Child maltreatment in Kenya, Zambia, and the Netherlands : a cross-cultural comparison of prevalence, psychopathological sequelae, and mediation by PTSS

(1)

: a cross-cultural comparison of prevalence,

psychopathological sequelae, and mediation by PTSS

Mbagaya, C.V.

Citation

Mbagaya, C. V. (2010, December 1). Child maltreatment in Kenya, Zambia, and the Netherlands : a cross-cultural comparison of prevalence,

psychopathological sequelae, and mediation by PTSS. Retrieved from https://hdl.handle.net/1887/16193

Version: Not Applicable (or Unknown)

License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden

Downloaded from: https://hdl.handle.net/1887/16193

Note: To cite this publication please use the final published version (if applicable).

(2)

Child Maltreatment in Kenya:

Prevalence, associated Sequelae, and Mediation by

Posttraumatic Stress Symptoms.

(3)

(4)

5.0. Introduction

In this chapter we present the results of our Kenyan sample. Descriptive statistics and bivariate associations for the variables that we investigated are provided. Next, we present findings on the multivariate associations between child maltreatment and psychopathological symptoms. Lastly, we present the results of our tests of mediation models with PTSD mediating the association between childhood maltreatment experiences and current psychopathological symptomatology.

5.1. Descriptives

Table 5.1 presents the means and standard deviations of the childhood maltreatment variables, posttraumatic stress symptoms (PTSS), social desirability and psychopathology variables. Male and female students differed significantly on their experiences of childhood maltreatment. Male students reported more experiences of childhood physical abuse (M = 2.22, SD = 0.90) than did female students (M = 1.96, SD = 0.74). These differences were also significant for sexual abuse; males reported more sexual abuse (M = 1.53, SD = 0.58) than females (M = 1.38, SD = 0.57). Similarly, males reported having experienced more neglect (M = 1.67, SD = 0.43) in their childhood than did females (M = 1.54, SD = 0.42). It is only in witnessing interparental violence that their experiences were comparable. The prevalence of CSA in the Kenyan sample was 31% (Males: 37%; Females: 24%) while that of CPA was 42% (Males: 48%; Females: 36%). Those who reported childhood neglect were 59% of the sample (Males: 62%; Females: 56%) while 20%

(Males: 22%; Females: 18%) reported witnessing interparental violence.

Table 5.1

Descriptive Data for Males and Females on Childhood Maltreatment Experiences and Personality Variables

Total (N=375)

Male (n=193)

Female

(n=182) t (df)

M SD M SD M SD

Physical abuse 2.10 0.83 2.22 0.74 1.96 0.90 3.02 (366.61)**

Sexual abuse 1.46 0.58 1.53 0.57 1.38 0.58 2.57 (373)**

Neglect 1.61 0.43 1.67 0.42 1.54 0.43 2.79 (373)**

Witnessing violence 1.59 0.94 1.67 0.87 1.50 0.99 1.77 (371.91) Social desirability 2.47 0.52 2.49 0.52 2.45 0.52 0.76 (373)

PTSS 2.51 0.49 2.52 0.53 2.51 0.45 0.23 (373)

(5)

Table 5.2. Bivariate Correlations of Child Maltreatment Experiences and Personality Variables 123456789101112 1 Physical abuse 2 Sexual abuse.25** 3 Neglect history.16**.32** 4 Witnessing violence.25**.26**.24** 5 Social desirability-.21**-.23**-.01-.12* 6 PTSS.28**.32**.11*.19**-.41** 7 Antisocial personality.19**.29**.35**.24**-.47**.41** 8 Criminal tendencies.35**.44**.27**.34**-.40**.40**.51** 9 Dating violence.15**.23**.23**.10*-.12**.20**.28**.32** 10 Externalizing problems.30**.42**.37**.30**-.43**.44**.79**.80**.70** 11 Depressive symptoms.23**.27**.39**.18**-.25**.40**.45**.35**.28**.47** 12 Borderline symptoms.25**.25**.21**.19**-.49**.53**.56**.40**.29**.54**.56** 13 Internalizing problems.27**.29**.34**.21**-.43**.53**.57**.43**.32**.58**.88**.88** * p < .05. ** p < .01

(6)

Psychopathology variables to some extent bore significant differences between male and female students. Males reported more criminal tendencies (M

= 1.89, SD = 0.57) than females (M = 1.64, SD = 0.56). In addition, males reported significantly more violent behaviour to their dating partners (M = 6.17, SD = 7.19) than did female students (M = 3.82, SD = 5.26). Correspondingly, the level of externalizing behaviour was higher for males (M = 0.15, SD = 0.79) than for females (M = -0.16, SD = 0.70). In contrast, both genders did not differ significantly on other psychopathological assessments like antisocial personality symptoms, borderline personality symptoms, depressive symptoms and overall internalizing behaviour. They were also comparable in posttraumatic stress symptoms (PTSS) and in social desirability.

