Comprehensive Summary
This study investigates the effectiveness of using a machine learning model to recognize and document safety events for patients diagnosed with dementia. Through the use of particular keyword searches and manual coding framework, 1,387 event reports classifying typical instances of safety events and failures due to human error were identified. The model was determined to be highly specific, with a strong ability to identify cases of human error but a weaker ability to classify positive ones. Because human errors tend to be under-reported and improperly described in safety event reports, more contextualized language models can be utilized in tandem with this machine learning model.
Outcomes and Implications
Patients diagnosed with dementia are more likely to be admitted to the hospital compared to those without dementia, increasing their risk of safety events. Documenting these safety events can be slow, expensive, and prone to error. By developing a machine learning model to make the process more efficient and identifying the factors associated with high-risk safety events, hospitals can more effectively allocate resources to help the most vulnerable populations.