Nursing activity recognition using an inexpensive game controller: An application to infection control.

Area of Science:

• Biomedical Engineering
• Healthcare Technology
• Infection Control

Background:

• Hospital-acquired infections (HAIs) cause significant patient mortality, with poor hand hygiene being a primary contributing factor.
• Current electronic hand hygiene reminders lack the ability to detect specific nursing activities, limiting their effectiveness.
• Improving hand hygiene compliance is critical to reducing HAIs and enhancing patient safety.

Purpose of the Study:

• To develop and evaluate a low-cost sensor system for identifying nursing activities to trigger timely hand hygiene prompts.
• To assess the accuracy of different machine learning classifiers in recognizing nursing tasks based on accelerometer data.

Main Methods:

• Utilized a 3-axis accelerometer from a wireless game controller, with five sensors attached to nurses' wrists, upper arms, and backs.
• Recorded sensor data during six distinct nursing activities performed by eight nurses.
• Extracted features like mean, standard deviation, energy, and inter-axis correlation.
• Compared 1-Nearest Neighbour (1-NN), Decision Tree (J48), and Naïve Bayes classifiers for activity recognition.

Main Results:

• The 1-Nearest Neighbour (1-NN) classifier demonstrated the highest performance in identifying nursing activities.
• Achieved an average accuracy of 84% ± 2% across different sensor locations.
• Successfully differentiated between various nursing procedures using wearable sensor technology.

Conclusions:

• Wearable accelerometers, repurposed from game controllers, can effectively identify nursing activities.
• This technology holds promise for developing intelligent hand hygiene reminder systems to reduce hospital-acquired infections.
• Accurate activity recognition is key to providing context-aware prompts for improved healthcare worker compliance.