Friday, November 15, 2013

This Is A Very Serious Problem That Needs A Solution If Clinical Decision Support Is To Make A Difference.

This appeared a little while ago.

Study: Half of CDS prescription alert overrides are inappropriate

October 31, 2013 | By Julie Bird
Providers override about half of the alerts they receive when using electronic prescribing systems, according to a new study that also finds only about half of those overrides are medically appropriate.
Researchers reviewed more than 150,000 clinical decision support (CDS) alerts on 2 million outpatient medication orders for the study, published online this week by the Journal of the American Medical Informatics Association (JAMIA).
The most common CDS alerts were duplicate drug (33 percent), patient allergy (17 percent) and drug interactions (16 percent.) Alerts most likely to be overridden, however, were formulary substitutions (85 percent), age-based recommendations (79 percent), renal recommendations (78 percent) and patient allergies (77 percent).
On average, 53 percent of alert overrides were considered appropriate, according to the study abstract. Only 12 percent of renal recommendation alert overrides were deemed appropriate, compared with 92 percent for patient allergies.
The researchers concluded that refining the alerts could improve relevance and reduce alert fatigue.
Alert fatigue and other misuses of EHRs can cause serious problems.
More here with links.
I look forward to suggestions as to what can be done to keep this incidence low. Insist on a reason for the override being recorded maybe?
What do you think?


Andrew McIntyre said...

One of the annoying things in CDS is how people want results now. What you quickly realize is that the quality of CDS is a N squared relationship with the quality of the data. A small increase in data quality leads to much better CDS, but any drop in data quality makes CDS much worse very quickly.

I have observed government sponsored CDS projects in Australia where all they want is something visual to show and want the results quickly. You end up with something pretty but dumb and very manual - like a web site.

At some point we need to abandon the grand result in 3 month mentality and just concentrate on the basics - like terminology and compliant messages and basic structuring of data.

Its not sexy, but in reality its not expensive, it just needs attention to detail and public servants with deeper knowledge and patience. Its like eliminating cholera by building the London sewers. Its not pretty or sexy and there is not a lot to see but the results speak for themselves.

The medical IT sewers are in a parlous state, but because no one in power ever goes down to that level to look it never gets the attention it needs. Lots of people would jump at the chance of doing CDS - if they had good data, it doesn't need a pile of government money.

The really sad part about health IT in Australia is the failure to do the cheap things that require some understanding. They poor money its the castles in the air and they crumble because no one attended to draining the swamp it was built on.

I suspect that a good dose of recession and frugality would actually be the right medicine. We need some governance, rather than fools with money.

Paul Fitzgerald said...

Most modern systems will provide the facility for the clinician to override any "suggestion", but should provide an audit trail as to who, what, why etc. At a minimum, it should require a reason so when everything goes pear shaped, there can be a root cause analysis done, with subsequent re-training if required. It shouldn't be used as a big stick, however, unless absolutely necessary.