Recall error and recall bias in life course epidemiology

摘要

OBJECTIVES I propose a distinction between recall error and recall bias and examine the the effect of childhood financial hardship on adult health, subject to such recall problems. Studying the effect of childhood hardship on adult health is a prototypical investigation in life course studies where both non-clinical factors and long-duration processes are at play in determining health outcome. These factors and processes are often elicited retrospectively. Unfortunately, retrospective information on childhood hardship is often subject to recall error and recall bias. There is surprisingly little methodological work on how to purge their effects in retrospective life course studies. METHODS I recast a variant of generalised latent variable models as covariate error measurement model to purge recall error in life course study. Additionally, I recast the endogeneous treatment model as a solution to the problem of recall bias. I apply both models to examine the effect of childhood financial hardship on adult health status of more than 359,000 European respondents from 23 countries. In addition, I validate the solutions using the National Child Development Study cohort where both prospective and restrospective information are available. RESULTS Childhood financial hardship has a strong effect on adult health status. Once recall error is accounted for in a generalised latent variable model, the effect reduced by an order of magnitude though remain statistically significant. Applying the endogeneous treatment model of recall bias suggests that childhood hardship is systematically misreported by respondents. Once this bias is purged, the effect of childhood deprivation on adult health increased markedly. Such an increase is consistent with multiple direct and indirect pathways linking childhood hardship and adult health. CONCLUSION Problems of recall error and recall bias are common in life course retrospective studies. Applied to data from 23 European countries, the proposed solutions recover the effect of childhood hardship on adult health outcome.