Objective: To compare the quality of data on previous pregnancies, based on maternal recall, to that of actual data in previous birth records, and to quantify to which extent recall based data bias results in epidemiologic studies on adverse reproductive outcomes.
Methods: Through linkage between agricultural censuses and the Medical Birth Registry, we identified 87,113 Norwegian female farmers with 174,764 single births in 1967-1991, including 533 late abortions (16-27 weeks). Data were linked into sibship records. In routine birth records, mothers provide information at each birth on previous fetal loss. Thus, we had information both based on recall and on the actual record.
Results: Among all mothers with a loss documented in previous records, 73.5% had a recall of previous loss (sensitivity), while 74.3% of all mothers reporting a previous loss also had a loss in previous sibship records (positive predictive value). Late abortion after a previous loss according to the sibship record was more frequent than after a previous loss based on maternal recall (18.3 vs. 10.8 per 1,000). There was a particular lack of recall for previous losses of short gestational age or when the current birth was a late abortion; both conditions tended to deflate the association between a previous and a current adverse outcome. Grain farming was associated with late abortion in mothers with previous loss in sibship records (odds ratio, 2.6) but not among mothers without previous loss (odds ratio, 1.0). When information on previous loss was based on maternal recall the odds ratios were 1.4 and 1.3, respectively.
Conclusions: Consecutive sibship records give more valid information than data derived from maternal recall.