Envisioning Migration from Conventional CAC (NLP) to AI: Part III

In my last article, I provided background for preparing your business cases to transition from a conventional computer-assisted coding (CAC)/natural language processing (NLP) environment to a technology platform to support coding and/or clinical documentation integrity (CDI) fueled by advanced artificial intelligence (AI), of which I defined new features and underlying technology requirements, along with the current state of features and advancements. This article will provide you and your organization with examples of where other organizations have delivered on the promise of AI to truly revolutionize CDI as we know it today. This will provide you and your stakeholders the opportunity to examine the early results of advanced AI to determine if it is worth the investment to migrate to, installing features on your current platform, or to select a new AI partner/system. 

Historically, solving CDI problems meant increasing people power – the theory being that more reviewers led to more compliant operations and consistent revenue. This reactive approach tackled immediate issues without addressing root causes, focusing on capturing complications and comorbidities (CCs) and major CCs (MCCs) to boost the case mix index (CMI) and reimbursement. Over time, automated tools improved efficiency, but remained largely transactional.

In many organizations, the objective is to achieve a 35-percent query rate. CDI specialists (CDISs) typically handle around 25 cases daily, aiming for a 90-percent physician response rate and an 85-percent agreement rate on queries. While these metrics gauge productivity, they do not identify areas needing improvement to accurately depict a patient’s acuity, and this leads to missed opportunities to capture revenue and assess risk accurately. This outdated approach to CDI results in inconsistent revenue and compliance performance.

This means that hospitals must transition from outdated tactical approaches to a modern, strategic approach. A strategic model is efficient, data-driven, and monitored with analytics. It is also supported by clinical audits and ensures consistent compliance and revenue integrity. By clearly defining problems and aligning projects with strategic objectives, hospitals can use resources efficiently and effectively to improve compliance, financial health, and overall patient care quality.

The challenges for CDI in healthcare are innumerable, associated with ensuring accurate, reliable, and compliant operations across the continuum. Denial management challenges also stem from poor documentation and inaccurate coding. Furthermore, there is often a lack of standardization and effective processes for managing appeals. The goal is to leverage technology to optimize CDI staff time, reduce burnout, ensure that our staff work at the top of their license, and reduce our reliance on contract backfill. Also, in non-acute-care settings, providers coding their own cases using non-standard templates and code capture can adversely impact the integrity of the documentation and code capture of the social determinants of health (SDoH) codes and other critical codes. In the acute-care setting, timely, complete, and specific documentation is critical to reflect the true patient population and complete code capture.

The CDI value proposition that advanced AI technology needs to encapsulate includes the entire continuum below to mitigate the above challenges, as we continue our transition from fee-for-service care to full capitation, as depicted in the diagram below:

CDI must be poised to transition to advanced AI to enable healthcare systems to instantly select, score, and prioritize high-impact cases for review to:

• Automate time-consuming case selection processes and provide guidance as to when new diagnosis-related groups (DRGs) or queries are warranted;
• Employ clinically driven algorithms to compliantly enhance DRG and revenue opportunities, increasing the return on investment (ROI);
• Enable results driven by analytics and benchmarking capabilities;
• Incorporate the latest healthcare regulations, coding guidelines, and payor contract rules; and
• Streamline workflows, allowing CDI, coding, and health information management (HIM) teams to concentrate on more crucial tasks.

The important point with the migration to advanced AI is that it augments, not replaces, the CDIS, enabling the CDI team to:

• Perform “at the top of their license” by letting them focus on clinical judgement instead of sifting through documentation;
• Find new revenue and capture meaningful codes without adding additional resources;
• Augment existing workflows – enabling a secondary review to complement concurrent CDI activities without ripping and replacing existing processes; and
• Facilitate a pre-bill review.

A CDI program’s financial impact relies on identifying revenue-increase and revenue-decrease opportunities, revealing vulnerabilities and educational needs. Aligning critical metrics to revenue goals is challenging, even with convention CAC/NLP tools, which measure the impact of queries. However, many queries are critical to the capture of value-based care and comparative quality system drivers, and do not carry a financial weight.