All the bivariate associations of childhood maltreatment variables, psychopathological characteristics, social desirability and PTSS were significant (see Table 5.2). Notably, social desirability was negatively related to all the variables measured. This means that the higher the respondents scored on social desirability, the lower they scored on child maltreatment, PTSS and psychopathological symptomatology. This association was however not significant for history of neglect. All the childhood maltreatment variables had significant associations with PTSS and the psychopathology assessments. The higher respondents scored on PTSS, the more psychopathological symptoms they reported.

(7)

5.2 Multivariate relations between child maltreatment variables and psychopathological symptoms

Hierarchical regression analyses were used to test whether childhood maltreatment experiences predicted externalizing and internalizing behaviour problems. In all analyses, social desirability was entered at step 1. At step 2, background variables (income, gender and age) were entered. At step 3, PTSS was entered and finally at step 4, all the childhood maltreatment experiences (physical abuse, sexual abuse, neglect and witnessing interparental violence) were added.

In the prediction of antisocial personality symptoms, the weight of social desirability was negative and significant, β = -.39 (p <.01). The higher the respondents scored on social desirability the lower they scored on antisocial personality symptoms. Of the background variables, age was significantly related to antisocial personality symptoms, with younger respondents reporting more antisocial personality symptoms than older respondents, β = -.20 (p < .01). Higher PTSS was also significantly associated with more antisocial personality symptoms, β = .19 (p < .01). After controlling for social desirability, background variables and PTSS, neglect history, β = .32 (p <.01) and witnessing violence, β = .09 (p = .04) were significant predictors (Table 5.3). The higher the respondents scored on either neglect or witnessing interparental violence, the higher they scored on antisocial personality symptoms. The predictors and covariates together accounted for 43%

of the variance in antisocial personality symptoms scores.

Table 5.3.

Regression Analysis Predicting Antisocial Personality Symptoms from Childhood Maltreatment Experiences with Social Desirability, Income, Gender, Age, and PTSS as Covariates

R R² R²Ch F_change df Beta¹ p

Step 1 .47 .22 .22 106.37 (1, 373) <.01

Social desirability -.39 <.01

Step 2 .52 .27 .05 8.20 (3, 370) <.01

Income -.01 .82

Gender -.06 .16

Age -.20 <.01

Step 3 .56 .32 .05 24.61 (1, 369) <.01

PTSS .19 <.01

Step 4 .66 .43 .11 17.98 (4, 365) <.01

Physical abuse -.04 .39

Sexual abuse .01 .79

Neglect .32 <.01

Witnessing violence .09 .04

1 Betas are derived from the final block of the regression model

(8)

Childhood maltreatment experiences, social desirability, PTSS, income, gender, and age together accounted for 42% of the variance in criminal tendencies (Table 5.4). Social desirability was significantly associated with higher scores on criminal tendencies, β = -.39 (p < .01).This implies that respondents who had higher scores on social desirability reported less criminal tendencies. The associations between criminal tendencies and gender and criminal tendencies and age were modest but significant. Younger respondents reported more criminal acts than did older respondents, β = -.17 (p < .01). Also, males reported more criminal acts than did females, β = -.12 (p < .01). PTSS was significantly related to criminal tendencies, β = .14 (p < .01). Higher scores on PTSS were associated with higher scores on criminal tendencies. After controlling for the effect of social desirability, background variables and PTSS, all the childhood maltreatment variables were significant predictors of criminal tendencies. Together, they accounted for 13% of the variance in criminal tendencies. Those who reported having experienced any one of the maltreatments in childhood; physical abuse, sexual abuse, neglect, and witnessing interparental violence were likely to report more criminal tendencies than those without such experiences.

Table 5.4.

Regression Analysis Predicting Criminal Tendencies from Childhood Maltreatment Experiences with Social Desirability, Income Gender, Age and PTSS as Covariates

Step 1 .40 .16 .16 69.51 (1, 373) <.01

Step 2 .49 .24 .09 13.84 (3, 370) <.01

Income -.03 .46

Gender -.17 <.01

Age -.12 <.01

Step 3 .54 .30 .05 28.17 (1, 369) <.01

PTSS .14 <.01

Step 4 .65 .42 .13 20.17 (4, 365) <.01

Physical abuse .12 <.01

Sexual abuse .20 <.01

Neglect .12 <.01

Witnessing violence .16 <.01

(9)

