It does seem as though this intervention had a measure of success – based on their primary outcome – but there’s more shades of grey throughout the article.
This is a prospective, controlled trial of a contextual computer decision-support (CDS) incorporated into the computerized provider order entry (CPOE) system of their electronic health record (EHR). They do a four-phase On/Off intervention where the CPOE either suggests alternative medications or dose reductions in patients >65 years of age. They look at whether the intervention changed the rate at which medication ordering was compliant with medication safety in the elderly, and then, secondarily, at the rate of 10-fold errors, medication cancellations, and adverse drug event reports.
The oddest part of this study is their choice of primary outcome measure. Ideally, the most relevant outcome is the patient-oriented outcome – which, in this case, ought to be a specific decrease in adverse drug events in the elderly. However, and I can understand where they’re coming from, they chose to specifically evaluate the usability/acceptability of the CDS intervention to verify the mechanism of intervention. There are lots of studies out there documenting “alert fatigue”, resulting in either no change or even increasing error rates.
As far as the main outcome measure goes, they had grossly positive findings – 31% of orders were compliant during the intervention periods vs. 23% of orders during the control periods. But, 92.5% of recommendations for alternative medications were ignored during the intervention periods – most commonly triggered by diazepam, clonazepam, and indomethacin. The intervention was successful in reducing doses for NSAIDs and for opiates, but had no significant effect on benzodiazepine or sedative-hypnotic dosing.
However, bizarrely, even though there was just a small difference in guideline-concordant ordering, there was a 4-fold reduction in adverse drug events – most of which occurred during the initial “off” period. As a secondary outcome, there’s much to say about it other than “huh”. None of their other secondary outcomes demonstrated any differences.
So, it’s an interesting study. It is consistent with a lot of previous studies – most alerts are ignored, but occasionally small positive effect sizes are seen. Their primary outcome measure is one of mostly academic interest – it would be better if they had chosen more clinically relevant outcomes. But, no doubt, if you’re not already seeing a deluge of CDS alerts, just wait a few more years….
“Guided medication dosing for elderly emergency patients using real-time, computerized decision support”
http://www.ncbi.nlm.nih.gov/pubmed/22052899