Maternal Recall Versus Medical Records of Metabolic Conditions from the Prenatal Period: A Validation Study.

To assess validity of maternally-reported diabetes and hypertensive disorders, and reliability of BMI measurements during periconception and pregnancy compared with medical records when mothers are interviewed 2-5 years after delivery. To investigate whether reporting accuracy differed by child's case status (autism, delays, typical development). Participants were mothers of 2-5 year old children with and without neurodevelopmental disorders from the CHARGE (CHildhood Autism Risks from Genetics and the Environment) Study who had both prenatal/delivery records and telephone interviews. Sensitivity and specificity of self-report in telephone interview was assessed by comparison with medical records; agreement was evaluated by kappa statistics. Deviations in reported BMI were evaluated with Bland-Altman plots and concordance correlation coefficient (CCC). Mothers of children with neurodevelopmental disorders (autism or developmental delay) reported metabolic conditions slightly more accurately than control mothers. For diabetes, sensitivity ranged from 73 to 87% and specificity was ≥98% across groups. For hypertensive disorders, sensitivity ranged from 57 to 77% and specificity from 93 to 98%. Reliability of BMI was high (CCC = 0.930); when grouped into BMI categories, a higher proportion of mothers of delayed children were correctly classified (κ(wt) = 0.93) compared with the autism group and controls (κ(wt) = 0.85 and κ(wt) = 0.84, respectively; P = 0.05). Multiparity was associated with higher discrepancies in BMI and misreporting of hypertensive disorders. For purposes of etiologic studies, self-reported diabetes and hypertensive disorders during periconception and pregnancy show high validity among mothers irrespective of child's case status. Recall of pre-pregnancy BMI is reliable compared with self-reported values in medical records.