Background: Several studies have suggested that maternal recall of offspring birthweight compares well to data from birth certificates or medical records. However, these studies describe relatively short recall periods and pertain to populations where hospitalized childbirth is the norm. Accuracy of maternal recall has not been confirmed after long recall periods or in populations where non-hospitalized childbirth is common.
Methods: Accuracy of recalled birthweights is assessed among 626 women interviewed at an average age of 43.3 years (standard deviation [SD] 1 year), at which time the average age of their 1297 offspring was 17.9 years (range 1-29 years, SD 4.8 years). One-third of these infants were delivered at home. Recalled birthweight was compared to hospital or Well Baby clinic records, available for 66% (861/1297) of all births.
Results: Record availability was not related to birthweight. For the 861 infants for whom both sources were available mean recorded birthweight was 3342 g (SD 586 g) and mean recalled birthweight 3340 g (SD 675 g). Recalled weights overestimated recorded weights by 109 g (95% Cl: 68-151) among infants weighing > 3750 g and underestimated recorded weights by 83 g (95% Cl: 54-111) among infants weighing < or = 3250 g.
Conclusions: As an outcome variable recalled birthweights were unbiased, but less precise than recorded birthweights. Use of recalled birthweight does not sacrifice study power however and may increase generalizability provided recalled weights are available for 30% more infants than recorded weights. For individual birthweight assessment and for use as an independent variable recalled weights are biased and may have to be adjusted using recorded data as available.