The association between birth weight and later life outcomes is of considerable interest in life-course epidemiology. Research often relies on self-reported measures of birth weight, and its validity is consequently of importance. We assessed agreement between self-reported birth weight and official birth records for Norwegian twins born 1967-1974. The intraclass correlation between self-reported birth weight and register-based birth weight was 0.91 in our final sample of 363 twins. It could be expected that 95% of self-reported birth-weight values will deviate from official records within a maximum of +446 grams and a minimum of -478 grams - around a mean deviation of 16 grams. Self-reported birth weight had a sensitivity of 0.78-0.89 and a positive predictive value of 0.59-0.85, and an overall weighted kappa of 0.71. We further assessed agreement by conducting two linear regression models where we respectively regressed self-reported birth weight and register-based birth weight on adult body mass index, a known association. The two models were not significantly different; however, there were different levels of significance in parameter estimates that warrant some caution in using self-reported birth weight. Reliability of self-reported birth weight was also assessed, based on self-reports in another sample of twins born 1935-1960 who had reported their birth weight in two questionnaires 34 years apart. The intraclass correlation was 0.86, which indicates a high degree of reliability. In conclusion, self-reported birth weight, depending on context and age when birth weight was reported, can be cautiously used.
Keywords: birth registry; birth weight; self-reports; twins.