Henry Ford Health Taps Into AI to Improve Medical Coding

In a Q&A, Joann Ferguson, the health system's VP of revenue cycle, explains how the technology saves time and money and improves revenue cycle and clinical processes.

Among the many uses for AI technology in the healthcare space is in medical coding, which affects both clinical and revenue cycle processes.

Detroit-based Henry Ford Health recently expanded its collaboration with CodaMetrix to include patient bedside visits, where abstraction takes an average of 40 minutes per patient and accounts for 20% of the health system's overall coding costs.

"Inpatient hospital stays due to serious medical conditions, injuries, surgical procedures, and medical emergencies such as strokes, heart attacks, broken bones and burns, routinely require bedside physician consultation," the health system said in a press release announcing the CodaMetrix deal. "Evaluations and management of patients at the bedside, by hospitalists and other specialists, as well as bedside procedures, need to be abstracted into medical codes for reimbursement. Depending on health systems’ policies, the coding function is usually performed by physicians, medical coders, or both. In scenarios where physicians are responsible for coding, not only is it an extra burden, but it increases the number of missed opportunities for accurate reimbursement."

To learn how AI can be integrated into the bedside procedures, HealthLeaders spoke virtually with Joann Ferguson, RN, BSN, MBA, CRCR, vice president of clinical revenue cycle at Henry Ford Health.

Q. How does Henry Ford use AI to improve the coding process?

Ferguson: As one of the nation’s premier academic and integrated health systems, Health Ford Health has more than 110,000 inpatient visits per year across five hospitals and 2,300 staffed beds. We hold ourselves to a high standard not just in our care, but operationally as well. We’re always looking for the best ways to support our teams, from our doctors and nurses to our coding and billing departments in their everyday workflows.

Joann Ferguson, RN, BSN, MBA, CRCR, vice president of revenue cycle, Henry Ford Health. Photo courtesy Henry Ford Health.

We began exploring new technological options across our revenue cycle operations because we were dealing with many of the same issues that are putting pressure on healthcare providers and employers across the country, including attrition via retirement and difficulty filling open positions. We needed to improve efficiency in our workflows, resulting in lower costs, reduced backlogs, and enhance patient and provider experiences.

After an extensive review of our internal work streams and technologies we decided to pursue implementing an AI coding solution for our bedside professional services. With more than 700,000 inpatient bedside services performed each year, it’s one of our highest volume specialties. We needed an alternative solution to keep up with rising volumes and to reduce backlogs.

AI improves our bedside medical coding process in several ways. First, it automatically codes the simplest procedures, taking that work off our coders’ plates. By 'simplest,' we mean the procedure notes that match closely or exactly with how the ICD codes themselves are written. It does this by bringing together all the complex information required to identify, understand, and code a bedside professional charge. It then predicts and assigns charges and diagnosis codes, automating cases directly to billing.

For bedside procedures where the AI platform does not reach our confidence level threshold, AI gives our coders an optimized view of the information required to code a case and pre-populates code suggestions for non-automated cases. The coder can then validate and edit from a selection of probable codes rather than start from scratch.

Additionally, before a coder releases the case for further processing, the case is checked against standard edits. This means the original coder resolves the edits rather than them being sent along to a standalone edit team, streamlining the process.

Q. How did the health system approach this process prior to using AI?

Ferguson: We had built a custom access database, which we used to aggregate coding information and look for charge gaps. Unfortunately, it was cumbersome to use and almost impossible to scale and maintain.

Q. What are the benefits to using AI in coding? What specific improvements are you seeing?

Ferguson: Inpatient bedside visit coding accounts for 20% of our overall coding costs. By implementing AI, we will increase workflow efficiency by reducing errors, missed charges, billing backlogs, and claim denials while lowering costs.

The platform also creates a nuanced understanding of our patient journey and can identify potential charge gaps where services were likely provided but there is no documentation. Once identified and routed to coders for follow up with providers, these estimated charge gaps can equal as much as 8% of overall bedside revenue that was previously left unbilled.

Workforce challenges are addressed, too. Because staffing is at a premium, by automating our bedside visit coding, we can shift resources to other areas of need. Regarding the big picture on the people side of Henry Ford Health, it reduces the daily workloads on physicians, medical coders, and billing administrators, driving better financial and operational performance while improving our coders’ job satisfaction.

Finally, Al improves the patient experience by reducing denials.

Q. What are the concerns or challenges to using this technology?

Ferguson: As with onboarding any new technology, the biggest challenges we face are overcoming staff nervousness about learning and using a new system, training staff to then use that system correctly, and ensuring the AI is coordinated with our other systems. However, we find we can get around some of the roadblocks and hesitation that come with using new technology by taking time to highlight the short-term and long-term benefits to employees’ everyday workloads, while also laying out how it helps the organization as a whole. When people see the benefits on both ends of the spectrum, we’ve found they’re very willing to make the leap to AI.

We also build trust with our coders by using a 'glass-box' AI partner. That means our team can see the evidence behind every code and every case, so we are not asking them to blindly trust the AI’s recommendations.

Q. How do you ensure accuracy and reliability with this technology?

Ferguson: The AI system we use through CodaMetrix is being built to learn and adapt over time based on the feedback provided by our medical coding teams, so it’s constantly improving. That’s the power of machine learning, which keeps the system from becoming brittle when new ICD and CPT codes are released throughout the year. 

We are in constant contact with CodaMetrix in every step of the build process to ensure we have a successful launch of the technology. We set our own quality standards for coding accuracy, giving us an additional layer of control of the AI. Through our partnership we are both committed to quality on Day 1 of implementation, having immediate access to prediction and automation information. This will keep Henry Ford Health’s revenue cycle running smoothly, while improving how we operate the system.

Q. What has surprised you, good or bad, about this technology or the outcomes you're seeing?

Ferguson: We like the level of control we have. We set the quality thresholds, which means we can use the coding AI to our standards rather than those set by someone outside of the organization. We also look forward to the transparency with which the platform operates. We’re able to see 'under the hood' at all times, so our medical coding team does not have to guess why CodaMetrix chose a particular code for a specific case. This helps build our team’s confidence in the AI solution, which we anticipate will speed ramp-up.

Q. How do you see this technology evolving? How and where else would you like to use it?

Ferguson: AI has been infiltrating the healthcare industry for years now, but recently it’s seemed to hit a critical mass. It’s already working its way into doctors’ notes and diagnoses via new products from Google and Microsoft, and it doesn’t seem like it will be long before AI is providing meaningful assistance to doctors making complicated decisions about the best way to treat their patients. It’s amazing to see everything unfold in real time and be at the center of it.

Regarding coding, as the technology matures and becomes more adaptable, I would like it to spread into new specialties and departments. Each hospital specialty has its own medical coding team, so the first step would be using AI across the entire hospital and patient billing departments. The same goes for billing and doctors’ notes, which all layer into a well-run revenue cycle. To have AI that makes revenue cycle management easier across the board via accurate automation is a big win.

Q. What advice would you give to another health system considering using this technology? What, in your opinion, would they be most likely to do wrong?

Ferguson: When it comes to using AI in medical coding, make sure to do your due diligence. Going for the quick fix or using last year’s technology because it’s cheaper will only make change more painful in the future. That means finding a partner who understands rev cycle operations and AI, and what your team needs to be successful.

AI products that will grow and can keep pace with the breakneck speed of tech innovation in healthcare are a must, not a 'nice to have.' Check out what the best and most innovative hospital systems are using; they’re usually at the vanguard of the industry and often choose the best practices.