SubHero Banner

The current health care payment process is unsustainable. An estimated $200B in administrative waste is generated annually that health care payers and providers are unable to reduce on their own.1

Imagine how we could chip away at this waste if, at the moment it happens, we caught an incorrect code entered for a critical drug, test or procedure — and could prevent the denial of services to a patient. Imagine if we could estimate the cost for that patient at the time they walk in the door and automatically capture a detailed breakdown of his or her encounter. What impact would that have on cutting down mistakes in filing, reviewing, denying and resubmitting service claims for reimbursement? By ensuring that billing after a visit is better anticipated and understood by the consumer, it’s possible to capitalize on a number of efficiencies.


With artificial intelligence (AI), this potential for accuracy and efficiency can be realized, and not a moment too soon. Critical applications are already in play at many points in the payer and provider payment process. The benefits of applying AI to the payment continuum go far beyond improving efficiencies and reducing waste: Applications across pre-service, at service and post-service are making a significant impact on improving the patient experience. 




Creating a complete pre-service picture

In a perfect world, providers and payers would have a holistic view of a patient’s insurance eligibility and benefit data — including policies, terms and coverage requirements — before they walk through the door.

Unfortunately, the harsh reality of our current health care system is that it lacks access to complete benefits data at the initial encounter. This disconnect hinders the ability to accurately estimate patient out-of-pocket costs and impacts coordination of benefits and payer liabilities. The results are confusion over cost of care for the patient and confusion over administrative costs for the professionals behind the scenes, which could otherwise be avoided. 

Perhaps one of the biggest challenges of a fragmented payment system is the lack of price transparency. More realistic expectations about costs before seeing a care provider can reduce billing confusion for patients and lead to streamlined communications between patients, providers and health plans about their respective financial liability. In fact, patients have been clamoring for better price transparency2 when it comes to health care services. Recent government regulations3 have brought even more attention and focus to the topic.

Studies show patients who receive an unexpected bill are more likely to switch from their preferred provider, costing the average hospital up to $100 million each year.


Once the right data is aggregated, AI technologies can help patients anticipate their financial obligations pre-service. Applications of AI could interpret policies, benefits and costs, and not only relieve the administrative burden associated with combing through that information, but also increase the accuracy of the cost estimate for the patient . Payers and providers could then educate and empower their consumers in a faster, more efficient way — and an informed patient is a loyal one. Studies show patients who receive an unexpected bill are more likely to switch from their preferred provider,4 costing the average hospital up to $100 million each year.5

Prior authorization (PA) is another frontier that can be streamlined by the application of AI. PA is a process that is intended to help curb unnecessary medical spending by requiring communication between the provider and the health plan before the service is rendered or a prescription is filled. Once the health plan approves the PA request, the treatment proceeds accordingly.

As with several of these other examples, the PA process today can be difficult to navigate. Communication across various stakeholders, barriers between different data systems and patient unfamiliarity with the process itself are just a few of the obstacles. With the proper infusion of AI, routine PA requests can be processed and approved with little manual intervention.  Providers can also use AI to predict if a PA claim will be denied — in that case, the provider can ensure adequate documentation is available to support the need for the treatment. That translates to less waste and administrative effort during the appeals process if there is disagreement between the provider and health plan.  

Technology can certainly help streamline the process for routine PA requests, but it should be noted that human oversight and decision-making still play an essential role. For more complicated cases or in the case of a denial, human judgment is critical. By automating the processing of routine requests, human reviewers can spend more of their time where it is needed most.


Accurately documenting patient care

Another pain point along the payment continuum where missteps often occur is in coding and documentation of a patient visit at the point of care. Incomplete or delayed clinical documentation and inaccurate coding of diagnoses, procedures and supplies involved can affect medical necessity determinations, reimbursement and quality scores. 

In a recent survey, 84% of hospital leaders named clinical documentation and coding as their top area of vulnerability for lost or decreased revenue. In the same survey, 68% of hospital leaders said they do not think their IT solutions are equipped to manage DRG* coding.6 

AI and advanced analytics have the potential to move health care from being reactive to proactive, offering the information needed for a complete patient picture and enabling technology to make appropriate recommendations for coding and documentation. For example, Optum coding technology prompts providers with a pop-up screen when an order isn’t coded correctly. Catching it in the moment leads to more accurate claims, resulting in fewer denials and faster reimbursement for services.


Along those lines, software that leverages natural language processing (NLP) and deep learning (DL) algorithms can improve the breadth and depth of medical necessity reviews. Specially trained physician advisors can spend hours reviewing individual cases to confirm medical necessity determinations, making manual review of the full case load cost prohibitive. DL algorithms trained on case reviews can review 100% of a hospital’s cases and predict which ones will require closer inspection by physician advisors. Not only is the hospital avoiding denials and achieving appropriate reimbursement — the physician advisors themselves are being deployed where their skills are most needed. 


Improved accuracy in documentation and coding also has the potential to reduce health care costs on a broader scale. 

For example, predictive analytics can identify members who are at higher risk for readmission before they are discharged. AI can instantly leverage millions of data records, going beyond physical health and medical history, to look at socioeconomic data drawn from ZIP Code-level research, environmental data, self-reported education level and much more. These social determinants of health can affect readmission rates,7 and a more comprehensive picture of a person’s health needs means providers and payers can create a treatment and post-discharge plan that both closes a potential gap in care and improves outcomes. 