CMI is a popular metric, but it doesn’t fully address performance due to factors like surgery schedules, patient acuity, and external events. A refined approach should enable hospitals to benchmark their CMI against comparable institutions, adjusting for patient acuity and care complexity. This method provides a clearer picture of actual performance by highlighting discrepancies between clinical evidence and recorded data. For instance, while a standard CMI might show a decline, this technology could reveal that the decline is due to an improved observation rate, not poor documentation or coding. This is just one of the many functions that enable this technology to drive real financial impact.

Again, advanced AI systems enable the analysis of top-performing peers to track your program’s progress. Organizations can start with cohort benchmarking to identify areas for improvement. Ensure that metrics are comparable within the cohort, and communicate the benchmarking process across teams, particularly with physicians, to foster understanding and engagement. Ideally, the aim to reach the top quartile of one’s peer group by comparing the 50th and 75th percentiles is an achievable target with this technology. Organizations must be committed to regularly revisiting and adjusting their cohort to maintain relevance as performance evolves.

Advanced AI technology will also enable the analysis to uncover root causes of documentation discrepancies to further drive improvements. For instance, consistent physician query responses of “unable to determine” often indicate the need for targeted education on specific diagnosis codes or general query responses. Regularly reviewing these data points reveals common areas to achieve the greatest ROI.

Hospitals can measure their impact in two ways: 

1. What was the impact on the cases reviewed?
2. What was the impact on the overall program?

Program improvement should be communicated in terms of dollars, differentiating between revenue-increase opportunities and revenue-decrease opportunities, as they represent different vulnerabilities, risks, or areas that require education, coupled with observed versus expected values resulting from their comparative quality applications. Again, this requirement is a feature of advanced AI CDI systems, without a technology vendor tacking on an analytics charge, requiring a background team.

Specific questions about data help identify what to track for improvement, and this should be facilitated by the analytics feature. These include questions like “Where do we have changes?” “What diagnosis codes are we using?” “Was it a principal diagnosis?” “Was it a secondary diagnosis?” and “What’s the nature of the change?” The metrics will help uncover root causes and identify program process improvements for targeted education across teams.

Lastly, evaluate the effectiveness of the review process itself. Are the right cases being selected for review? Is there a good hit rate that justifies the continued performance of these reviews, ensuring that they pay for themselves and generate additional value? 

Advanced AI for CDI will enable the case review process to be more efficient, reducing the workload for CDI, HIM, and coding teams, incorporating CDI methodologies into proprietary algorithms based on comprehensive CDI insights, DRG methodology, and clinical knowledge. Also, the application should include encounter data characteristics, review outcomes, and changes in coding rules and guidelines. Finally, it should possess an extensive knowledge base, which allows the scoring engine to pinpoint smaller clinical populations accurately, identifying opportunities efficiently and flagging cases for expert reviews.

Although it’s difficult to demonstrate the technology in an article, a demo from available technology vendors should adequately demonstrate the easing of the documentation and review burden by both the provider and the CDIS, utilizing nudge technology, leveraging either direct documentation in the electronic medical record (EMR) or speech/ambient recognition/listening.  

Advances with documentation captured through ambient listening interoperability with the EMR and the CDI application is a game-changer. Also, many organizations have reported successful strategies with discrete field population utilizing these techniques, which will further amplify data analytics capabilities described above. 

Below is a depiction of the high-level interaction from the provider’s documentation, dictation (captured successfully now via speech recognition or ambient listening technology):

The next article (Part IV) will highlight use cases on the coding front to mitigate current technology and resource challenges, and the final article (Part V) will provide a blueprint for preparing your business case and setting up your implementation for success.

Programming note:

Listen to Cassi Birnbaum report this story live today during Talk Ten Tuesday, 10 Eastern, with Chuck Buck and Dr. Erica Remer.