Hierarchical regression analysis was also conducted for the prediction of dating violence (Table 5.5). The prediction of dating violence by social desirability, income and age was not significant. However, gender was significantly associated with dating violence, with males more likely to report being aggressive towards their dating partners than females, β = -.13 (p = .01). Higher scores on PTSS were also associated with dating violence, β = .13 (p = .03). After controlling for the effects of social desirability, age, income, gender and PTSS, a history of neglect, β

= .12 (p < .01) and sexual abuse in childhood, β = .11 (p < .05) remained significant predictors of dating violence. This means that those who reported having been either neglected or sexually abused in childhood were more likely to be violent to their dating partners. Conversely, physical abuse, witnessing interparental violence, PTSS, social desirability, age and income were all nonsignificant predictors. All the predictors and covariates in the regression however, accounted for 13% of the variance in dating violence.

Table 5.5.

Regression Analysis Predicting Dating Violence from Childhood Maltreatment Experiences with Social Desirability, Income Gender, Age and PTSS as Covariates

Step 1 .12 .01 .01 5.25 (1, 373) .02

Social desirability -.05 .41

Step 2 .25 .06 .05 6.16 (3, 370) <.01

Income -.05 .28

Gender -.13 .01

Age .09 .09

Step 3 .30 .09 .03 10.57 (1, 369) <.01

PTSS .13 .03

Step 4 .36 .13 .04 4.31 (4, 365) <.01

Physical abuse .04 .51

Neglect .15 <.01

Witnessing violence -.02 .71

(10)

Table 5.6 presents the results of the hierarchical regression predicting depressive symptoms from childhood maltreatment experiences and covariates.

Together they accounted for 31% of the variance in depressive symptoms. Social desirability was negatively related to depressive symptoms, β = -.11 (p =.03). The higher the respondents scored on social desirability, the lower they scored on depressive symptoms. Of the three background variables, only gender was a significant predictor of depressive symptoms, β = .09 (p = .04). Female respondents were more likely to report depressive symptoms than were male respondents.

Higher PTSS scores were significantly associated with more depressive symptoms, β = .35 (p < .01). After controlling for social desirability, background variables and PTSS, only a history of neglect, β =.35 (p < .01) was a significant predictor of depressive symptoms. Thus, higher scores on a history of neglect were associated with more depressive symptoms. None of the other maltreatment variables was significantly related to depressive symptoms when PTSS, social desirability and background variables were taken into account.

Table 5.6.

Regression Analysis Predicting Depressive Symptoms from Childhood Maltreatment Experiences with Social Desirability, Income Gender, Age and PTSS as Covariates

Step 1 .25 .06 .06 25.00 (1, 373) <.01

Social desirability -.11 =.03

Step 2 .29 .08 .02 2.50 (3, 370) .06

Income -.06 .22

Gender -.09 .04

Age -.05 .32

Step 3 .43 .18 .10 44.50 (1, 369) <.01

PTSS .28 <.01

Step 4 .56 .31 .13 17.38 (4, 365) .02

Neglect .35 <.01

(11)

The results of the hierarchical regression analysis conducted to predict borderline personality symptoms are presented in Table 5.7. The predictors and covariates together accounted for 41% of the variance. Social desirability was significantly negatively related to borderline personality symptoms. The higher the respondents scored on social desirability the lower they scored on borderline personality symptoms, β = -.33 (p < .01). Of the background variables, age was significantly associated with borderline personality, β = -.13 (p < .01).

Younger respondents reported more borderline personality symptoms than did older respondents. After controlling for social desirability and the background variables, PTSS was significantly related to borderline personality symptoms, β = .35 (p < .01). Higher scores on PTSS were associated with higher scores on borderline personality symptoms. Of the childhood maltreatment variables, only a history of neglect was a significant predictor of borderline personality symptoms, β = .17 (p < .01). Higher scores on childhood neglect were associated with higher scores on borderline personality symptoms.

Table 5.7.

Regression Analysis Predicting Borderline Personality Symptoms from Childhood Maltreatment Experiences with Social Desirability, Income, Gender, Age and PTSS as Covariates

Step 1 .49 .24 .24 114.48 (1, 373) <.01

Step 2 .51 .26 .03 4.94 (3, 370) <.01

Income -.01 .79

Gender .00 .92

Age -.13 <.01

Step 3 .62 .38 .12 69.76 (1, 369) <.01

PTSS .35 <.01

Step 4 .64 .41 .03 4.97 (4, 365) <.01

Sexual abuse -.02 .63

Neglect .17 <.01

5.3. Mediation of the association between child maltreatment and psy- chopathological symptomatology by PTSS

All hierarchical regression analyses showed that PTSS significantly contributed to the variance in psychopathology outcomes. In order to examine whether the association between the various childhood maltreatment variables and psychopathological outcomes could in part be explained through PTSS, mediation

(12)

analyses were done for all the maltreatment variables and psychopathology outcomes examined in this study. In accordance with the mediation analysis procedure summarized by Baron and Kenny (1986), four steps were made in order to test for mediation by PTSS. First, the childhood maltreatment variable should significantly predict PTSS (path a); second, PTSS should also significantly predict the psychopathology outcome (path b); third, there should be a significant association between the child maltreatment variable and psychopathology outcome (total effect, path c); and lastly, the association between the maltreatment variable and psychopathology outcome ought to decrease when PTSS is controlled for (direct effect, path c1). The mediational model that we used in our study is illustrated below (Figure1).

