Trust, security, and impact: Transparent AI in healthcare

From accelerating quality improvement to increasing precision in risk adjustment, using artificial intelligence (AI) in healthcare holds a multitude of benefits for value-based care organizations. Yet without transparency and explainability, AI-powered technology falls short of its promise to deliver cost savings, increased efficiency, and better outcomes for patients while keeping organizations compliant with changing regulations.

The promise of AI in healthcare

AI-powered technology has already demonstrated its potential to drastically increase efficiency, make sense of large volumes of data, and generate highly accurate insights. Its potential is only increasing as the technology advances, data sets expand, and models continue to adapt.  

For value-based care organizations, AI can accelerate and increase the precision of various processes such as:

• Extracting data from unstructured clinical notes and medical records

• Uncovering hidden risk factors  

• Standardizing diverse data sources  

• Extracting clinical evidence

• Providing clinicians with insights and suggestions at the point of care

• Boosting risk adjustment accuracy  

• Optimizing quality improvement metrics  

When AI falls short: lack of transparency

Despite the promise of AI to enhance efficiency, accuracy, and compliance in healthcare, without a high level of transparency, AI-powered technology falls short of its promise. Explainable AI—or AI-powered technology that seeks to make its processes as transparent as possible—is an increasingly large part of the total AI technology market. Predictions suggest it will grow at a rate of over 20% in coming years as users demand ever more control, insight, and accountability for AI technology.

Forty-one percent of physicians are both excited and concerned about using AI in clinical care. Many are skeptical of its ability to deliver highly accurate insights, and even more so when those insights are not backed by evidence. To make the greatest impact, AI-powered technology should assist healthcare providers, medical coders, and other staff by delivering insights that are trackable back to source documentation. This level of transparency in AI is critical to ensuring that value-based care organizations make the best use of the technology. It also ensures users don’t have to sacrifice the trust and integrity necessary to feel confident in their decisions and the basis for those decisions.

What to look for in transparent AI technology

When AI systems are truly transparent, they present a clear view of the logic and decision-making processes used. They also offer ways to address any errors or biases occurring within the systems.

A high level of accuracy

Transparent AI technology generates the highest level of accuracy when it comes to its insights and predictions. More opaque AI models carry a far greater risk of errors. When AI is transparent, however, users can help to identify biases, flag errors, and correct faulty reasoning. These interactions help to quickly refine machine learning algorithms, making them more precise and more relevant for future users.

Links to supporting evidence

The key to transparency in AI is the ability of human coders to review the evidence used in the AI system’s decision-making process. AI-powered technology should have clear documentation on how it works, including types of data used for initial training, data science techniques, and how certain data types and attributes are weighted.  

For any prediction, users must also be able to trace back to the source evidence quickly and easily. For example, if AI-powered technology suggests an HCC for diabetes to a risk adjustment coding team, the team should be able to easily access the evidence for that suggestion in source documentation. ‍

Opportunities for expert input

Users working alongside AI should be able to go a step beyond accessing supporting evidence—they should be able to quickly identify incidents of bias. In this way, AI technology serves as a true assistant to healthcare experts, offering suggestions, backing them up with evidence, and ultimately deferring to the expertise of human staff. In every case, users should have the final decision-making authority, approving or rejecting AI-generated predictions.

Highly transparent AI technology used in diagnosis recommendations can go so far as to predict the likelihood of coder diagnosis validation based on similar activities and patterns from previous clinical reviews. Using confidence scores, it can consider the likelihood of clinician or coder agreement when it raises a suspected diagnosis. This goes beyond simple diagnosis prediction and moves towards emulating the user over time to improve the accuracy and impact of its predictions.

Rigorous testing and security protocols

Finally, any technology incorporating AI—perhaps especially in healthcare—should adhere to strict security and compliance protocols. These protocols should be clearly documented so that healthcare organizations can rest assured they are remaining HIPAA-compliant and protecting patient data with the highest level of security.

Audits and peer reviews of AI systems can support these security protocols and help improve the accuracy of the technology’s predictions. AI-powered technology systems should conduct rigorous testing before deployment as well as continuous monitoring afterward, demonstrating the results to its users.

Consider EVE™: Reveleer’s reliable (and transparent) AI assistant

Supporting Reveleer’s value-based care enablement platform is our Evidence Validation Engine, EVE. EVE harnesses natural language processing for data ingestion and machine learning for clinical review, helping healthcare organizations:

• Enhance data capture and analysis

• Streamline clinical workflows

• Excel in value-based care

EVE is incredibly effective. She can generate clinical insights with up to 99% accuracy and reduce coding duration by 42.5%. However, EVE does more than provide actionable insights for risk adjustment and accelerate critical clinical review processes with intelligent automation. She is transparent, auditable, and secure.  

Not only does the Reveleer Platform adhere to strict data privacy regulations and use advanced security measures to protect sensitive patient information, but it also prioritizes explainable AI. Users have easy access to clear insights into how its models arrive at their predictions, enabling human oversight of AI decision-making and improving the accuracy of AI models. This high level of transparency also builds trust with users, increasing the usability and impact of the platform.

Interested in learning more about transparent AI in healthcare that can power your value-based care goals? Schedule a demo of the Reveleer platform today.