PreMo: Clinical and Vital Sign Data for Very-Low-Birth-Weight Infants from 4 NICUs

Deidentified vital signs and clinical data for 3829 very-low-birth-weight infants from four neonatal intensive care units. The dataset includes REDCap form exports for demographics, discharge outcomes, intubation, and sepsis/NEC evaluations, alongside high-frequency heart rate and SPO2 vital sign recordings. It was created by researchers including Ian German Mesner and colleagues, with a manuscript pre-print dated 2026.

Use Cases

• Develop predictive models for sepsis or necrotizing enterocolitis (NEC) based on clinical evaluation data.
• Analyze patterns in heart rate and SPO2 vital sign time-series data to identify physiological precursors to clinical events.
• Study correlations between demographic/perinatal factors and discharge outcomes for very-low-birth-weight infants.
• Benchmark signal processing or machine learning algorithms on deidentified, high-frequency neonatal vital sign data.

Strengths

• Includes data from 3829 patients, providing a substantial cohort for analysis.
• Combines high-frequency (0.2-1 Hz) vital sign time-series with structured clinical event and demographic data.
• Data originates from four different neonatal intensive care units (NICUs), suggesting multi-site representation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download from the provided metadata file.
• Row counts for individual data files are unknown, which may limit suitability assessment for specific analyses.
• All dates have been deidentified into ages or timestamps, limiting precise temporal analysis of events.

Provenance

Source

University of Virginia (Libra Data) Harvested Dataverse, authored by Ian German Mesner et al.

Collection Method

Data collected from the PreMo 4-NICU study of very low birth weight infants, exported from REDCap and deidentified.

Freshness

Last updated 2026-04-26 06:10:07.

Geography

Data collected from four neonatal intensive care units (NICUs); specific locations are not provided due to deidentification.