How Confidence Scoring Can Improve Your Risk Adjustment ROI

Confidence scoring is an AI-driven approach to hierarchical condition category chart review that helps payers prioritize which records are most likely to yield new or updated codes, reducing unnecessary manual review while improving risk adjustment accuracy.

Claims costs are rising for payers. The aging population alone is expected to contribute an additional 0.5% to 1.0% increase in claims costs by 2029. At the same time, enrollment declines in Medicaid and Affordable Care Act (ACA) plans, along with persistent margin pressures, are creating added financial strain.

Weathering these challenges and positioning for the future necessitates that payer organizations rethink traditional operating models, improve performance, and embrace technology, according to McKinsey. In a recent industry analysis, the firm argues that the key to competitiveness will be AI-enabled back-end transformation.

One operational area especially ripe for transformation is hierarchical condition category (HCC) chart reviews. Risk adjustment and coding workflows have become a priority focus as organizations seek new ways to improve accuracy, reduce administrative burden, and protect margins. Conventional chart review systems continue to rely heavily on manual processes, which are time-consuming and prone to errors.

According to one recent KFF study, nearly 30% of manually reviewed charts contain new or unsubstantiated HCCs.

Where Most NLP Tools Fall Short

As payers seek to improve efficiency and scalability, automation is becoming increasingly necessary. Yet relying on rule-based or narrowly trained systems can create new limitations.

That’s because natural language processing (NLP) engines typically focus on missing HCC codes rather than on documentation quality, nuanced clinical context, and coding accuracy. For example, an automated review may identify diabetes but fail to distinguish between uncomplicated diabetes and diabetes with chronic complications, which can significantly affect the accuracy of adjustment.

NLP tools can also miss important nuance and context. In some cases, the automated review may identify clinical findings that support a diagnosis while overlooking other documentation that contradicts or weakens the clinical evidence.

In addition, high-precision systems trained on a limited set of predefined “covered” coding scenarios can also introduce blind spots by extracting diagnoses that the provider did not clinically substantiate. An unsupported condition can create both exposure and the potential for significant financial impact, including clawbacks.

Another challenge is that incomplete or poorly tuned NLP engines can produce significant “false positives,” leading coders to override or ignore technology recommendations over time. Rules-based engines are also not as nimble as more advanced agentic AI offerings, and creating rules requires extra time that results in delayed revenue generation.

To move beyond the limitations of traditional NLP automation, payers need an approach to chart review that can evaluate clinical context more comprehensively rather than simply identifying keywords or suspected diagnoses.

Confidence Scoring: A More Targeted Approach to Risk Adjustment Review

Rather than replacing human review entirely, many payer organizations are beginning to recognize the value of a hybrid approach, one that combines AI-driven prioritization with clinically informed oversight. This model can help organizations improve efficiency while maintaining coding accuracy and compliance integrity.

A hybrid approach guided by AI for “confidence scoring” is ideal, as it combines the speed of automated review with the expertise of human review in a more targeted way.

With the right technology, coders can gain insights into a record’s potential before they even open a file, helping them decide where to spend their time and resources.

Reducing workload with high accuracy. ECLAT’s proprietary Confidence Score uses agentic AI to analyze each record like a coder, using clinical documentation review to assess which records are the best candidates to review or skip based on the likelihood of new codes or information. Having this confidence in where to focus helps users reduce first-pass chart reviews by up to 70% without sacrificing performance.

Better decision-making. The quality of the review process also improves. The technology identifies new HCC codes, flags records without codes, and validates existing diagnoses against reported HCCs. Payer organizations can use these insights to guide better decision-making and help coders strategically prioritize what records they review, and when. Staff remain the key drivers and decision-makers, manually validating adds and deletes throughout the review, as desired, thereby enhancing the accuracy of risk adjustment factor (RAF) scores and reducing audit risk.

With confidence scoring and human-in-the-loop validation, payers can better prioritize and move through chart review queues more effectively. ECLAT users see more than 2x ROI compared with other code review offerings, driven by efficiency gains from fewer unnecessary first- and second-pass reviews and higher code capture from increased RAF scores.

Moving Toward More Intelligent Review Workflows

As payer organizations modernize risk-adjustment operations, success will depend not only on adopting AI, but on implementing processes that improve accuracy, transparency, and clinical integrity. Organizations that combine intelligent automation with experienced human oversight will be better positioned to create scalable, compliant workflows while focusing resources where they can have the greatest impact.

See how ECLAT’s AI-enabled chart review approach can help your team reduce workload with greater confidence. Request a demo.