Cheng et al analyzed planned home birth and hospital outcomes (notably 5 minute Apgar, neonatal seizures) using national birth certificate data for 2008. Readers should consider several issues when weighing the validity of these findings. First, although the 2003 revision of the standard certificate of live birth classification differentiates planned from unplanned home birth, only 27 states adopted it by 2008, and several states had very limited experience with it. Although large, the final sample reported by Cheng et al after exclusions for missing data actually constitutes less than 50% (2,081,753 of 4,247,694) of all 2008 US births and excludes several large states (eg, Illinois, New York, and North Carolina).

Second, although most states now utilize electronic systems for birth certificate registration, the quality and timeliness of certification of home compared with hospital births has not been subjected to rigorous review. Whereas most hospitals employ dedicated birth certificate clerks, responsibility for completing birth certificates for home deliveries falls to the birth attendant, typically using worksheets that are then entered into the state vital statistics system. Why does this make a difference? Several studies have documented that stray marks or improper use of checkboxes on paper forms and data entry screens lead to false-positive reports. Even one case of neonatal seizures among planned home births in Table 2 of the report recorded in error would result in a crude odds of 2.30 (95% confidence interval, 1.03–5.17); 2 erroneous cases would result in a crude odds of 1.92 (95% confidence interval, 0.79–4.65). This level of error is well within that observed in many vital statistics validation studies.

Home birth certificates may also represent a longer period of exposure to the outcome of newborn seizure: although hospital births are registered electronically within 2 days of delivery, home births are often registered 5-10 days later.

Researchers often take for granted that data contained in public-use databases receive the same level of scrutiny accorded clinical research data. Although vital statistics data are occasionally validated, we have never seen any scientific assessment of the quality of birth certificate data comparing home with hospital births. Clinical data elements have limited sensitivity and poor positive predictive value. Data elements such as neonatal seizures should be used only with extreme caution with accompanying sensitivity analyses. Readers should interpret the findings of Cheng et al circumspectly and should not hold them as additional support for ‘things I believe’ without regard for evidence.