The recent article comparing maternal and newborn morbidity among births at home, hospital, or birth centers by Wax et al, reported that infantss born at home more frequently experienced 5-minute Apgar scores below 7. The methodology used brings into question the validity of this conclusion.

This study used 2006 US Standard Certificates of Live Birth. Multiple exclusions were applied, resulting in inclusion of only 36.0% (745,690/2,073,368) of women in participating states. Inclusion of only slightly more than one-third of the potentially eligible population raises questions about the ability of birth certificates to identify women at low risk and consequently the generalizability of study findings.

Secondly, ascertainment of the type of birth attendant is missing for 4801 women or 0.6% of the sample. It is possible that at least some of these births were unattended. If this is indeed the case, then these births, which would be expected to have high rates of suboptimal outcomes, might be overrepresented in the home birth group, where failure of an attendant to arrive on time might be more likely. In addition, some women may have deliberately chosen to have an unattended home birth.

Because only 75% of the home births were recorded as attended by a physician or midwife, fully one-quarter may have been unplanned. Unplanned home births are well known to be at higher risk for adverse outcomes.

Lastly, the authors acknowledge that births for which complications necessitated transfer to hospital are attributed to hospital rather than to home or birth center births. In contrast to the above biases, this bias would favor home births. They also acknowledge that perinatal mortality was not measured, which eliminates deaths occurring during labor.

In view of these serious flaws, the statement that this study provides a “robust evaluation of maternal and newborn outcomes that is generalizable and reflects actual practice” cannot be supported. Without internal validity, placed in question by missing data and the inability to attribute births to planned place of birth, the issue of external validity or generalizability is irrelevant. Recent studies of place of birth in Canada and the Netherlands have used population-based perinatal databases with mandated participation by midwives and documentation of intended place of birth and attendant, as well as relevant outcomes, including intrapartum fetal death. American studies of place of birth must meet this new standard in order to draw valid conclusions and permit international comparisons.