MPOG CPT Prediction Tool

A Machine Learning tool that predicts anesthesia CPT codes when billing data is

missing or delayed

2

What is the CPT Prediction Tool?

• A Machine Learning tool developed to predict the Anesthesia CPT code when one

is not available

• Even though this tool can be used, MPOG still requires billing data to be submitted when it is

available. This will help the tool ‘learn’ from the codes and improve predictions.

− Errors can also occur within the file (i.e., file name format)

o Missing billing files at the local site level

3

4

• How the algorithm selects from the top 3 predicted codes

• Rule 1: If the predicted confidence interval differs by more than 1.6 times between the 1st

and 2nd code, use the 1st code only

• For example: 01922 at 70 percent, 00700 at 20 percent, 01936 at 10 percent - use 01922 only

• Rule 2: If the 1st and 3rd codes differ by more than 1.6 times, use the 1st and 2nd codes only

• Rule 3: For labor epidurals and c-sections, use the 1st predicted code only

• What to expect: CPT prediction is better than no codes, but it is not perfect

• Site feedback lets us manually improve the algorithm over time

5

• The CPT Prediction Tool fills the gap when anesthesia billing codes are missing or

• Predicted codes feed MPOG Measures and Phenotypes, then refresh once actual

support@mpog.zendesk.com