USE OF BERNOULLI CENSUS AND LOG-LINEAR METHODS FOR ESTIMATING THE PREVALENCE OF SPINA BIFIDA IN LIVEBIRTHS AND THE COMPLETENESS OF VITAL RECORD REPORTS IN NEW YORK STATE

Abstract
Data from birth certificates (BC), death certificates (DC) and medical rehabilitation files (MR) were analyzed to estimate the livebirth prevalence of spina bifida in Upstate New York in 1969–1974 and the completeness of the data sources. Birth certificates listed about 68% of cases, death certificates about 27% and medical rehabilitation files about 25%. The three sources together, it is estimated, included only about 80% of cases in the population. For each source, comparisons of estimates of completeness derived using each of the other two as reference sources were found to be useful for evaluating the likelihood of source dependence. The estimated livebirth prevalence rate, adjusting for incomplete reporting and the observed negative dependence of MR and DC sources, was 0.85 per 1000 livebirths by both Bernoulli census and log-linear methods. Taking into account in addition evidence for a BC-DC positive dependence, the resulting prevalence rate estimates were slightly higher, 0.88 per 1000 by log-linear methods and 0.90 per 1000 by the Bernoulli census approach. In view of the likely BC-DC positive dependence, it is suggested that Bernoulli census estimates derived using only these two sources without some ancillary third data source are likely to be biased to a false low figure. Nevertheless, estimates from BC and DC alone may still be useful in establishing that the prevalence rate is above some minimum figure, for example a “breakeven” prevalence rate, in cost-benefit analyses of a possible prevention program.