In determining whether PTSS was indeed a mediator of the association between childhood maltreatment variables and antisocial personality symptoms, a series of regressions and Sobel tests were completed (Table 5.8). The model involving childhood physical abuse (CPA) showed that the association between CPA and PTSS was significant, β = .21 (p < .01). Similarly, the association between PTSS and antisocial personality symptoms was significant, β = .39 (p < .01).

The regression weight of the total effect of CPA and antisocial personality was also significant, β = .19 (p < .01). After PTSS was included in the regression, the regression weight of the direct effect dropped to a nonsignificant level, β = .08 (p

= .09).This implied that the mediation of PTSS was complete. The Sobel test (z) showed that PTSS significantly mediated the association of CPA and antisocial personality symptoms, z = 2.72 (p < .01).

In the same way, regressions and the Sobel test were completed in order to find out whether the association between childhood sexual abuse (CSA) and antisocial personality symptoms was mediated by PTSS. The mediation model showed a significant association between CSA and PTSS, β = .21 (p < .01), and between PTSS and antisocial personality symptoms, β = .36 (p < .01). The regression weight of the total effect of CSA on antisocial personality symptoms was β =.29 (p < .01).

After controlling for PTSS, the regression weight of the direct effect dropped to β = .18 (p < .01). The Sobel test showed that the partial mediation by PTSS was

Figure 5.1. Mediational Model with PTSS Mediating the Association between Child Maltreatment and Psychopathological Symptomatology

In determining whether PTSS was indeed a mediator of the association between childhood maltreatment variables and antisocial personality symptoms, a series of regressions and Sobel tests were completed (Table 5.8). The model involving childhood physical abuse (CPA) showed that the association between CPA and PTSS was significant, β = .21 (p < .01).

Similarly, the association between PTSS and antisocial personality symptoms was significant, β = .39 (p < .01). The regression weight of the total effect of CPA and antisocial personality was also significant, β = .19 (p < .01). After PTSS was included in the regression, the regression weight of the direct effect dropped to a nonsignificant level, β = .08 (p = .09).This implied that the mediation of PTSS was complete. The Sobel test (z) showed that PTSS significantly mediated the association of CPA and antisocial personality symptoms, z = 2.72 (p

< .01).

In the same way, regressions and the Sobel test were completed in order to find out whether the association between childhood sexual abuse (CSA) and antisocial personality symptoms was mediated by PTSS. The mediation model showed a significant association between CSA and PTSS, β = .21 (p < .01), and between PTSS and antisocial personality symptoms, β = .36 (p < .01). The regression weight of the total effect of CSA on antisocial personality symptoms was β =.29 (p < .01). After controlling for PTSS, the regression weight of the direct effect dropped to β = .18 (p < .01). The Sobel test showed that the partial mediation by PTSS was significant, z = 3.77 (p < .01).

Child maltreatment Psychopathology

PTSS

a b

c c¹

Figure 5.1. Mediational Model with PTSS Mediating the Association between Child Maltreatment and Psychopathological Symptomatology

(13)

between witnessing inter-parental violence in childhood and PTSS β = .10 (p < .05), and between PTSS and antisocial personality symptoms β = .38 (p < .01). The total effect of witnessing inter-parental violence on antisocial personality symptoms was β = .24 (p < .01). After controlling for PTSS, the direct effect dropped to β

= .17 (p < .01). The Sobel test indicated that the partial mediation by PTSS was significant, z = 2.01 (p < .05).

In contrast, PTSS did not mediate the association between a history of neglect and antisocial personality symptoms. This was because of the nonsignificant association between a history of neglect and PTSS, β = -.04 (p < .39).

Table 5.8.

