It is possible this measure was not a valid or reliable measure of parenting stress which may have skewed the results in the study. Furthermore, despite the Barrett study having a large sample, the percentage of CSA survivors in this sample was actually the smallest out of all the studies included in this review (11%, see Table 1). This increases the possibility of a type II error as it may seem as though there was no effect of CSA on parenting stress when the sample size of CSA survivors was not sufficient to detect any effect.
Only two studies limited their inclusion criteria to contact CSA only: Alexander et al. (2000) who did not find any association between CSA and later parenting stress and Douglas (2000) who found that mothers in their CSA group reported significantly more stress overall than their comparison group. This difference in results may be due to the methodological quality of the studies: Alexander et al. was rated as weak in quality and Douglas was of moderate quality. An alternative explanation may be that the Douglas study used a clinical sample from a mental health outpatient clinic where participants may be experiencing elevated stress due to their mental health difficulties rather than due to parenting per se, whereas Alexander et al. recruited from the community where there may be less variation in the data. Lastly, the Douglas study contained a greater proportion of CSA survivors (54%) compared to the Alexander et al. study (21%) which may have enhanced the potential of identifying an association between CSA and parenting stress.
In summary, there is no strong, consistent evidence of a direct association between CSA and later parenting stress. However, the results suggest that contact-only CSA may produce a significant association with parenting stress and that studies including both contact and non-contact CSA may need larger sample sizes to detect smaller effects. Several studies suggest elevated stress on the parenting domain of the PSI but not the child domain which suggests participants were more likely to attribute parenting stress to their own characteristics rather than the characteristics of the child.
Possible mediating factors between CSA and parenting stress
Depression was highlighted in 10 of the studies as having a significant association with both CSA and parenting stress. The results of eight of these studies suggest there may be a potential indirect path from CSA to parenting stress through current level of depression (Buist and Janson 2001; Douglas 2000; Ethier et al. 1995; Lutenbacher 2000; Mapp 2006; Pazdera et al. 2013; Schuetze and Eiden 2005; Wright et al. 2005). Five of these studies were of moderate quality and three were weak in quality. The other two studies, both rated as weak in quality, found a significant association between depression and parental stress, though the association between CSA and depression was not significant (Harmer et al. 1999; Lang et al. 2010). Of the eight studies which found significant associations between CSA, level of depression and parenting stress, three of these used the same primary data set (Mapp 2006; Pazdera et al. 2013; Schuetze and Eiden 2005) and hence the same measure of depression; the Center for Epidemiologic Studies Depression Scale (CES-D; Radloff 1977). This scale was also used in the Lutenbacher (2000) and Wright et al. (2005) study while the Beck Depression Inventory (BDI; Beck et al. 1961) was used in both the Buist and Janson (2001) and the Ethier et al. (1995) study. Buist and Janson also used the Hamilton Rating Scale for Depression (HDRS) and Douglas (2000) found a significant association between the depression subscale on the General Health Questionnaire (GHQ-28) and parenting stress for both the CSA group and the comparison group. The results of these eight studies, which used different but reliable methods of measuring depression, suggest depression is a significant factor in the association between CSA and parenting stress.
With regard to the two studies which found a significant association between depression and parental stress, yet not between CSA and depression, Lang et al. (2010) found depression was significantly negatively related to defensive responding and parental distress on the PSI-SF at one year postpartum. This means that participants reported less parental distress than they may actually be experiencing. However, conclusions from Lang et al. should perhaps be interpreted with some caution because the study was of weak quality overall and retained only 31 out of 44 participants for the postpartum follow-up. Such attrition may result in a biased sample at follow-up and this small sample size is not particularly representative, making analysis susceptible to type II errors. Similarly, Harmer et al. (1999) was rated as methodologically weak and reports that some mothers chose not to complete all measures. The number of participants per measure ranged from 39 to 46 and five participants chose to complete the measures with the assistance of a researcher, which increases possibility of demand characteristics. Furthermore, approximately half of the remaining participants had missed occasional questions when they returned the measures, which the researcher subsequently supported them to complete, again elevating the risk of bias.
