Recent reports of payers using AI algorithms to automatically deny claims raise the stakes for health care finance executives.
My health care career began in the 1980s when I was a clinical coder for a small Pittsburgh hospital. Even then, before the Prospective Payment System and DRGs, a steady drumbeat of payer denials echoed throughout the business office, HIM, and finance departments of most health care provider organizations.
Today, the drumbeat of payer denials has become a deafening roar. And the noise continues to intensify as payers implement new artificial intelligence (AI), robotic process automation (RPA), and machine learning (ML) algorithms to automatically tag cases for denial, audit, or revenue recoupment.
While some use of automation by payers is beneficial to providers and patients, such as expediting prior authorizations for commonly approved treatments, some is not. For example, UnitedHealth Group was recently in the news regarding its acquisition of a popular Medicare Advantage algorithm that predicts the amount of care needed by seriously ill patients and triggers denials for continued coverage. Cigna Health is allegedly using an algorithm to automatically deny claims in bulk, according to a recent complaint filed in the Eastern District of California. And a UnitedHealth Group subsidiary also faces a lawsuit filed by the US Labor Department for not “applying a prudent layperson standard” and incorrectly denying emergency department and urinary drug screening claims for thousands, according to the complaint.
The latest CMS transparency data on claims denials adds another layer of confusion, with 77% of health plans classifying the reason for denials as “all other reasons,” as noted in a recent Kaiser Family Foundation article.
With these types of technology-driven denials on the upswing, it’s even more urgent for hospitals and health systems to become more assertive and proactive with denial defenses. Every available tool, best practice, and workflow must be deployed to retain provider revenue and protect margins.
For example, by using AI, data mining, analysis, and automated decision engines, revenue cycle teams can streamline many processes and witness a major impact on reducing denials in real time. Intelligent systems work by exception to assess each claim and calculate its likelihood of being denied. Predictive algorithms also evaluate the probability of a denial being overturned.
Revenue cycle teams are in a much better position to fight back against a new era of payer denials and focus on high-value work when equipped with better intelligence. Here are eight specific defense tactics to consider.
TACTIC #1: Ace Your Prior Authorization Requests
The Office of Inspector General recently published an audit of seven MCO parent companies. It found prior authorization denial rates ranging from 2% to 41%. The Office of Inspector General recommends targeted state oversight of prior authorization denials. But until states create new oversight programs, what can provider organizations do to proactively prevent these denials?
According to Jason Coffin, CHFP, CRCE, revenue cycle healthcare leader at TruBridge, a CPSI company, “Many facilities don’t track prior authorization denials, and prevention must begin with better insights.”
Upcoming rules require payers to electronically receive and transmit back prior authorization information starting in 2024. This change will “increase transparency and initiate automated feedback loops back to the provider,” Coffin says. Until then, he offers three tips to improve the prior authorization process:
- Craft unique authorization requests based on knowledge of each payer’s coverage rules for treatments and procedures for which authorizations are most often requested.
- Use automation to electronically submit and process requests to decrease reliance on human processing. For example, ML can digitally prepare and send the prior authorization request based on unique payer-specific parameters defined by the provider organization while also electronically reading the payer response.
- Analyze data produced by the technology to define criteria where precertification may be needed and teach bots to automatically route these cases to patient access specialists for human intervention and expertise.
TACTIC #2: Predict Reimbursement Up Front and Escalate Risky Cases
Existing collection practices are limited in the ability to predict reimbursement accurately and proactively. The new primary objective for revenue managers should be to ensure expected reimbursements are correctly calculated up front and identify any potential trends for underpayment to group claims together farther along in the billing process.
According to Jose Loza, divisional vice president of revenue cycle management partner transformation at Acclara, “Knowing which cases have potential for underpayment reduces the amount of downstream work effort to recover revenue.” Loza recommends that organizations establish a payer escalation program as part of the revenue cycle continuum to reduce aging and improve claims throughput. In partnership with Matthew Thomas, founding partner of RemedyIQ, Loza suggests the following five steps for an effective payer escalation program:
- Launch an expected reimbursement integrity team and equip them with automation.
- Create algorithms to limit the number of human touches while empowering staff to identify potential trends or issues.
- Write an escalation pathway that includes pursuit levels, best practices, and technology workflows to standardize payer escalation.
- Build a payer escalation team focused on scrubbing data and grouping similar issues for escalation to the payers.
- Establish a legal referral program and standing agreement with a local attorney.
The long-term goal is to create a subset of predictable payer behaviors and automate as much as possible, allowing teams to preemptively respond without human intervention. This strategy extends to denial tracking and appeals.
TACTIC #3: Use Data to Get Smarter About Denials and When to Appeal
With denial volumes on the rise and reasons for denials often unclear, it’s imperative for health systems to spend their time and resources wisely. Some payers seem to deny just about everything, anticipating that providers don’t have adequate bandwidth to appeal. Other payers are often correct in denying a claim, and their case is justified. The difficulty for providers lies in knowing the difference.