PTSS Mediating the Association between Child Maltreatment Variables and Antisocial Personality Symptoms

Child Maltreatment variable Regressions B SE β

Physical abuse

Mediation test 3.84 (p < .01)

Physical abuse – PTSS¹ 0.12 .03 .21**

PTSS - Antisocial personality¹ 0.34 .04 .39**

Physical abuse – Antisocial

personality¹ 0.10 .03 .19**

Physical abuse - Antisocial

personality controlling for PTSS 0.04 .03 .08

Sexual abuse

Sexual abuse - PTSS¹ 0.18 .04 .21**

PTSS - antisocial personality¹ 0.31 .04 .36**

Sexual abuse – Antisocial

Sexual abuse - Antisocial

personality controlling for PTSS 0.13 .04 .18**

Neglect

Neglect - PTSS¹ -0.05 .06 -.04

Neglect - Antisocial personality¹ 0.36 .05 .35**

Neglect - Antisocial personality

controlling for PTSS 0.31 .05 .31**

Witnessing violence Mediation test 2.01 (p =.04)

PTSS - Witnessing violence¹ 0.05 .03 .10*

Witnessing violence - Antisocial

personality controlling for PTSS 0.08 .02 .17**

* p < .05. ** p < .01.

1 controlling for the third variable

Table 5.9 shows the regression coefficients and the Sobel test results of the mediation models involving childhood maltreatment variables and criminal tendencies. In the model testing CPA, PTSS and criminal tendencies, the association between CPA and PTSS was significant, β = .16 (p < .01). The association between PTSS and criminal tendencies was also significant, β = .33

(14)

(p < .01). Further, the regression weight of the total effect of CPA and criminal tendencies was significant, β = .35 (p < .01). After controlling for PTSS, the weight of the direct effect dropped but remained significant, β = .26 (p < .01). Sobel test showed that the partial mediation of the association between CPA and criminal tendencies was significant, z = 2.85 (p < .01).

The model testing CSA, PTSS and criminal tendencies showed significant associations between CSA and PTSS β = .17 (p < .01) and between PTSS and criminal tendencies β = .29 (p < .01). The regression weight of the total effect of CSA on criminal tendencies was also significant, β = .44 (p < .01). After controlling for PTSS, the regression weight of the direct effect dropped but remained significant, β = .35 (p < .01). The Sobel test indicated that the partial mediation by PTSS was significant, z = 2.94 (p < .01).

In contrast, PTSS did not mediate the association between a history of neglect and criminal tendencies because a history of neglect was not related to PTSS. In addition, PTSS did not mediate the association between witnessing violence and criminal tendencies because of the nonsignificant relation between witnessing violence and PTSS, β = .07 (p = .19).

Table 5.9.

PTSS Mediating the Association between Child Maltreatment Variables and Criminal Tendencies

Child maltreatment Regressions B SE β

Physical abuse

Physical abuse - PTSS¹ 0.09 .03 .16**

PTSS - Criminal tendencies¹ 0.38 .06 .33**

Physical abuse - Criminal

tendencies¹ 0.25 .03 .35**

Physical abuse - Criminal

tendencies controlling for PTSS 0.18 .03 .26**

Sexual abuse

Sexual abuse - Criminal tendencies¹ 0.44 .05 .44**

Sexual abuse - Criminal tendencies

Neglect history

Neglect - PTSS¹ .00 .06 .00

Neglect - Criminal tendencies¹ 0.37 .07 .27**

Neglect - Criminal tendencies

(15)

Table 5.10 shows the regression coefficients and the Sobel test results of the mediation models involving childhood maltreatment variables and dating violence. The model testing CPA, PTSS and dating violence indicated a significant association between CPA and PTSS, β = .26 (p < .01) and between PTSS and dating violence, β = .17 (p < .01). The regression weight for the total effect of CPA and dating violence was also significant, β = .15 (p < .01). After controlling for PTSS, the regression weight of the direct effect dropped to a nonsignificant level, β = .10 (p = .06), implying complete mediation of the association between CPA and dating violence by PTSS. The Sobel test showed that this mediation was significant, z = 2.72 (p < .01).

The model testing CSA, PTSS and dating violence showed a significant association between CSA and PTSS, β = .29 (p < .01) and between PTSS and dating violence, β = .14 (p < .01). The regression weight for the total effect of CSA on dating violence was β = .23 (p < .01). After controlling for PTSS, the regression weight of the direct effect dropped but remained significant, β = .18 (p < .01). The Sobel test indicated that the partial mediation by PTSS was significant, z = 2.37 (p < .01). In the same way, the model involving witnessing interparental violence, PTSS and dating violence showed that the association between witnessing interparental violence in childhood and PTSS was significant, β = .18 (p < .01). Similarly, the association between PTSS and dating violence was significant, β = .18 (p < .01).

