Friday, 14 January 2005 - 12:00 PM

This presentation is part of: Poster Session I

Child Maltreatment and Insecure Attachment: A Meta-Analysis

Purpose: This study builds upon the empirical knowledge of attachment and child maltreatment. A review of the literature on maltreated infants (age 1 to 48 months) indicates that they are more likely to have an insecure attachment style than non-maltreated children as measured by Ainsworth’s Strange Situation protocol(Ainsworth & Wittig, 1969; see Morton & Browne for a full review). While a number of studies report the co-occurrence of maltreatment and attachment, the strength of this relationship is unclear. The purpose of this paper is to evaluate the effect size of the relationship between child maltreatment and insecure attachment. Method: Study exclusion criteria were: no control groups, variation in operational definition of maltreatment, and samples drawn from the same population. Only the study with the largest N from each population was included. Because data in the reviewed studies reported on abuse, neglect, as well as a combination of abuse and neglect, for this analysis, all categories were collapsed into a Maltreated category. Attachment status was dichotomized into Secure and Insecure categories. Studies were coded for the following characteristics: sampling, maltreatment criteria, Ainsworth’s method of classifying attachment style, statistical analysis, and results. A number of the original coding categories were discarded due to lack of information, or lack of variability between studies. Most of the discarded categories (i e., the country in which the research was conducted, the method of assessing attachment, and coding protocols) were dropped because they were consistent across studies. Analyses: Biostat’s Comprehensive Meta-Analysis software was used to compute effect size and significance levels. The program uses inverse variance weights to compute a mean odds ratio effect size. Results: As was hypothesized, we found that there is a statistically significant relationship between child maltreatment and insecure attachment style. In the preliminary analysis (N=5), the N’s of the studies ranged from 43 to 190, with a total of 517 individual cases. The maltreated children ranged in age from one to 48 months (M = 20.7). The Q test was significant, 19.68 (4), p < .01, indicating that a random effects model should be used. A significant effect size was found, N = 5, OR = 6.68, p < .01, with a large confidence interval (95%CI: 1.81 to 24.67). Egeland and Sroufe’s study had the largest N (190), and theirs were the only non-significant findings. Because it was an outlier, this study was removed from the final analysis. The N’s of the remaining studies ranged from 43 to 182, with a total of 327 individual cases. This improved the homogeneity of the distribution, Q= 2.91 (3), p > .05. Therefore, a fixed effects model could be used, finding a significant effect size, N=4, OR=12.28, p < .01, with a smaller confidence interval, (95%CI: 6.41 to 23.50). Discussion: The results validate previous findings of a relationship between maltreatment and insecure attachment style and suggest the need for further studies. Additionally, these findings have clinical implications for the assessment and treatment of attachment relationships in maltreated children and their families.

See more of Poster Session I
See more of Oral and Poster

See more of Celebrating a Decade of SSWR (January 13 - 16, 2005)