A Radiology Decision-Support Fail

It's harder than it looks!

Wouldn’t you like to have your radiology ordering system abuse you like this?

Turns out, the clinicians in Germany – for whom this custom, imaging-appropriateness decision-support was designed – did not appreciate its brutalist beauty, either.

Reducing the burden of inappropriate imaging requests is always a struggle – with a small, persistent minority choosing imaging modalities poorly matched to the clinical question being asked. In this cluster-randomised, differences-in-differences trial, the hospitals involved had baseline rates of inappropriate imaging requests of around 6-7%. After roll-out of this CDSS at the three intervention hospitals, the rates of inappropriate imaging requests dropped to 5-6% – but dropped “similarly” at both the intervention and non-intervention hospitals:

Interestingly, the authors also tracked the effect of the CDSS on whether imaging requests were changed from an initial order to a “more appropriate” option. Only 1% of imaging orders were changed – and, amusingly, more than half of those changed were actually changed to a “less appropriate” option.

Overall, it’s a lovely article to explore, and yet another example of the necessity of measuring any new decision-support for effectiveness and/or unintended consequences.

All hospitals discontinued use of this decision-support intervention following the trial.