Helping to Eliminate Mistakes in Medical Treatment: Our Challenges (Part 2)
By Loren Mack on Aug 15, 2007
Loren Mack is a design architect in xDesign who creates strategic and tactical designs for the Service Oriented Architecture/Business Integration group at Sun.
For this project, our design task involves finding a way to help health care professionals match records that the existing automated system can't.
The system is made up of two parts. The behind-the-scenes part consolidates a patient's records from various data sources to produce a single, complete, and up-to-date record of a particular patient. The system is even smart enough to see that records match even when some of the data in the records don't match.
The second part of the system is a user interface that reports records that might match, but that can’t be matched automatically. In this situation, the system needs some help from a "live operator standing by." Working with non-technical end users and providing them with awesome tools is one of the fun parts of this project.
When the system can’t resolve a conflict, the user interface alerts the health care professional and provides decision support to resolve the conflict. For example, when a baby is born, the hospital uses the father’s social security number as the baby's social security number on the birth certificate. Once common, this practice is now quite a headache for health care professionals later on, because the father and baby appear to be the same person. It's also a hard problem for the system to fix since the records of the father and baby may share the same data in many fields (like social security number and address), but the data in key fields are different (like name and birth date).
While there’s already a tool that lets a live human review these potential matches, it has many usability problems. It’s hard to tell what portions of a record don’t match. It’s hard to see information across more than one duplicate record (such as three systems all having similar, but slightly different data that could all be part of one person’s medical history). And it’s just plain slow.
These issues make it much harder for people to quickly and effectively handle records that the system can’t, and, in many cases, it takes much longer than necessary. Today, the people who do this work full-time print out huge lists of duplicate records and then spend hours reviewing the hard copies to make sure they’re matching the right information. The existing user interface could resolve these issues, but it doesn’t support their tasks well enough to be useful.
To be continued...