Use of AI in coding and risk adjustment: 4 key recommendations

Tools compatible with artificial intelligence (AI) such as natural language treatment (NLP) have been integrated into a wide range of applications, including risk adjustment coding tools, for greater efficiency and greater precision in the health care industry. For Medicare Advantage (MA) plans, these tools can considerably improve the accuracy of diagnostic and category of hierarchical conditions (CHC) necessary to support risk adjustment programs and help guarantee appropriate reimbursement.

Prepare new RADV changes

With NLP tools, MA plans can discover errors during retrospective reviews of graphics before a risk validation audit (RADV). Once required only for around 10% of MA plans each year, the RADV audits will now affect all MA plans, because the centers for Medicare & Medicaid Services (CMS) make efforts to reduce overpaye.

As part of its aggressive strategy, CMS will also have a greater number of records – up to 200 recordings per plan. The change of policy highlights the need for both precision and efficiency for the MA plans.

The expansion of audit RADV follows other important changes in policy, which now allow CMS to extrapolate its audit results from the sample of medical files examined with the entire plan contract – potentially putting a single contract in danger for millions if the agency decides that the files do not adequately support the diagnostics of registrants. The elimination of the Act Compensation Expert (FFS) also increases the burden of plans to guarantee precise and complete HCC reports or risky extrapolated sanctions.

How can the AI help the plans my

For MA plans which have not undergone RADV verification, these changes offer an appropriate opportunity to integrate AI into their coding practices and establish policies and procedures appropriate to technology.

By incorporating compatible tools AI into their workflows, MA plans can prioritize critical documentation and ensure that their coding teams focus on the most relevant fields of long and complex medical records. For example, these tools can easily identify current errors such as HCC reported in the wrong parameter (hospital against outpatient) or by bad specialty. NLP compatible tools can also help coders quickly find cases in which medical records recovered for two different members have been accidentally merged, which creates inaccuracies for retrospective graphics reviews or the processes of submission of RADV graphics.

Strategies to deploy compatible AI tools

Here are the best practices for plans to consider when they implement compatible tools AI to improve the accuracy of their coding and risk adjustment programs.

• Launch an AI governance committee for human surveillance. The plans should establish a framework for verifying and supervising new uses of AI or NLP in their organizations. By creating a governance committee of clinical, technical and coding experts, the plans can examine different use cases for AI and have a forum to raise concerns concerning potential inappropriate uses. To guide organizations in health care and other industries, the AI Institute of Manager offers best practices for AI governance structures, as well as the principles of examining AI projects. Membership of the advice of industry defense groups can help managers ensure the ethical implementation of AI in coding and other areas.

• Create a “sandbox” environment so that the coders test the tool. Providing coders with test documents so that they can experience the tool can help them practice the workflows they will experience in real life. Plans can also provide a user control list to help coders simulate various scenarios and record any problems related to performance or conviviality.

• Publish a dashboard with measurements to measure performance in a global manner. Managers must maintain continuous commitment to assess the performance of AI compatible tools. Plans should visualize their performance globally and follow individual and precision productivity and precision measures. Potential red flags are coders who, compared to peers, are exceptionally slow or fast when using AI tools. Plans should also look for panels suggesting an excessive exception on AI, such as a coder who accepts the suggestions generated by AI almost 100% of the time. The specific benchmarks defined by the plans should depend on factors such as their sector of activity, the type of software used and if the data is extracted from electronic medical files (DME) or digitized PDF files. Plans should examine their measures at least monthly to identify improvement opportunities and share results with the main stakeholders.

• Take advantage of the end user comments for continuous improvement. The request of coders comments is essential to ensure a positive user experience. Sometimes coding tools that generate excess recommendations for coders can slow them down, hinder productivity and create frustration. Coding “superutilizers” submitted suggestions to managers and leadership can help you constantly refine technology and procedures.

• Align performance expectations with suppliers. If the plans take advantage of compatible software AI via a coding partner, they should have performance guarantees related to system performance, availability / stop measures and PNL accuracy with deadlines and potential penalties for delays. This can help protect plans from system failures and other problems that could potentially have their project deliverables derail and report deadlines.

Prepare new CMS audit efforts for the plans of my

While CMS increases its RADV initiatives in the coming months, the plans should guarantee that their risk adjustment programs meet the highest precision and compliance standards. Prospective and retrospective analysis improved by AI can help plans to work with suppliers to optimize the documentation at the care point and identify coding errors during the preparation of the audit. Plans may also want to consider carrying out a second level examination of the coding results, which allows them to correct unpaid CHCs which could easily be neglected during the first level exam. By combining tools compatible with AI with monitoring of experts, plans can improve the success of these efforts because they encounter greater regulatory monitoring in the future

Photo: Thanakorn LappattaNan, Getty Images

This message appears through the Medcity influencers program. Anyone can publish their point of view on business and innovation in health care on Medcity News through Medcity influencers. Click here to find out how.