Well we haven’t even got CPOE systems in more than a small percentage of hospitals (4%-15% depending on who you believe) and their validity as a major weapon against medical errors is already being questioned. A study in Annals of Internal Medicine looked at all medication orders in a large hospital in Chicago, over the course of one week in 2002 and tried to figure out what difference a CPOE system would have made had one been in place:

A total of 1111 prescribing errors were identified (62.4 errors per 1000 medication orders), most occurring on admission (64%). Of these, 30.8% were rated clinically significant and were most frequently related to anti-infective medication orders, incorrect dose, and medication knowledge deficiency. Of all verified prescribing errors, 64.4% were rated as likely to be prevented with CPOE (including 43% of the potentially harmful errors), 13.2% unlikely to be prevented with CPOE, and 22.4% possibly prevented with CPOE depending on specific CPOE system characteristics.

As the authors note somewhat dryly:

Prescribing errors are common in the hospital setting

The implications are laid out in this article in Health-IT World. Some half of the serious medication errors would not be caught by a standard CPOE system unless it had high level decision support software combined with it. In fact most CPOE systems do, even if the authors study poo-poos them.

The issue in practice is that if you set the warning levels too low, the clinicians can make errors. If you set them too high, the physicians just hit over-ride all the time, and make errors. Physicians and hospital IT folks I’ve talked to about CPOE have tended to start on the lowest levels of “warnings” and slowly ratchet up. But any computer use needs intelligence, and CPOE is no exception. That doesn’t mean that it shouldn’t be tried, and slowly it will become more common.