As a healthcare enterprise looks after patients, information is gathered about various events that take place. Information about notable events, Admissions and Discharges, for example, is recorded in Hospital Information Systems or Patient Administration Systems. These systems typically broadcast event information in a form of HL7 messages for use by other enterprise systems, for example laboratory or diagnostic imaging. A stream of HL7 messages can be intercepted and processed to derive all sorts of interesting information.
The solution developed in this walkthrough deals with Excessive Length of Stay. Length of stay is defined as the period between patient’s admission to and discharge from the hospital. Statistical average expected length of stay is typically available for different kinds of patients presenting with different kinds of conditions. A significant variation from the average length of stay for specific patients may indicate complications, treatment errors, infections and other kinds of issues that the hospital needs to investigate. Notification of such incidents may help the hospital in addressing these issues and prevent future occurrences.
In this solution the Intelligent Event Processor is used to calculate the continuously updated average length of stay over a period of time and use it to compare against each event’s length of stay. It passes, to the downstream component, all events where the length of stay exceeds the average by 1 ½ times and ignores all others.
In the initial iteration, the solution reads a stream of discharge messages, containing admission date, discharge date, length of stay, and a bunch of other fields from a file and passes them to the IEP process. The IEP process keeps the window on the last 10 seconds worth of records and continuously calculates the average length of stay over all records in that window. As records are added to and removed from the window the average is recalculated. As each record is seen its length of stay is compared to the average length of stay of all records in the window at the time. If the length of stay in the current record is less then or equal to 1 ½ times the average at the same time the record is discarded. If the average is greater the record is ejected to the output and ultimately written to a file of exception records.
In a subsequent iteration the solution is modified to accept messages from a JMS Queue. This modification allows the solution to use the stream of discharge messages produced by the HL7 Processor solution, discussed in “HL7 Processor Demonstration - GlassFish ESB v2.1”.
In a further modification the solution is configured to send notification messages to another JMS Queue. Notification messages are processed by a different solution and sent to an email recipient.
The article and referenced materials are available at http://blogs.czapski.id.au/2009/09/glassfish-esb-v2-1-excessive-length-of-stay-healthcare-iep-demonstration