Natalie McGarry, Imaging Performance Partnership
In a recent study, researchers found an extremely low rate of error in clinical decision-support input with over 90% of orders entered correctly. Of the input errors, only 4.2% could be construed as physicians deliberately entering incorrect information to “game” the system.
Isolating a CDS gatekeeper
The study, published in the Journal of the American Medical Informatics Association, evaluated emergency department (ED) clinicians’ use of clinical decision-support (CDS) software when ordering CT angiography (CTA) exams for pulmonary embolism (PE).
Researchers chose CTA in order to isolate the input of D-dimer results, which acts as a cut off point for the CTA imaging request for PE. First, ED clinicians calculate the patient’s Wells score, which is used to determine the risk of PE.
If the patient has a high risk of PE, the CDS allows the CTA request, if the patient has a low pretest risk of PE, the system requests the clinician enter a D-dimer value. An elevated D-dimer value enables the imaging order, a normal D-dimer value in a low PE risk patient recommends not obtaining a CTA.
The clinical pathway is as follows, where the D-dimer value is only necessary if the patient has a low risk of PE, and therefore acts as the gate keeper for the CTA request:
Few errors, few docs gaming the system
In 2011, of the 1,296 patients who received CTA studies, 1,175 or roughly 90.7% of patients had accurate D-dimer values recorded by clinicians. The authors noted that on the basis of their findings in addition to a small number of studies, there is likely a ceiling of 90-%95% accuracy in clinician data-entry.
Of the 121 cases with data entry errors, 55 were entered incorrectly in such a way as to avoid CDS alerts. These 55 cases represent 45% of all data-entry errors, but only 4.2% of the total cases. In 15 of these cases, clinicians indicated an elevated D-dimer value when it was normal, while the other 40 involved clinicians claiming that no D-dimer result was available when the laboratory reported the result as normal.
Clinicians eventually canceled 21 of the 55 inappropriately ordered imaging requests, while the remaining 34 completed scans found no PE.
The researchers were surprised by how low the deliberate error rate was that they questioned the benefit of investigating clinical motivation for error entry.
They also noted that adherence to evidence-based guidelines improved by 75% with CDS and that investigating how to further improve this number would inform further research. One suggestion was auto-populating data entry with clinical data from the electronic medical record.
Learn More About CDS at the National Meeting
To learn more about CDS, attend the Partnership’s national meeting, On the Brink of Accountability, for helpful case studies and a study on preparing for risk-based payment.
Want more from this author? See all of Natalie's blog posts.