Wednesday, September 23, 2020

Using Artificial Intelligence to Help the Clinician Make the Right Decision

Physicians and Advanced Practitioners work hard to make the correct diagnosis and help each patient. They use their knowledge and experience to decide what information to gather, but we are all human and imperfect. No one can know everything, and everyone has biases in their reasoning. 

There are many studies showing that clinicians often don’t order the best imaging studies given the patient’s signs and symptoms. Sophisticated imaging studies like CT scans and MRI’s are priced fairly high because of the many associated expenses, and therefore the total cost to payers is enormous. When an inappropriate test is ordered, it is a prime example of waste. 

The Protecting Access to Medicare Act (PAMA) of 2014 established a program to increase the rate of appropriate advanced diagnostic imaging studies provided to Medicare beneficiaries. Examples of advanced imaging studies include:

Computed tomography (CT)
Positron emission tomography (PET)
Nuclear medicine, and
Magnetic resonance imaging (MRI).

The legislation requires that whenever a practitioner orders an advanced imaging study, he or she must use “a qualified Clinical Decision Support Mechanism (CDSM) with appropriate use criteria (AUC).” The CDSM is an interactive electronic tool that provides a determination of whether the order adheres to the AUC. It takes the input from the clinician and provides analysis and feedback to the ordering provider about whether the test is appropriate. This requirement is currently in the testing and education phase and will be fully implemented in January 2022.

At GBMC, we have used our core competency of redesigning care and made numerous enhancements to our CDSM/AUC workflow. By analyzing our data, we identified that the use of “free text” indications accounted for most of the non-compliant/indeterminate orders. We learned from discussion with providers that the “free text” field is used when a patient's condition did not match the available displayed diagnosis. We also learned that it is used to provide additional clinically-relevant information to the radiologist reading the study.

By working with the vendor of our CDSM software, we deployed their artificial intelligence tool to streamline the ordering workflow and to move us closer to our aim of increasing the rate of appropriate advanced imaging studies. The new artificial intelligence tool reads the “free text” field and presents discrete diagnoses to the provider so that he or she can pick one during the process of ordering advanced imaging tests. The pilot was done with providers in the Emergency Department because they had the highest utilization of the “free text” field. 

The chart below shows the “special cause” improvement in reducing the number of indeterminate studies in the pilot. This test of the use of level 2 mistake proofing  (a reminder in the moment or an affordance) shows that the use of artificial intelligence significantly improved ordering. 

My hat goes off to my colleagues in Radiology and in the Emergency Department for this great use of the Model for Improvement in moving us closer to our vision! 

Wash Your Hands
This week is International Clean Hands Week. It’s a good time to remind ourselves of the importance of good hand cleaning habits. We all know that proper hand hygiene is one of the most effective actions to reduce the spread of pathogens and prevent infections, including the COVID-19 virus. So please, wash your hands. Thank you!

No comments:

Post a Comment

Thank you for taking time to read "A Healthy Dialogue" and for commenting on the blog. Comments are an important part of the public dialogue and help facilitate conversation. All comments are reviewed before posting to ensure posts are not off-topic, do not violate patient confidentiality, and are civil. Differing opinions are welcome as long as the tone is respectful.