Conceptualizing self-reported adverse childhood experiences: From epidemiologic exposure to psychometric latent variable

In this paper, we apply a measurement science perspective to explore both the epidemiologic and psychometric frameworks for the conceptualization, operationalization and assessment of self-reported adverse childhood experiences (srACEs). The epidemiologic paradigm suggests that srACEs are 'exposures', while the psychometric paradigm views responses on srACEs instrumentation as 'indicators'. It is the central premise of this paper that srACEs cannot be both exposures and indicators of scales. We review issues of reliability and validity from both perspectives, examine the degree of agreement between objective and subjective reports of childhood maltreatment and the implications of poor agreement, and conclude that the for the assessment of ACEs via self-report, the epidemiologic paradigm is not a good fit. We then review a number of reflexive and formative latent variable models that might be usefully fit to srACEs data for purposes of modeling structural properties of assessments, and/or to model ACE-health relationships. We highlight the mismatch and limitations of the reflexive measurement model for srACEs and conclude by endorsing either a formative latent variable model or application of latent class analyses. We emphasize the importance of considering and potentially formally testing competing measurement models and conducting both rationale analysis, conceptualization and hypothesis-testing to assess the fit of these models.