3-Sarah Rosetti
Introduction
Speaker and team acknowledgment.
Funding source: National Institute for Nursing Research and American Nurses Foundation.
Multisite collaboration: Columbia, Mass General Brigham, Vanderbilt University, Washington University in St. Louis.
Clinical Patient Safety Problem
Preventable inpatient deaths in hospitals are a major concern:
Estimated 200,000 cardiac arrest deaths annually in hospitals.
Estimated 130,000 sepsis-related deaths could be prevented.
Critical need for improved patient outcomes in hospitalization.
Nurses play a key role in detecting subtle signs of patient deterioration.
Nursing Documentation Contribution
Nurses increase documentation in response to patient concerns.
Documentation patterns provide insights into at-risk patient states.
Differences in perception of patient stability between nurses and physicians.
Physicians rely on physiological signs from EHR while nurses may detect early warning signs.
Communication barriers exist between nursing and medical staff regarding patient status.
Importance of Interdisciplinary Communication
Nurses' observations lead to increased surveillance and communication with physicians.
Challenges arise when physicians are not present to observe patient changes directly.
Effective communication among care teams is essential to avoid missed signals that indicate patient risk.
EHR Data Utilization
Analysis of flow sheet data from nursing documentation.
Importance of recognizing patterns of documentation beyond just values of vital signs.
Example: Comments made by nurses in EHR clarify patient context and significance of data.
Metadata analysis is crucial for understanding clinician documentation patterns.
Concern Study Objectives
Three aims for the Concern study:
Analyze nursing documentation patterns to determine notification thresholds for at-risk patients.
Develop and integrate a smart app notification system for risky patient states.
Evaluate the impact of the concerned smart app on patient outcomes (mortality, length of stay, etc.).
EHR Data Types and Challenges
Key data types collected for analysis:
Patient demographics, encounter information, flow sheet data, medication administration records, laboratory data, etc.
Data harmonization challenges between different EHR systems (e.g., Allscripts vs. Epic).
Need for structured and standardized data for optimal analysis and insights.
Health Care Process Modeling
The significance of understanding clinician behaviors through documentation patterns.
Expert knowledge in nursing contributes to improved predictive modeling for patient deterioration.
Certain behaviors (e.g., increased documentation frequency) may indicate worsening patient conditions.
Health care process modeling aims to capture and quantify these behaviors.
Predictive Modeling Approach
Utilization of machine learning for prediction of adverse patient outcomes.
Focus on capturing early warning signals through enhanced data points from EHR.
Importance of temporal dynamics in modeling (e.g., time of day, length of stay).
Early Warning System Development
Need for systems that predict deterioration based on nursing observations rather than solely physiological signs.
Early intervention capabilities enhanced by predictive modeling leading to better patient outcomes.
Interdisciplinary Focus Groups
Engaged nursing and physician input to inform the development of predictive models.
Importance of ensuring usability of predictive tools in clinical practice.
Concern Smart App Implementation
Presenting the smart app within the existing EHR workflow.
Alerts designed to minimize alert fatigue and focus on actionable patient risks (red, yellow, green scoring).
Continuous training and outreach to ensure clinician awareness and usage of the smart app.
Conclusion
Demonstrated the capability of using nursing documentation patterns as a predictor of patient deterioration.
Encouragement of further research into metadata and nursing insights.
Emphasized the necessity of clinician collaboration in modeling healthcare data effectively.