Considering the Source of Information in the Evaluation of Maltreatment Experiences-Reply

In Reply We thank Goltermann et al for their thoughtful Letter in response to our meta-analysis showing low agreement between prospective and retrospective measures of childhood maltreatment.¹ Goltermann et al are interested in the reasons for such low agreement. As highlighted in the Discussion section of our article,¹ we agree that low agreement may be partly attributable to methodological differences between prospective and retrospective measures.

In the meta-analysis, we tested whether agreement was moderated by the assessment instrument as well as other measurement characteristics. We found that the agreement did not differ according to whether child maltreatment was prospectively assessed through records (eg, child protection records or medical records), reports (eg, questionnaires or interviews), or mixed measures (eg, records and reports). However, agreement was higher when retrospective measures were based on interviews rather than questionnaires.

It was not possible to directly test whether the source of information moderated the agreement between prospective and retrospective measures because only 2 studies^2,3 used the same informant (self-reports) for both prospective and retrospective measures (Table 1¹). These studies found poor to fair agreement (κ = 0.34; 95% CI, 0.23–0.45² and κ = 0.05; 95% CI, 0.02–0.09³) similar to the overall meta-analytic estimate (κ = 0.19; 95% CI, 0.14–0.24¹). Furthermore, jackknife sensitivity analyses did not show that meta-analytic estimate was sensitive to the exclusion of these (or any other) studies. We also agree with the suggestion from Goltermann et al that future longitudinal studies using both identical reporting sources and instruments for prospective and retrospective measures would help to delineate the temporal stability (ie, test-retest reliability) of child maltreatment measures. However, in practice this is likely to be difficult owing to ethical and practical challenges in asking children to self-report on their maltreatment exposure via instruments used in adults.

Goltermann et al suggest that the agreement between prospective and retrospective measures may be affected by the age of children at the time of prospective assessment. It was not possible to test moderation by age at prospective assessment because this information was not available from many of the primary studies, partly reflecting the recurrent nature of maltreatment experiences. However, we found that agreement was not moderated by age at retrospective report.

Goltermann et al also highlight that retrospective reports may underestimate the prevalence of childhood maltreatment. We found that retrospective reports captured a higher prevalence of child maltreatment than prospective measures (eFigure 1 in the Supplement¹), indicating that retrospective measures may have greater sensitivity than prospective measures. Nevertheless, our findings suggested underdetection in both prospective and retrospective measures. For example, morethan half of individuals with prospective measures of maltreatment did not retrospectively report it. Likewise, more than half of individuals retrospectively reporting childhood maltreatment did not have concordant prospective measures.

In summary, our meta-analysis highlighted that current prospective and retrospective measures of childhood maltreatment identify largely different groups of at-risk individuals. We hope that future research will address the origins of these differences (including the methodological points highlighted by Goltermann et al), as well as the mechanisms underlying risk in different groups.