The regression weight of the total effect of witnessing interparental violence on dating violence was also significant, β =.10 (p < .05). After controlling for PTSS, the direct effect dropped to a nonsignificant level, β = .07 (p = .20) implying complete mediation by PTSS. The Sobel test showed that PTSS significantly mediated the association between witnessing interparental and dating violence, z = 2.50 (p

= .01). Conversely, PTSS did not mediate the association between a history of neglect and dating violence because neglect history was not significantly related to PTSS, β = .07 (p =.20).

(16)

Table 5.10.

PTSS Mediating the Association between Child Maltreatment Variables and Dating Violence

Regressions B SE β

Physical abuse

PTSS - Dating violence¹ 2.20 .69 .17**

Physical abuse - Dating violence¹ 1.14 .40 .15**

Physical abuse - Dating violence

controlling for PTSS 0.77 .41 .10

Sexual abuse

Mediation test 2.37 (p = .02)

Sexual abuse - Dating violence¹ 2.51 .56 .23**

Sexual abuse - Dating violence

Neglect

Neglect - PTSS¹ 0.08 .06 .07

Neglect - Dating violence¹ 3.46 .76 .23**

Neglect - Dating violence

Witnessing violence Mediation test 2.50(p = .01)

Witnessing violence - PTSS¹ 0.09 .03 .18**

Witnessing violence - Dating

violence¹ 0.70 .35 .10*

Witnessing violence - Dating

violence controlling for PTSS 0.46 .36 .07

* p < .05. ** p < .01

The regression coefficients and the Sobel test results of the mediation model involving childhood maltreatment variables and the overall score for externalizing problems are presented in Table 5.11. The model testing CPA, showed significant associations between CPA and PTSS, β = .16 (p < .01) and between PTSS and externalizing problems, β = .40 (p < .01). The regression weight of the total effect of CPA on externalizing problems was also significant, β = .30 (p < .01). After controlling for PTSS, the regression weight of the direct effect dropped but remained significant, β = .20 (p < .01). The Sobel test showed that PTSS partially mediated the association between CPA and externalizing problems, z = 3.16 (p

(17)

In contrast, PTSS did not mediate the association between a history of neglect and externalizing problems because a history of neglect was not significantly related to PTSS, β = -.07 (p =.19). In the same way, PTSS did not mediate the association between witnessing interparental violence and externalizing problems because witnessing interparental violence was not significantly related to PTSS, β = .07 (p = .16).

Table 5.11.

PTSS Mediating the Association between Child Maltreatment Variables and Externalizing Problems

Physical abuse

PTSS - Externalizing¹ 0.60 .07 .39**

Physical abuse – Externalizing¹ 0.28 .05 .30**

Physical abuse - Externalizing

Sexual abuse

Sexual abuse – Externalizing¹ 0.56 .06 .42**

Sexual abuse - Externalizing

Neglect

Neglect - PTSS¹ -0.08 .06 -.07

Neglect – Externalizing¹ 0.67 .09 .37**

Neglect - Externalizing controlling

for PTSS 0.59 .08 .33**

Witnessing violence

Witnessing violence - PTSS¹ 0.04 .03 .07

Witnessing violence – Externalizing¹ 0.04 .04 .30**

Witnessing violence - Externalizing

* p < .05. ** p < .01.

The model testing CPA, PTSS and depressive symptoms showed a significant association between CPA and PTSS, β = .20 (p < .01) and between PTSS and depressive symptoms, β = .36 (p < .01) (see Table 5.12). Further, the regression weight of the total effect of CPA on depressive symptoms was significant, β = .23 (p < .01). After controlling for PTSS, the regression weight of the direct effect dropped but remained significant, β = .12 (p < .01). The Sobel test results showed that PTSS partially mediated the association of CPA and depressive symptoms, z

= 2.85 (p < .01). Similarly, the model testing CSA, PTSS and depressive symptoms showed a significant association between CSA and PTSS, β = .23 (p < .01), and between PTSS and depressive symptoms, β = .35 (p < .01). The regression weight

(18)

for the total effect of CSA on depressive symptoms was β =.27 (p < .01). After controlling for PTSS, the regression weight of the direct effect dropped but remained significant, β = .16 (p < .01). The Sobel test showed that the partial mediation of the association between CSA and depressive symptoms by PTSS was significant, z = 2.94 (p < .01).

In the same way, the model involving witnessing interparental violence in childhood, PTSS and depressive symptoms, showed significant associations between witnessing interparental violence and PTSS, β = .13 (p < .01), and between PTSS and depressive symptoms, β = .38 (p < .01). The regression weight of the total effect of witnessing interparental violence on depressive symptoms was β

=.18 (p < .01). After controlling for PTSS, the direct effect dropped but remained significant, β = .10 (p = .03). The Sobel test showed that PTSS partially mediated the association between witnessing interparental and depressive symptoms, z = 2.54 (p < .01). In contrast, PTSS did not significantly mediate the association between a history of neglect and depressive symptoms because childhood neglect was not related to PTSS, β = -.06 (p = .27).

