Completeness of Nebraska's 2004 hospital discharge data — how much is missing?

Background: Hospital discharge data (HDD) are used to track multiple public health conditions. Despite its nationwide importance, studies that evaluate reporting completeness of HDD are limited. In Nebraska, a 2004 change in hospital data submission forced the state's hospital association to alter procedures for compiling HDD. Consequently, a discharge reporting decline of 10% from 2003 to 2004 might have compromised public health programs that use HDD to track conditions and assess the impact of interventions. Methods: We compared Nebraska 2004 HDD with two vital records databases to estimate the number of missing HDD records. Nebraska birth and death certificate records were matched with HDD at the facility level. These records were linked by using common variables: admission date, discharge date, age, sex, zip code, and delivery method for births. Results: Of 24,620 singleton in-hospital births identified by using vital records, 19,915 (80.9%) were contained in HDD. Of 5,972 in-hospital deaths identified by using vital records, 4,031 (67.5%) were contained in HDD. For 65 hospitals providing obstetric services, reporting completeness ranged from 20.6% to 104.9% (median, 80.3%). For 88 hospitals with one or more deaths, reporting completeness ranged from 0% to 100.0% (median, 58.0%). Conclusions: These results indicate substantial underreporting of both in-hospital singleton births and deaths in Nebraska's 2004 HDD and calls into question reporting completeness for other health events requiring hospitalization. Public health programs that rely on these data should interpret results of HDD analysis with caution. Further study is needed to characterize which records are underreported and why.