Why AI Chart Review Isn't a Decisive Advance for RAF Capture
AI in chart review for RAF capture offers the obvious time savings improvement over manual chart review, but the questions industry analysts will ask are whether performance improvement goes along with the time improvement — the old maxim of “faster but poorer” comes to mind.
Why Chart Review Matters
Portions of the RAF score are calculated using HCC (Hierarchical Condition Category) codes. However, the HCC code does not have value unless it is evaluated and documented per a specific criteria standard. Documentation of a chronic condition must be evaluated and addressed at a minimum of once within 12 months and recorded in order to contribute to a patient’s risk score. Failure to document a condition substantiates the effective disappearance of the condition in the record and reimbursement for the condition.
The Limitations of Traditional Chart Review
The Overview
For manual chart review, a coder or reviewer must read the record and cross-check the record against prior year documented conditions to determine if the record is complete. It is an effective method, but, as already noted, it is painfully slow.
Where the AI Chart Review Process May Be a Game Changer
Transcending the Limitations of the Traditional Method
AI would likely be most impactful in reviewing prior years' records. Analysis of prior year records against the current year record for completeness could have an immediate impact on documentation of required evaluation and treatment of chronic conditions.
The Real Costs
Time per chart: Doing a thorough manual review of a chart can take 20-30 minutes, and doing this for each chart in a patient panel of thousands will take a long time.
Reviewer fatigue: Keeping review attention on repetitive documentation increases the chance of documentation being missed due to attention limits.
Inconsistency: Due to varying interpretation of ambiguous documentation, a single chart can be reviewed differently by different persons on a team, leading to differing capture rates.
How AI-Driven Chart Review Solves Similar Issues
Speed at Scale
AI can review the charts of a patient panel of potentially thousands of records in a fraction of the time needed by a manual review team. AI can find documentation of likely missing or incomplete HCCs across thousands of records at one time rather than one at a time.
Pattern Recognition Across Records
AI really excels at recognizing patterns that humans might miss. AI can be used to find chronic conditions by looking at medications, and can spot a condition that was diagnosed in the previous year, but has not been documented this year.
Where AI Needs a Human
AI-reviewed charts should be reviewed by a clinician. AI is not a substitute for a review of a patient by a clinician. The purpose of the chart review is to surface charts for review, not to make recommendations for codes.
Cost Comparison: What Actually Changes
While the technology itself isn't the largest expense, the shift in reviewer time is the largest cost driver. Instead of reading through completed charts that are ‘perfectly’ completed, clinical reviewers now have time to review the incomplete charts AI has flagged. While the per-chart cost of a full manual review doesn’t drop, the total number of charts that require a full manual review significantly drops.
Why Chart Review Matters So Much for RAF Accuracy
In regard to accuracy, AI can review more charts and review more consistently than a human reviewer; however, a human reviewer will still have the ability to capture more nuanced and context-specific clinical information. AI can also review a full panel, unlike medical reviewers who will typically only review a sampled subset of a panel due to time constraints.
The Realistic Combination
Most organizations that have noted positive RAF capture results do not choose AI over manual review. Rather, they pair the two. These organizations utilize AI to determine priority and order of reviews, and then perform a manual clinical review on the charts that are recommended. This is usually a value-based care software add-on, as opposed to a separate software, since the accuracy of RAF directly correlates to the software’s risk-adjustment quality metrics and value-based care outcomes. This combination has demonstrated an improvement over either used independently.
Conclusion
This isn’t really AI vs. manual chart review. This is more of a case of complete AI screening with some manual review vs. only manual review. For RAF capture specifically, which includes cost and accuracy tradeoffs for review of thousands of charts per year, this will most likely place a practice’s reimbursement accuracy in the spotlight.

Comments
Post a Comment