On-Demand Learning Lab: Advanced Data Science for Healthcare Quality Improvement-Learning Lab Handout

Learning Lab Handout - March 2025

Pdf Summary

The Duke Institute for Health Innovation (DIHI) presentation explains how advanced data science and AI can strengthen healthcare quality improvement (QI), safety, efficiency, and patient outcomes—without replacing QI professionals. It defines key terms (AI, machine learning, large language models) and emphasizes that AI’s value depends on thoughtful integration into real clinical workflows, supported by strong, safety-focused governance.

Speakers describe the evolution of QI practice (Model for Improvement, Lean/Six Sigma, root-cause analysis, human-centered design) and argue that traditional QI is often retrospective and manually intensive, limiting real-time decision-making. AI can reduce waste and cognitive load by automating data curation, enabling predictive analytics, and delivering point-of-care decision support—if implemented with clinician trust and careful lifecycle management.

Several real-world examples illustrate impact:
- <strong>Medication safety</strong>: a wrong-concentration event revealed deeper workflow issues beyond an individual failure to scan.
- <strong>Pediatric sepsis</strong>: collaboration between clinicians and data scientists produced human-centered ED alerts, improving early identification and speeding antibiotics, reducing sepsis-related harm.
- <strong>SEP-1 quality reporting</strong>: automating CMS measure data curation reduced costly manual review and outsourcing while maintaining accuracy.
- <strong>Operational and access improvements</strong>: LLM tools support surgical pre-authorization documentation, oncology intake summarization using retrieval-augmented generation, and inbasket message classification to reduce burnout and improve continuity.
- <strong>ED flow and Sepsis Watch</strong>: machine learning models support admission prediction and real-time sepsis risk monitoring, improving bundle compliance and mortality trends.

The presentation outlines organizational capabilities needed across technology integration, measurement/governance, regulatory compliance, patient-centered outcomes, workforce readiness, process improvement, and risk management. It concludes with a call for QI leaders to actively partner with data scientists, start with a single high-value pain point, iterate rapidly (PDSA/MVP mindset), align with strategic priorities, and “own the AI conversation” to ensure safe, equitable, patient-centered adoption.

Keywords

healthcare quality improvement (QI)

clinical AI integration

data science in healthcare

machine learning decision support

large language models (LLMs)

patient safety governance

predictive analytics for sepsis

SEP-1 measure automation

workflow automation and efficiency

human-centered clinical alerts