Table 5.12.

PTSS Mediating the Association between Child Maltreatment Variables and Depressive Symptoms

Child Maltreatment Regressions B SE β

Physical abuse

Mediation test 1.33 (p = .19)

Physical abuse – Depressive

symptoms¹ 0.12 .03 .20**

PTSS – Depressive symptoms¹ 0.33 .05 .36**

Physical abuse – Depressive

symptoms¹ 0.12 .03 .23**

Physical abuse - Depressive

symptoms controlling for PTSS 0.07 .03 .12*

Sexual abuse

PTSS - Depressive symptoms¹ 0.32 .05 .35**

Sexual abuse – Depressive

symptoms¹ 0.21 .04 .27**

Sexual abuse – Depressive

symptoms controlling PTSS 0.13 .04 .16**

Neglect

Neglect - PTSS¹ -.07 .06 -.06

PTSS - Depressive symptoms¹ 0.33 .04 .36**

Neglect – Depressive symptoms¹ 0.38 .05 .39**

Neglect - Depressive symptoms

(19)

The regression coefficients and the Sobel test results of the mediation models involving childhood maltreatment variables and borderline personality symptoms are shown in Table 5.13. The model testing CPA, PTSS and borderline personality symptoms showed that the association between CPA and PTSS was significant, β = .16 (p < .01). Similarly, the association between PTSS and borderline personality symptoms was significant, β = .50 (p < .01). The regression weight of the total effect of CPA and borderline personality symptoms was also significant, β = .25 (p < .01). After controlling for PTSS, the regression weight of the direct effect dropped but remained significant, β = .11 (p = .01).The Sobel test indicated that PTSS partially mediated the association of CPA and borderline personality symptoms, z = 3.35 (p < .01).

In the same way, the model testing CSA, PTSS and borderline personality symptoms showed significant associations between CSA and PTSS, β = .20 (p <

.01), and between PTSS and borderline personality symptoms, β = .50 (p < .01). The regression weight for the total effect of CSA on borderline personality symptoms was β = .25 (p < .01). After controlling for PTSS, the regression weight of the direct effect dropped to a nonsignificant level, β = .09 (p < .06), implying complete mediation by PTSS. Consequently, the Sobel test indicated that the mediation of CSA and borderline personality symptoms by PTSS was significant, z = 4.21(p <

.01).

The model involving witnessing interparental violence showed a significant association between witnessing interparental violence in childhood and PTSS, β = .10 (p = .03) and between PTSS and borderline personality symptoms, β = .51 (p < .01). The regression weight of the total effect of witnessing interparental violence on borderline personality symptoms was significant, β =.19 (p < .01).

After controlling for PTSS, the direct effect dropped to β = .09 (p = .04). The Sobel test showed that PTSS partially mediated the association between witnessing interparental and borderline personality symptoms, z = 2.18 (p = .03). In contrast, PTSS did not mediate the association between a history of neglect and borderline personality symptoms because a history of neglect was not related to PTSS.

(20)

Table 5.13.

PTSS Mediating the Association between Child Maltreatment Variables and Borderline Personality Symptoms

Physical abuse

PTSS – Borderline symptoms¹ 0.45 .04 .50**

Physical abuse – Borderline

symptoms¹ 0.14 .03 .25**

Physical abuse - Borderline

symptoms controlling for PTSS 0.06 .03 .11**

Sexual abuse

Sexual abuse – Borderline

symptoms¹ 0.20 .04 .25**

Sexual Abuse - Borderline

symptoms controlling for PTSS 0.07 .04 .09

Neglect

Neglect - PTSS¹ .00 .05 .00

Neglect – Borderline symptoms¹ 0.23 .06 .21**

Neglect - Borderline symptoms

Witnessing violence Mediation test 2.18 (p = .03)

Witnessing violence - PTSS¹ 0.05 .02 .10*

Witnessing violence – Borderline

symptoms¹ 0.09 .03 .19**

Witnessing violence - Borderline

controlling for PTSS 0.04 .02 .09*

* p < .05. ** p < .01

(21)

Table 5.14 presents the regression coefficients and the Sobel test results of the mediation models involving childhood maltreatment variables and the overall score for internalizing problems. In the model testing CPA, PTSS and internalizing problems, the association between CPA and PTSS was significant, β

= .15 (p < .01). Similarly, the association between PTSS and internalizing problems was significant, β = .49 (p < .01). The regression weight of the total effect of CPA on internalizing problems was also significant, β = .27 (p < .01). After controlling for PTSS, the regression weight of the direct effect dropped but remained significant, β = .14 (p < .01). The Sobel test showed that PTSS partially mediated the association between CPA and internalizing problems, z = 3.09 (p <.01). In the same way, the model testing CSA, PTSS and internalizing problems showed significant associations between sexual abuse and PTSS, β = .18 (p < .01) and between PTSS and internalizing problems, β = .48 (p < .01). The regression weight for the total effect of CSA on internalizing problems was β = .29 (p < .01). After controlling for PTSS, the regression weight of the direct effect of CSA on internalizing problems dropped but remained significant, β = .14 (p < .01). Consequently, the Sobel test showed that the partial mediation of the association between CSA and internalizing problems by PTSS was significant, z = 2.89 (p < .01).

