I've been nosing around my hospital bugging lots of people trying to them interested in computerised clinical decision support system (CDSS).
They say things to me like "that sounds like a lovely theory Erich, but how do you apply it?" And this is where I get stuck. In a big way. In late 2011, in my institution, it turns out to be hard to find good examples where some generic CDSS will immediately add value. This is because the simple opportunities already have well established systems (human and computer), and the complex example require some knowledge engineering infrastructure that doesn't yet exist.
So what makes a domain an attractive target for computerised clinical decision support system (CDSS)?
Point 3, getting information out, is also not so hard with email and SMS, coupled with some basic data like 'treating team' for inpatients and 'family doctor' for outpatients. This is good enough to deploy messages, even if it doesn't provide good process integration for messages that need to arrive at 'just the right time'.
Point 4, the knowledge representation problem, is easier in some domains than others and is probably subject to the law of diminishing returns.
So this leaves Point 5 as the major current stumbling block. Unfortunately there is a degree of "chicken and egg" with recording structured clinical information in computers. Without online concurrent CDSS there's not much of a business case for building the systems and getting clinicians to enter the data. Without a decent corpus of structured clinical data it is very hard to demonstrate value from CDSS.
They say things to me like "that sounds like a lovely theory Erich, but how do you apply it?" And this is where I get stuck. In a big way. In late 2011, in my institution, it turns out to be hard to find good examples where some generic CDSS will immediately add value. This is because the simple opportunities already have well established systems (human and computer), and the complex example require some knowledge engineering infrastructure that doesn't yet exist.
So what makes a domain an attractive target for computerised clinical decision support system (CDSS)?
- high error rate or high cost of non-computerised error prevention
- high cost if error is made
- ability to transmit CDSS conclusion to appropriate person in a timely manner
- ability to capture and represent the clinical knowledge in a computer processable format
- necessary patient information available within computer systems in a computer processable format
Point 3, getting information out, is also not so hard with email and SMS, coupled with some basic data like 'treating team' for inpatients and 'family doctor' for outpatients. This is good enough to deploy messages, even if it doesn't provide good process integration for messages that need to arrive at 'just the right time'.
Point 4, the knowledge representation problem, is easier in some domains than others and is probably subject to the law of diminishing returns.
So this leaves Point 5 as the major current stumbling block. Unfortunately there is a degree of "chicken and egg" with recording structured clinical information in computers. Without online concurrent CDSS there's not much of a business case for building the systems and getting clinicians to enter the data. Without a decent corpus of structured clinical data it is very hard to demonstrate value from CDSS.
No comments:
Post a Comment