Using Machine Learning to Predict Mental Health Call Priority
Author Information
Author(s): Rana Rajib, Higgins Niall, Haque Kazi Nazmul, Burke Kylie, Turner Kathryn, Stedman Terry
Primary Institution: University of Southern Queensland
Hypothesis
Can machine learning accurately predict the priority of mental health calls based on voice characteristics?
Conclusion
The study demonstrates that machine learning can effectively classify mental health call priorities with a balanced accuracy of 92%.
Supporting Evidence
- The model achieved a balanced accuracy of 92%, indicating strong performance in identifying call priorities.
- High precision of 90% shows that most identified high-priority calls were indeed high-priority.
- The study utilized a dataset of 459 call records from a mental health helpline.
Takeaway
This study shows that computers can listen to people's voices and help decide how urgently they need help when they call for mental health support.
Methodology
The study used deep learning neural networks to analyze voice features from 459 call records to classify call priority.
Potential Biases
The AI model may lack the nuanced understanding of human emotions that clinicians possess.
Limitations
The model's effectiveness relies on the quality of triage categories, and it has not been tested in diverse populations.
Participant Demographics
198 males and 261 females, mean ages of 39 and 36 years respectively.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website