Nurse Stress Prediction — Wearable Sensors in Hospital [11.5M Records]

Overview

This dataset was collected from 15 female nurses aged 30 to 55 during regular hospital shifts across two collection phases (April–August 2020 and October–December 2020). Participants wore an Empatica E4 wristband that continuously recorded galvanic skin response (EDA), blood volume pulse (BVP), heart rate (HR), skin temperature (TEMP), and three-axis orientation data.

What makes this dataset valuable is its real-world clinical context: stress events were not lab-induced but occurred naturally during nursing shifts. Periodic smartphone surveys captured subjective stress levels and contributing contextual factors, providing labeled ground truth for supervised learning models.

The dataset is hosted on the Dryad Digital Repository and mirrored on Kaggle, making it openly accessible for occupational health, affective computing, and IoT wearable research.

Column Schema

Key Statistics

• Total Records: approximately 11,500,000
• Features: 9 columns
• Subjects: 15 nurses across two data collection phases
• File Format: CSV
• Device: Empatica E4 wristband
• Time Period: April 2020 – December 2020

Use Cases

• Occupational stress detection in healthcare workers using wearable IoT sensors
• Classification of physiological signals into stress vs. non-stress states
• Feature importance analysis for EDA, HR, and TEMP in stress prediction
• Benchmarking ML models (Random Forest, Decision Tree, LSTM) on real hospital data

Source & Attribution

The original dataset is hosted on the Dryad Digital Repository and was created as part of a hospital-based stress monitoring study using Empatica E4 wearables. A cleaned version is available on Kaggle contributed by Priyank Raval.