Nursing documentation — care plans, shift notes, discharge summaries — contains rich clinical insights that structured codes cannot capture. We apply BERT-based NLP models trained on healthcare text to automatically extract, classify, and analyze these narratives at population scale, identifying quality signals invisible to manual review.

Nursing diagnoses represent a standardized taxonomy of patient conditions that guide care planning. We use transformer models — including BEHRT (BERT adapted for EHR data) — to automatically and accurately classify nursing diagnoses from ICD-coded administrative claims, enabling scalable quality measurement without manual coding.

Using longitudinal patient and nurse data, we build predictive models that surface early warning signals for adverse outcomes, care gaps, and quality deficiencies — enabling proactive intervention before problems escalate. Models are designed to be clinically interpretable, not just accurate.

Large nursing datasets — claims files, NCLEX records, state administrative data — require industrial-grade data engineering before any AI model can run. We design and build the ETL pipelines, data cleaning protocols, and storage architectures that make research possible at population scale.