The model testing the mediation of witnessing interparental violence and internalizing problems by PTSS also showed significant associations between witnessing interparental violence and PTSS, β = .09 (p =.05) and between PTSS and internalizing problems, β = .51 (p < .01). Further, there was a significant association between witnessing interparental violence and internalizing problems β = .21 (p < .01). After controlling for PTSS, the regression weight of the direct effect of witnessing interparental violence on internalizing problems dropped but remained significant, β = .11 (p =.01). The Sobel test showed that PTSS partially mediated the association between childhood physical abuse and internalizing problems, z = 1.97(p = .05). In contrast, PTSS did not mediate the association between a history of neglect and internalizing problems because a history of neglect was not significantly related to PTSS β = -.08 (p = .09).

(22)

Table 5.14.

PTSS Mediating the Association between Child Maltreatment Variables and Internalizing Problems

Physical abuse

PTSS – Internalizing¹ 0.40 .04 .49**

Physical abuse - Internalizing¹ 0.13 .02 .27**

Physical abuse - Internalizing

Sexual abuse

Sexual abuse - Internalizing¹ 0.20 .03 .29**

Sexual abuse - Internalizing

Neglect

Neglect - PTSS¹ -0.09 .05 -.08

Neglect - Internalizing ¹ 0.32 .05 .34**

Neglect - Internalizing controlling

for PTSS 0.27 .04 .29**

Witnessing violence Mediation test 1.97 (p = .05)

Witnessing violence - PTSS¹ 0.05 .02 .09*

Witnessing violence - Internalizing¹ 0.05 .02 .21**

Witnessing violence - Internalizing

* p < .05. ** p < .01.

5.4. Mediation of child maltreatment variables and psychopathology outcomes by social desirability

Similar analyses were conducted to determine if social desirability significantly mediated the association between childhood maltreatment variables and psychopathology outcomes. In more than half of the mediation models, social desirability was not a significant mediator. It is noteworthy however, that the mediation by social desirability was significant in the models that involved CPA

(23)

the association between CPA and borderline personality symptoms, (z = 2.02, p

= .04), CSA and borderline personality symptoms (z = 2.50, p = .01), and a history of neglect and borderline personality symptoms (z = -2.05, p = .04). The analyses also showed that social desirability was a significant mediator of the associations between CPA and internalizing problems (z = 2.15, p = .03), CSA and internalizing problems (z = 2.28, p = .02), and a history of neglect and internalizing problems (z = -2.91, p < .01). In the Kenyan sample therefore, PTSS was a more a potent mediator of the association between child maltreatment variables and psychopathology outcomes than was social desirability.

In sum, the results of our Kenyan sample show that males had more externalizing psychopathological symptoms than females. They also did report having experienced more child maltreatment than did female respondents.

Males and females were however comparable in their experience of PTSS and internalizing behaviour problems. The hierarchical regression models showed that males were more likely to report antisocial personality symptoms, criminal tendencies and dating violence than were females. Younger respondents were also more likely to report antisocial personality symptoms, criminal tendencies and borderline personality symptoms than were older respondents. Notably, PTSS was significantly related to all psychopathology assessments in this study.

After controlling for social desirability, the background variables and PTSS, a history of neglect was the most prominent predictor of all the psychopathological symptoms. CSA was predictive of criminal tendencies and dating violence only.

Witnessing violence on the other hand was associated with antisocial personality symptoms and criminal tendencies while CPA was related to criminal tendencies only.

Further, the mediation models showed that PTSS significantly mediated the association between CSA and all assessments of psychopathological symptoms. Similarly, the mediation of the association between CPA and all psychopathological symptoms was significant. This implies that the effect of CSA and CPA on the various psychopathology outcomes was in part or completely through PTSS. The mediation of the association between witnessing interparental violence and psychopathology outcomes by PTSS was significant for all psychopathology variables, except for the models which involved criminal tendencies and externalizing problems. It is remarkable that PTSS did not mediate any associations between a history of neglect and the psychopathology outcomes among Kenyan university students.