We partnered with a large primary care practice (circa 22,000 registered patients) with the goal of finding at least an additional $15,000 of pay-for-performance income using CodePilot. The practice is well run, with a professional administrative team, and best-available technology for data management and clinical coding. They recently completed a project with an external team of consultants to ensure that certain key registriess, such as those for dementia and diabetes, are up-to date and not missing any patients. Confident that they run a tight ship but open-minded to new improvements, the practice partners were very interested to see what additional income our AI could find.
To minimise disruption to the practice, in the first instance we supplied a set of EMR queries to generate completely anonymous patient data for batch analysis. In the future, however, we are providing a full EMR integration to allow deeper analysis, and for coding errors and missed income to be caught in near real time. We have identified over fifty potential areas of investigation, but began with a subset of twelve as a starting point. The initial focus was to identify patients diagnosed with certain conditions or prescribed medications used almost exclusively to treat certain conditions, who are missing key clinical codes that trigger their inclusion on that condition’s registry. Correcting these coding oversights represents a lift in pay-for-performance income and additional reimbursements for items such as unclaimed flu vaccines. The results were classified as either ‘near-certain’ or ‘probable’ depending on the confidence with which we believe each record will require a coding change subject to clinical review.
Our analysis of the first twelve areas of investigation identified over $30,000 of additional income for the practice. A breakdown of these findings is shown in the table below. We found 676 records with coding changes required, including over 100 patients missing from the dementia registry recently reviewed by external consultants. This analysis shows that potential errors and oversights can linger in patient records even after they have been subject to specialist review, and highlights the benefit of having an automated system to double check coding and perform ongoing data quality control.
We are now working with the practice to enable clinical review of the project findings, such that the coding changes can be made and the potential earnings realised. We are also working to expand the analysis to include all 50 areas of investigation, complete a full EMR integration, and roll-out CodePilot to the other three practices in their primary care network. The partners at the practice expressed their satisfaction with the results of the project:
