BACKGROUND
Poorly managed pain can lead to addiction, depression, suicide, worsening health, and increased use of health services. Most pain assessments occur in clinical settings, far removed from patients’ natural environment. Advances in smart home technology can, potentially, allow observation of pain in the home setting. Smart homes can be trained to recognize human behaviors and may be useful for quantifying functional pain interference thereby creating new ways of assessing pain and supporting people with pain who are living at home.
OBJECTIVE
To determine if a health-assistive smart home can detect pain-related behaviors to perform automated assessment and support intervention.
METHODS
A preliminary retrospective study was nested in a larger longitudinal smart home study. Archived ambient sensor and weekly nursing health assessment data from 11 independent older adults reporting pain across 1-2 years were co-analyzed. Two nurses using qualitative methods interpreted 27 unique pain events (acute, flare, chronic) represented in sensor-based data and illuminated 13 clinically-relevant associated behaviors. Data were used to train a machine learning model to recognize pain-related behaviors. The model extracted 550 behavioral markers for each sensor-based data segment. These were treated as both a binary classification problem (event, control) and a regression problem.
RESULTS
Qualitative analysis revealed six pain-related behavioral themes. Quantitative results were classified using a clinician-guided random forest technique which yielded a classification accuracy of 0.70, a sensitivity of 0.72, a specificity of 0.69, an area under the ROC curve of 0.756, and an area under the PRC curve of 0.777. The regression formulation achieved moderate correlation with r=0.42. Without clinician guidance, using standard anomaly detection techniques, an accuracy of only 0.16 was achieved.
CONCLUSIONS
Findings reveal that a pain-assessing smart home may viably support pain management. Utilizing clinicians’ real-world knowledge when developing machine learning models for healthcare delivery could improve the model’s performance. A larger study is warranted to improve model performance. Automated assessment may improve health outcomes for persons with chronic pain by facilitating timely contact with the healthcare team.