Results

The study comprised 649 children (372 girls) born in Newcastle upon Tyne between 17 March 1980 and 9 August 1989 in two hospitals. The birth weight as recorded in the hospital records ranged from 1700 g to 4995 g (mean, 3380), and there was some evidence that the figures had been rounded off to 0 and 5. The difference between recalled birth weight and recorded birth weight ranged from −1190 g to +1390 g (mean, −11; median, −6).

Recalled and recorded birth weights were highly correlated (r = 0.95) and the differences were not significant (p = 0.71). Table 1 shows that over 90% of recalled birth weights were within 200 g of that recorded. No significant associations were observed between the difference in birth weights and social class (p = 0.65), sex (p = 0.87), hospital (p = 0.32), age of child at time of interview (p =  0.49), or planned feeding method (p = 0.39). Table 2 divides the recorded birth weight into various groups, and the recalled birth weight at either extreme was not significantly better than the middle group. Accuracy of recalled birth weight did not vary according to the recorded birth weight (p = 0.44).

Discussion

Obtaining hospital birth records to verify birth weight can be tedious, particularly if the individual was born a number of years ago or in another part of the country. When asked, parents are usually quick to remember their child's birth weight. This is not surprising because this piece of information is usually eagerly awaited at the time of delivery and also repeatedly relayed to friends and family, thereby imprinting it to memory. However, time can blur the details,10 11 and the figure may be “rounded off”12 or confused with other deliveries.

A number of studies on birth weight3-8 have used maternal recall as the basis of their data and it is obviously important that the validity of this is checked. An Irish study13 of 99 mothers concluded that recall accuracy was better for the higher social classes and became worse as the child's age increased. A systematic error in birth weight recall where the error is greater for lower social class could introduce significant bias into studies of risk factors for ischaemic heart disease, where social class is in itself an important risk factor. Three UK studies14-16 found that the accuracy of recalled birth weight was good, but that even a few months after delivery there was a tendency for mothers to report slightly higher birth weights.14

Analysis of data from the USA diethylstilbestrol study17found good to excellent recall of birth weight, but there was a significant difference in accuracy between the study parents and the mothers of a control group. The reasons for this difference between groups were not explored.17 A study from Israel of 97 women, who had at least seven children each, found that about 75% of reported weights were accurate to within 100 g.18 Recall was best when children were less than 4 years old but was not related to maternal age or education. A Swedish study of laboratory workers found that after about 7 years, recalled birth weight was within 50 g of the recorded weight in 87% of the study population.19However, these results of studies of specific ethnic18 and occupational groups19 might not be applicable to the general population, and further studies were needed.

We found in a relatively large study, representing all social classes, that parental recall was good, with 75% of recalled birth weights being within 50 g of that recorded in the hospital records. We found that recall was not different for different social classes and was good for children born up to 15 years previously. Other studies have found more inaccuracy at the extremes of weight,12 17 but this was not evident in our study. In general, it appears that recalled birth weights are sufficiently accurate for use in epidemiological investigations. In particular, because our study found no association with social class or the age of the child, any error will be random and therefore not likely to introduce systematic bias.