Artificial intelligence (AI) captured the imagination of many in 2023. AI gets a lot of attention but little understanding or appreciation of what it can do to move revenue cycle management (RCM) forward, improve the patient experience, and answer the question, “Are you receiving the appropriate reimbursement?”
More than 3 of 5 companies are still experimenting with AI, according to a recent survey by Accenture. Only one in four are innovating or achieving stated objectives) and only one in eight firms have advanced their AI maturity enough to achieve superior growth and business transformation. That percentage drops to only 3% for healthcare organizations. This is not surprising. Accenture argues healthcare companies are lagging behind because they tend to be late adopters of digital transformation technology for administrative purposes. But AI mature healthcare software-as-a-service (SaaS) vendors can offer ways to reduce the lag.
AI is not new. There are numerous examples of AI taking a foothold in the healthcare and clinical diagnostics domain, including digital pathology and AI in NextGen Sequencing (NGS) analytics, as well as speech recognition and conversion to clinical notes. But there is still so much that can be achieved with AI.
Removing friction from the patient experience
The upfront information gathering from a patient is rife with friction. Prior authorizations, eligibility, benefits coverage determination and insurance discovery all generally require detailed information exchange.
Consider for a moment a patient who has Blue Cross Blue Shield insurance. The patient provides the policy number at the time of service, but that number alone isn’t sufficient to confirm eligibility or benefits coverage, which is essential to providing an accurate estimate of potential out-of-pocket expenses. The insurance information required can be far more extensive than the patient has readily available.
Complicating matters is that it can be difficult for a patient to put the right information in the patient portal or application that determines authorization, according to the American Medical Association. Reading the insurance card, hunting for the information being requested, typing it all in, and getting it right, according to Forbes, can be difficult for patients. Even just getting the correct payor name is not simple, according to WebMD.
According to a recent poll, many are turning to Robotic Process Automation (RPA). What is RPA? London School of Economics Professor Leslie Willcocks describes RPA as “a type of software that mimics the activity of a human being in carrying out a task within a process. It can do repetitive stuff more quickly, accurately, and tirelessly than humans, freeing them to do other tasks requiring human strengths such as emotional intelligence, reasoning, judgement, and interaction with the customer.”
Unfortunately, RPA is limited to mimicking human actions, including the automated replacement of human keystrokes or application programming interface (API). RPA can help by automating keystrokes. But to truly remove friction from the patient experience, organizations need to look beyond RPA and adopt AI to remove keystrokes and other steps, such as putting the onus on the patient and the provider to provide the information. AI applied in the right places can uncover the underlying payor details are needed to process a claim.
Simplifying interactions with payors
For each payor response there is in many cases a need for manual intervention, requests for additional information, unnecessary cognitive load, and pressure to resolve not be impacted by timely filing deadlines. There are a myriad of acknowledgments, denials, and reason codes.
Understanding the payor requires having an agent on the phone to get the information from the patient. This is where AI can help. AI can uncover the underlying payor details, including eligibility, coverage, and patient responsibility for a particular claim. It can also discover the payor plan details for that claim to be processed without manual intervention.
How can we use the small amounts of information the patient has and get to where we need to be? Through optical character recognition (OCR). OCR can interpret the insurance card image and text data and feed that into an AI that can lead to eligibility determination. AI can discover the RCM payor and details for that claim so that can be processed without manual intervention.
Machine learning-based historic data models can also assist with healthcare claim acknowledgment responses and use natural language processing (NLP) to translate them to the appropriate reason codes.
Translating payor responses into actionable next steps
Another AI financial game changer to the RCM process is the ability to determine how likely there is a problem with a particular claim, and proactively red flag or even solve the problem.
AI reduces noise, accelerates resolution, and can automate parts of the RCM process that previously required manual intervention.
An accurate picture of expected payor reimbursement is crucial to many RCM and financial functions. Contracted plans can be complex. Even harder to evaluate are non-contracted health plans. Machine learning models, trained on recently adjudicated claims, can overcome those challenges and provide accurate information based on rule history that may not be published:
- Expected allowed amount.
- Estimated copay
- Estimated coinsurance
- Risk of coverage limitations
AI can also help with exception processing prioritization. Imagine an AI engine that could assign and prioritize claim exceptions based on:
- Likelihood of payment collection
- Billing team member expertise or efficacy
- Claim value
- Timely filing deadlines
With edited configurable rules, AI can determine if a claim is likely rejected because of incorrect or incomplete payor information or patient ineligibility and use automation to resolve many issues.
Rich and highly configurable AI can then quickly determine the probability of reimbursement to help prioritize the claims that still require intervention and then redirect those needing human attention to the best available team member.
Because there will be times when an agent is involved in handling exception processing that needs to be acted upon manually, AI-supported RCM can produce assignments daily to determine manual work.
AI can prioritize and determine the billing team member who is most effective at resolving various denial types and route new denials to that person most likely to get the best outcome. AI-driven workflow automation significantly reduces the manual work required to escalate, mobilize, coordinate, and resolve claim disputes.
RCM platform support for algorithms or AI can also drive efficient automation of workflow adaptation to payor changes. AI can help determine problems and then use those same models to provide updates and to determine if information is inconsistent.
Future-Ready RCM Infrastructure Saves Time, Money
Dirty or unstructured data leads to unintelligent AI. Purposeful data modeling in preparation for AI use requires constant vigilance in every step of the process to ensure the integrity of the data and the results.
AI reduces the number of claims needing to be touched by a billing team member and corrects input errors, and ultimately gives back time and focus to diagnostic leaders. AI helps teams eliminate pain points, shortens turnaround time, reduces cost, and enables easier dollar recoupment that was either previously lost or underpaid by payors.
Saving time and money on tasks means a greater focus on what is most important to clinicians – the patient. It means better insights, less expense, and more opportunities to take on additional workloads and deliver better results.
About Jeff Carmichael
Jeff Carmichael is the Senior Vice President of Engineering at XiFin, Inc., a provider of SaaS-based healthcare revenue cycle management (RCM) and workflow automation solutions. Jeff Carmichael’s engineering leadership spans over 20 years and encompasses networking, security, and Healthcare software and systems. He brings a career long focus on data-driven insights and prediction through advanced data modeling across several industries. Prior to joining XiFin, Jeff led worldwide software development for the network and security division of LSI Corp. He has held senior level leadership positions at several successful startups, and divisional leadership positions at Intel. Jeff holds a B.A. in Mathematics from San Jose State University.