Résumé :
|
[BDSP. Notice produite par INIST-CNRS wVn6R0xf. Diffusion soumise à autorisation]. The author evaluated the validity of four historically based variables collected on Georgia birth certificates : outcome of preceding pregnancy, history of delivery of a low- (<2,500 g) or high- (>4,000 g) birth-weight infant, and death of the baby resulting from the preceding pregnancy. Data were derived from birth and fetal death certificates that were linked for the first and second deliveries of 231,075 women in Georgia from 1980 through 1995. Deaths that occurred during the infant's first year of life were also linked to the birth certificate. For all but the survival variable, the outcome of the first birth as reported on the certificate for the second birth was compared with the outcome recorded on the certificate for the first birth, which was assumed to be correct. Except for ascertainment of death of the firstborn infant, sensitivities for the history of poor outcomes were low. Furthermore, sensitivities were higher when an extremely adverse outcome occurred in the first pregnancy or an adverse outcome recurred. The only high sensitivity was for past infant death (85.4%). These results suggest caution when using these variables to identify high-risk subsets for further research or control for confounding. Am J Epidemiol 2001 ; 154 : 883-8.
|