Development and validation of a retrospective self-report measure of childhood neglect

Background: Several measures are available to assess childhood physical and sexual abuse, but few measures focus specifically on neglect and little psychometric research on measures exists. This paper aims to fill a gap in the field by describing a new instrument to measure childhood neglect retrospectively and providing information about construct, predictive, and discriminant validity using adults with documented histories of childhood neglect. Methods: Data are from a large prospective, longitudinal study of abused and neglected children and matched controls followed up and assessed in adulthood. The current sample (N =717) includes 370 individuals with histories of childhood neglect and 347 demographically matched controls without those histories. Self-reports of childhood neglect were collected in in-person interviews at approximate age 40. Participants responded to a pool of items representing neglect. Missing responses were treated as substantive information in analyses. An optimal set of items was selected using Support Vector Machine (SVM) – a machine leaning algorithm. Neglect severity, diversity and SVM-based propensity scores were tested for predictive, construct and discriminant validity. Results: The optimal item subset included 10 items. The propensity scale measured with this optimal subset passed all validity tests, showing high predictive validity for neglect, discrimination between documented cases of neglect and abuse, and significant correlation with violence in adulthood (construct validity). The simple severity and diversity scores failed in at least one of the validity tests. Conclusions: This new instrument shows promise in detecting experiences of childhood neglect retrospectively. Missing responses were found informative in recollections of childhood neglect.