The Journal recently published an article suggesting a causal association between electronic fetal monitoring (EFM) and decreased infant mortality. We have strong reservations regarding the ability of that paper to offer guidance regarding the effectiveness of EFM because it is inappropriate to use vital statistics data to make the leap from statistical association to causation.

Several years ago, two of us contributed to a set of American Journal of Obstetrics and Gynecology commentaries discussing the reasonable use of secondary vital statistics data. Those commentaries raised important issues and limitations that should be considered, acknowledged, and addressed.

The first issue is the potential inaccuracy of the EFM data. Validation of birth certificates has found that obstetric procedures, including EFM, are particularly poorly reported. Although random underreporting of EFM would bias estimates of the association between EFM and mortality toward finding no difference, it may be optimistic to assume that underreporting is random. We are aware of no study evaluating differential reporting of obstetric procedures, but it is plausible that reporting of intrapartum events may be less complete when, for example, an ill neonate is transferred shortly after birth. Differential reporting could bias the relative risk in either direction.

The second issue is confounding. Even if EFM were perfectly reported, we doubt that the factors controlled for by Chen et al, maternal age, race/ethnicity, education, marital status, self-reported tobacco and alcohol use, and infant sex, are the most relevant characteristics to explain how the 11% of labors that were not monitored differed from the 89% that were. Of great concern is the inability to ascertain the clinical circumstances of the labors. Among the many unmeasured potentially confounding factors are precipitous delivery or delivery in settings without easy access to EFM (and, perhaps, specialized obstetric care). These factors would be associated with both lack of EFM and adverse outcome.

Lastly, there is inappropriate use of causal language and inference in the description and interpretation of these cross-sectional administrative data. Within the limitations of birth certificate data, the paper by Chen et al may provide a helpful description of EFM use. However, it should not be used as a basis for clinical decision making, nor should the association between EFM and decreased infant mortality be deemed a causal one on the basis of this analysis.