Using hospital discharge data for determining neonatal morbidity and mortality: a validation study.

Area of Science:

• Neonatal Health
• Population Health Data
• Medical Record Accuracy

Background:

• Limited reports exist on the accuracy of neonatal diagnostic/procedure codes in population health data.
• Hospital discharge data is widely used but its accuracy for neonatal conditions is under-examined.
• This study addresses the need to validate routinely collected data against a specialized audit.

Purpose of the Study:

• To assess the accuracy of hospital discharge data for identifying neonatal morbidity.
• To compare routinely collected data with a statewide audit of neonatal intensive care unit (NICU) admissions.
• To determine the suitability of discharge data for neonatal risk factor analyses.

Main Methods:

• A validation study comparing linked hospital discharge/birth data with NICU audit data.
• Included 2,432 infants admitted to NICUs in New South Wales, Australia (1994-1996).
• Calculated sensitivity, specificity, and positive predictive values (PPVs) for 12 diagnoses and 6 procedures.

Main Results:

• Specificities for all items exceeded 85%; PPVs ranged from 70.9% to 100%.
• High PPVs (>92%) were observed for in-hospital mortality, low birthweight, retinopathy of prematurity, respiratory distress syndrome, meconium aspiration, pneumonia, pulmonary hypertension, major anomalies, mechanical ventilation, and specific surgical conditions.
• Transient tachypnea of the newborn (70.9%) and drainage of air leak (83.6%) had the lowest PPVs.

Conclusions:

• Routinely collected hospital discharge data demonstrates high PPVs for most neonatal morbidities, making it suitable for risk factor analysis.
• While under-ascertained, the data's accuracy is sufficient for many epidemiological purposes.
• Procedures were generally recorded more accurately than diagnoses in the discharge data.