Registered Nurse Strain Detection Using Ambient Data: An Exploratory Study of Underutilized Operational Data Streams...

Area of Science:

• Nursing Informatics
• Healthcare Operations Research
• Machine Learning in Healthcare

Background:

• Registered nurses (RNs) experience strain when demand exceeds capacity, risking patient care.
• Monitoring RN strain is crucial but challenging due to a lack of comprehensive workload data.
• Ambient data from electronic tools may offer predictive insights into RN strain.

Purpose of the Study:

• To evaluate ambient workplace data for automated sensing of registered nurse (RN) strain.
• To determine the utility of non-electronic health record (EHR) data in predicting RN strain.

Main Methods:

• An exploratory retrospective study analyzed 1 year of ambient data from nurse call, medication dispensing, and communication systems.
• Supervised machine learning models were developed to classify work shifts based on unplanned overtime.
• Models predicted overtime at 8, 10, and 12-hour intervals within shifts.

Main Results:

• Classification accuracy for predicting unplanned overtime ranged from 57% to 64%.
• Predictive accuracy was highest at the end of the shift and lowest at 10 hours.
• Key predictors included communication device usage and medication delivery methods.

Conclusions:

• Ambient data streams can signal registered nurse (RN) strain, proxied by unplanned overtime, as early as 8 hours into a shift.
• This approach offers a pathway for early detection of RN strain.
• Proactive interventions can be developed to prevent missed or delayed patient care.