The Quality of National Birth Certificate Data for Neonatal Seizure Epidemiology
Abstract number :
3.343
Submission category :
16. Epidemiology
Year :
2016
Submission ID :
195759
Source :
www.aesnet.org
Presentation date :
12/5/2016 12:00:00 AM
Published date :
Nov 21, 2016, 18:00 PM
Authors :
Kristen Berry, Weill Cornell Medical College; Michael Pesko, Weill Cornell Medicine; Dale Hesdorffer, GH Serviesy Center and Department of Epidemiology, Columbia University; Renée A. Shellhaas, University of Michigan; Joanna Seirup, Weill Cornell Medicine
Rationale: Seizures are a common manifestation of neurological dysfunction in neonates and carry high risk for mortality and adverse long-term outcomes. National birth certificates are a potentially valuable source to study the epidemiology of neonatal seizures. However, the quality of the data is understudied. Methods: We reviewed national birth records from 2003-2013 to describe evidence of underreporting, rates of missing data, and effect of the 2003 revision of the birth certificate form. To measure underreporting, we compared estimates to a published reference rate of 0.95 neonatal seizures per 1000 live births - the "California rate" (Glass et al., Pediatr 2009; 154(1); 24-28 e21). We evaluated missingness by state and year using a two-way ANOVA, and evaluated missingness by demographic, infant health, and medical care factors using bivariate analyses. We developed 3 criteria for data utility and plausibility: (1) use of revised form for at least 4 years, (2) less than 1% missing data, and (3) reported seizure rate within 50% of the California rate. We reported which states met these criteria. Results: Of 22,834,395 live term births (≥36 weeks) recorded using the revised form from 2005-2015, there were 5,875 with neonatal seizures, suggesting a rate of 0.26 per 1000 term births, a quarter of the expected rate. The overall rate of missing seizure data was low (0.5%). However, missingness varied significantly by state and year (p < 0.001), as well as by demographic, infant health, and medical care factors (Table 1). After the 2003 birth certificate form revision, missing data (p < 0.001) and underreporting (Figure 1) increased. Nine states met criteria for utility and plausibility: IA, MT, NE, OR, SD, UT, VT, WA, and WY. Conclusions: Results show that birth certificate data are limited by underreporting, biased missingness, and changes in reporting subsequent to the 2003 revision. Data from nine states merit further investigation for their use in neonatal seizure epidemiology research. Funding: No external funding for this abstract
Epidemiology