Analyzing when a denial is appropriate vs when it should be appealed is an ideal role for automation. For example, Kristy Evans, senior vice president of coding and clinical documentation integrity at e4health, agreed with payers 83% of the time during her former tenure as coding manager and quality coding specialist. “It was knowing when to fight for the other 17% that was key to effective use of our time and resources.”
Evans suggests organizations begin with centralized tracking and trending of all denials. Once centralized, ML should be applied to identify when payer denials are “most correct.” These cases can be written off, while those falling outside of the algorithm are electronically routed to denial and appeal specialists.
Reconciling the claim to the expected payment amount and to the estimation of benefits also gives staff greater insights into denials and payer performance. “This is especially important when flagging which cases to pursue or not to pursue for appeal,” adds Bill Knox, vice president of product management at FinThrive. “Advanced denial mapping intelligence is the first step to smarter appeals.”
Denial mapping engines should go beyond payer reason and remark codes to evaluate all data within the estimation of benefits, properly identify each scenario, and be used to streamline the triage of cases. This includes a false-denial indicator for cases where the payer uses an incorrect code and improperly triggers a denial.
“Claims data tells a story about the patient’s severity of illness and treatment and also provides valuable insight into payer behavior when a claim is rejected,” says Susan Gatehouse, vice president of revenue integrity at AQuity Solutions. Gatehouse encourages providers to fully understand each payer’s behavior. Knowing payer behavior is the impetus to proactively quell the increasing volume of denials before a claim is released for billing. “Take a closer look at payers with a high denial rate, verify the validity of the denials, and identify trends.”
Gatehouse emphasizes that centralized denial data helps organizations arrange the puzzle pieces and determine which payers have the highest volume of denials, reasons for denials, and appeal overturn rates. Knowing which specific MS-DRGs carry a higher rate of denials also supports educational efforts with physicians, clinical documentation integrity professionals, and coders. Managed care departments should also be involved in the denial analysis process as insights become useful at the time of contract assessment or renewal.
ML also plays a role in improving denial intelligence. ML can be used to predict the best course of action to resolve the issues at hand, expediting cash collections. According to Carrie Bauman, vice president of marketing at WhiteSpace Health, “When these ML-generated recommendations are followed, process consistency is created, revenue is enhanced, and denials (and other areas of revenue leakage) are resolved faster. Altogether, this creates a positive impact on cash collections.” And, as the data model is expanded over time, the intelligence generated by ML gets even better. Here is a brief overview of the steps involved:
- ML technology finds complex patterns in historical data where similar issues were successfully addressed.
- • These learnings are then packaged into guided steps that even the newest staff members can follow, equipping them to resolve the issue at hand with the highest probability of success.
- Benefits also extend to the front end of the revenue cycle. When ML-generated evidence is shared with upstream colleagues, it clarifies how process and behavioral changes will result in less revenue leakage.
- By consistently incorporating ML into revenue cycle workstreams, this new, higher level of financial performance is sustainable.
TACTIC #4: Bulk and Automate Appeal Letters
Identifying and managing denials in bulk is an effective way to expedite reimbursement. However, payers often reject bulk appeals and settlements, leaving providers with little recourse except to invest more time and resources. Knox offers a different perspective.
FinThrive believes the key to winning bulk appeals is the ability to identify groupings and classifications of accounts, use these data to inform ML tools, and then have the system act on this bulk as a project within the system vs manually. This includes composing and sending automated appeal letters in bulk.
“TruBridge is targeting the use of ChatGPT to compose customized appeal letters. The company is seeing success in the testing of tailored, automated appeal letters for claims reprocessing and to receive accurate provider reimbursement,” according to Tyler Houston, revenue cycle management product strategist at TruBridge. This includes use of ChatGPT for bulk appeals.
TACTIC #5: Move Toward More Automated Data Exchange With Payers
In addition to needing clinical documentation during the denial process, payers increasingly request medical records for risk adjustment, the Healthcare Effectiveness Data and Information Set, and other reviews and audits. Payer requests for records have increased enormously in recent years. And the manual processing of these requests is costly and burdensome for health care providers.
New technology automates components of the payer-provider communication process to reduce phone tag and repeated sending of documents, thereby increasing efficiency for both sides of the reimbursement process. Another benefit of more automated payer-provider data exchange is an instant and electronic record of every clinical data element, piece of patient information, and communication thread already shared with the payer. Payer delay tactics are cut short when providers are armed with an irrefutable log of digital communications.
Mo Weitnauer, chief product officer at MRO, emphasizes that manual intervention may still be required in some cases, but many payer requests can be satisfied in seconds or minutes instead of weeks. MRO makes the most of FHIR-based application programming interfaces to reduce staff needs during data exchange, which has become essential as most health care organizations face staffing challenges.
TACTIC #6: Know When Payers Are Out of Bounds
Providers want to hold payers accountable. But payers are notorious for sporadic, unpredictable denial behavior and fluctuating reimbursement rules. This keeps providers on shaky ground and limits their ability to take strong defensive actions.
Data, metrics, and analytics are essential to establish better operational relationships with payers and engage in corrective conversations during contract negotiations. For example, ML can analyze the entire claim lifecycle and identify all the steps taken from original claim to final outcome. Armed with this type of intelligence, providers understand such factors as recovery rates, likelihood of denials, which cases should be worked, and trends in bulk denials. According to Knox, “Intelligence helps providers make better decisions about which cases to fight, which to outsource, and which to write off.”
