Background: Email communication between patients and healthcare providers is gaining popularity. However, healthcare providers are concerned about being inundated with patient messages and their inability to respond to messages in a timely manner. This work provides a text mining decision support system to overcome some of the challenges presented by email communication between patients and healthcare providers.
Method: A decision support system based on text mining algorithms was developed and tested to triage real world email messages into medium and highly urgent messages that are routed to health provider staff, or low urgency messages that could be routed to an automated response system, responding to the messages in a timely and appropriate way.
Results: Due to the length of email messages, feature reduction algorithms are inadequate in this context. Therefore, in this work, several different classifiers were combined and tailored to build a high performance classifier that supports this type of classification. The system was tested and proved to perform well with real-world patient messages that were exchanged with healthcare providers during a hypertension management study.