Five studies conducted mediation analysis with their data (Mapp 2006; Pazdera et al. 2013; Pereira et al. 2012; Schuetze and Eiden 2005; Wright et al. 2005) though only three report CSA and parenting stress as predictor and outcome variables and depression as a mediator, which are the three studies which use the same primary data set (Mapp 2006; Pazdera et al. 2013; Schuetze and Eiden 2005). The other two studies report mediation using different outcome variables including maternal sensitivity (Pereira et al. 2012) and resilience domains (Wright et al. 2005). Mapp (2006) reported the results of a path analysis which indicated the only significant route from CSA to elevated scores on the PSI was through the level of current depression. This study also noted locus of control impacted scores on the PSI both directly (r = .47) and through depression (r = .45). Both Pazdera et al. (2013) and Schuetze and Eiden (2005) included other variables in their mediation models which precludes clear conclusions being made regarding whether depression mediates the association between CSA and parenting stress. Pazdera et al. (2013) conducted a multiple mediation model which included CSA as predictor, parenting sense of competence and depression as mediators, and parenting stress and maltreatment behaviour as outcome variables. They reported the fit of the model to the data was relatively poor (χ
2(7) = 36.17, p = <.001). Similarly, Schuetze and Eiden (2005) found that partner violence, along with depression, mediated the association between CSA and the outcome variables which were parenting attitudes (including both parenting stress and parenting competence) and punitive discipline. However, the model did not fit the data particularly well (χ
2(21) = 38.17, p = <.05). These results suggest variables other than depression may impact the association between CSA and parenting stress, though investigation of these relationships was only conducted in studies which used the same primary data, demonstrating a need to replicate these findings in different samples.
As indicated above, the studies included in this review measured a number of other variables alongside CSA, depression and parenting stress. There was little homogeneity between studies in terms of variables measured, but several studies indicated significant associations with other factors. Positive belief systems were found to be negatively associated with parenting stress in six studies (Buist and Janson 2001; Lutenbacher 2000; Mapp 2006; Pazdera et al. 2013; Renner et al. 2015; Schuetze and Eiden 2005). For example, higher self-esteem was negatively associated with stress in the Lutenbacher (2000) study (r = −.48, p = <.001) and higher scores on parenting satisfaction and self-efficacy were associated with lower scores on parenting stress in Pazdera et al. (2013) and Schuetze and Eiden (2005) (associations ranged between −.41 to −.68, p = <.01). Similarly, higher social support and/or relationship satisfaction were associated with lower parenting stress for CSA survivors in three studies (Alexander et al. 2000; Harmer et al. 1999 and Wright et al. 2005). Such factors may therefore be potential mediators or moderators of the relationship between CSA and parenting stress, though were not tested as such in the studies.
Seven studies included measures of various other forms of childhood maltreatment, including neglect and physical and emotional abuse (Alexander et al. 2000; Barrett 2009; Ethier et al. 1995; Harmer et al. 1999; Lang et al. 2010; Lutenbacher 2000; Pereira et al. 2012). Different types of childhood maltreatment were associated with each other in most of these studies and parenting stress was associated with the experience of childhood physical abuse in four studies (Barrett 2009; Ethier et al. 1995; Lang et al. 2010; Pereira et al. 2012), with neglect/negative home environment in two studies (Ethier et al. 1995; Harmer et al. 1999) and emotional abuse in two studies (Lang et al. 2010 and Pereira et al. 2012). Furthermore, current partner violence was also associated with stress in two studies which included a measure of this (Lutenbacher 2000; Schuetze and Eiden 2005), though was only associated with CSA in Schuetze and Eiden (2005).
Finally, only six of the 14 studies reported characteristics of the CSA experienced by their participants (Alexander et al. 2000; Buist and Janson 2001; Douglas 2000; Lutenbacher 2000; Schuetze and Eiden 2005; Wright et al. 2005). Despite the range of experiences within the categorisation of CSA, only Douglas (2000) reported analyses using these different types of experience, finding no significant difference between scores on the PSI for intra and extra-familial abuse. No studies included analysis of other potential moderators, such as age or severity of abuse, so conclusions regarding different aspects of CSA and the effects on later parenting stress could not therefore be inferred.