Crump suggests maintaining a close eye on new payer rules and measuring the impact on your organization. “Payers employ different processes and procedures, and they require varying documentation to approve a claim.” With that in mind, she provides four recent egregious payer denial trends to monitor:
- Claim letters are going digital. United-Healthcare will no longer mail claim letters. Instead, it is posting claim letters on its portal for providers to access manually or through an application programming interface. Monitoring payer portals is another new task for teams or future automation.
- Readmission reviews expand dates of service. Humana reviewers are asking for one specific date of service in the claim letter, then also requesting the provider send information from any other dates of service 30 days prior.
- Improper page counts trigger denials. Medicaid auditors are denying claims due to page counts on attestation/certification not matching the page counts they received.
- New technical denials emerge. Payers are identifying more technical denials. But upon further evaluation, it was found that claim letters simply were not sent to the correct provider contact.
Crump reiterates the need for expert oversight at each provider organization to understand how each payer behaves and monitor the resultant impact on systems setup, workflow, and data capture. Support to track these changing payer rules is an essential task well suited for automation.
TACTIC #7: Ensure Strong Oversight
New technologies are valuable defense strategies, but a knowledgeable captain must steer the ship. Oversight from experienced revenue cycle and HIM professionals is still imperative to ascertain the accuracy of the process and ensure validity of the final product.
According to Barbara Hinkle-Azzara, senior vice president of HIM operations at HRS, “Health information professionals are well suited for overseeing automated processes in the revenue cycle.” HIM and health informatics knowledge intersect when AI, RPA, and ML are used to automate revenue cycle processes.
Julie Boomershine, HRS director of coding, notes, “We have consulted with clients who were depending on their automated process to produce a more efficient and enhanced work product, only to find the results were inaccurate. We took a closer look at the automated process and were able to pinpoint the issues and errors and provide corrective action steps to ensure accurate results.”
TACTIC #8: Upskill Today’s Staff for More AI Ahead
AI is not a silver bullet for financial sustainability and cash flow protection. However, it is the future of our industry as more technology-driven revenue cycle processes are ahead. For example, KLAS Research cites 13 operational targets for new technology implementations in a recent survey of 49 health care executives. Patient access, workflow, and process efficiency all rank at the top of the list in the August 2023 report.
With this future in mind, savvy leaders are proactively investing in their teams. Lindsey Hall, human resources people partner at CPSI, describes how the health care technology company is leaning into its people-first culture with an evolution to a skills-based organization. CPSI sustains a strong focus on learning and performance with the use of skills and capabilities to drive development, career opportunities, and employee satisfaction across the organization. Using skills as a common language, CPSI employees can imagine and consider broader career pathways and development opportunities within their current roles to help deliver business outcomes, prepare for the future of work, and grow their careers. Hall provides the following practical ideas:
- Categorize the skills needed for the future.
- Create learning pathways for entry point roles and critical roles.
- Provide employees with career path options and how to achieve them.
- Reduce dependency on recruiting staff and upskill instead.
- Build a culture where AI isn’t scary but is seen as a digital helper.
- Treat AI like an actual employee; give bots a name.
- Quality review the AI output, just like you would an employee.
Put Digital Employees on the Front Lines: Save Staff Smarts for the Tough Stuff
AI, RPA, and ML success stories abound within the health care revenue cycle and other operational areas. Digital employees, or bots, are happily and safely handling some of health care’s most laborious and repetitive tasks, thereby freeing skilled staff to focus on more important functions.
When used as guides to focus more valuable work, digital employees drive greater efficiency and stronger denial defenses. And stronger defensive lines are exactly what is needed amid an increase in tech-driven payer denials.
Put your bots on the front lines of health care revenue cycle operations as the first line of defense against technology-driven denials and other nefarious payer practices.
— Beth Friedman is a senior partner at FINN Partners. As part of the company’s Global Health leadership team, she’s a host of FINN Voices on Healthcare NOW Radio and is a frequent health care industry author.
CENTRALIZED DENIAL DATA: An HIM Checklist
Dawn Crump, senior director of revenue cycle solutions at MRO, also advocates for centralized denial management and optimizing denial data. Crump provides a valuable checklist to evaluate centralized denial and appeal management platforms:
- Know the denial reasons most appealed and tap internal experts to verify which remark and reason codes are most appropriate for automated processing.
- Track every detail of the denial and appeal—including denial date, payer, denial reason, appeal data, appeal submission, follow-up actions taken, and more. This information builds the intelligence required to conduct root cause analysis and implement preventive measures. Detailed documentation holds payers to their contractual obligations.
- Note which cases you don’t appeal and why. Perform corrective action if you see trends.
- Auto-assign appeal letters, use templates, and use technology to auto-flag when remits arrive for cases in the appeal process.
- Optimize analytics to assess all denial and appeal details, and be ready to validate your data to payers.
- Dashboard wins and losses across payer types to inform continual process improvement.
By Beth Friedman
For The Record