This collaboration between providers and payers is key. Correct documentation of the treatment and after-care plan results in more accepted claims by health plans and swift implementation of the patient’s care plan. When patients receive the right after-care resources, it also lowers the likelihood of a future hospital admission— a win-win scenario that combines better outcomes with lower total costs of care. 

* Diagnosis -related group (DRG) is any of the payment categories that are used to classify patients for the purpose of reimbursing hospitals for each case in a given category with a fixed fee, regardless of the actual costs incurred.


Closing the gap post-service

Now that you have applied AI to payment processes before service and during service, two more hurdles remain on the payment continuum: claims processing and patient payments post-service. 

One common problem that often leads to service claim denials is missing or incomplete codes. Claims submissions with inaccurate, incomplete or missing information require costly follow-ups and rework. In fact, the Centers for Medicare & Medicaid Services (CMS) estimate coding errors resulted in $31.6B in improper payments in 2018.8


Many providers already use technology with NLP to identify gaps in coding that can lead to denials. NLP takes what a physician says — captured as unstructured data in clinical notes — and translates it into discrete digital data that machine learning (ML) algorithms can interpret and analyze. ML can quickly process large amounts of data into manageable, meaningful amounts, then identify patterns in the data sets and make predictions. In this example, the software uses ML to suggest the appropriate code based on millions of reference cases. It also helps identify where additional information, detail and support are needed for claim approvals. 


Whether an erroneous claim is submitted intentionally or due to human error, it all contributes to the mounting costs of fraud, waste and abuse in the health care system. In 2016, the cost of waste represented an estimated 31% to 41% of all health care expenditures in the United States.9

Human error is not the only source of waste. Fraudulent billing activities in health care also contribute but are not easy to identify or validate, which is another area where machine learning can help. 

When applied to post-service billing activities, ML can widen the net of what is being examined within health care claims and more accurately assess the risk of the claim being incorrect. Health plans that effectively use ML technology and extensive data assets can lower claims risk — helping to reduce the cost of administrative waste generated each year.

Patient billing is another area that commonly results in delayed, incomplete or missing payments. In 2016, 68% of patients failed to pay off their full medical bill balances, and that number is expected to reach 95% by 2020.10 Most businesses today have adopted more tech-savvy means of accepting consumer payments, such as payments via smart phone apps or online payment portals, in order to fit seamlessly into their consumers’ mobile lifestyle and create easier channels for payments. However, many health care organizations still rely on printed and mailed bills, payable only by credit card or check.

Using ML, organizations can take a more personalized approach to patient billing. Variables such as whether the member typically pays by check, credit card or e-payment, which day of the month do they usually pay, what day of the week, etc., help paint a clear picture of consumer payment habits and which billing channels are best suited to reach them. 


Driving payment integrity, improving patient satisfaction

AI, NLP and ML are more than data and analytics buzzwords. They are the way forward in delivering seamless, efficient and effective care that exceeds a patient’s expectations, takes waste out of the system and lessens the administrative burden on care providers.  

AI provides opportunities to create tools and transform a process that often puts undue physical, financial or emotional stress on consumers, and creates friction in relationships between providers, payers and patients alike.

Health care organizations are at varying stages of progress when it comes to implementing an AI strategy, but that doesn’t mean improving the payment continuum is only an option for more advanced organizations. 

Taking a strategic look at each phase of the payment continuum allows us to focus on how the practical application of new technologies can make improvements today while keeping an eye toward future opportunities in payment integrity for payers and revenue cycle management for providers. By chipping away at one of the largest sources of waste, we can bring down costs and also improve patient and provider experiences with the health system.

Accordion Block
  • Applying AI in the Health Care Payment Nexus

    Benefits — Insurance Verification

    Pain point: Lack of access to complete benefit data hinders the ability to accurately identify patient out-of-pocket costs, coordination of benefits and payer liabilities.

    Solution: Using AI to interpret policies, benefits and costs not only relieves the administrative burden, but also increases the accuracy of the cost estimate for the patient.

    Eligibility — Prior Authorization

    Pain point: Communication across various stakeholders, barriers between different data systems, and patient unfamiliarity with the process can create myriad obstacles.

    Solution: With the proper infusion of AI, routine prior authorization requests can be processed and approved with little manual intervention; more complicated cases can be flagged for human review.

    Case Review — Medical Necessity

    Pain point: Physician advisors can spend hours reviewing cases to confirm medical necessity, making manual review of all cases cost prohibitive.

    Solution: AI can review 100% of cases and predict which ones will require closer inspection, thereby avoiding denials, achieving reimbursement, and deploying physician advisors where their skills are most needed.

    Coding — Documentation and Coding

    Pain point: Incomplete or delayed clinical documentation and coding can have a far-reaching impact on revenue, from inaccurate reimbursement to negative HCAHPS or NPS scores.

    Solution: AI provides the data and analytics needed to assemble a complete patient picture and make appropriate recommendations to the clinician for more accurate documentation and coding.

    Claims Processing — Billing and Adjudication

    Pain point: Claims submissions with inaccurate, incomplete or missing information require costly follow-ups and rework.

    Solution: Natural language processing (NLP) can identify where additional information or documentation is needed for claims approvals. Software can then suggest how to fill those gaps by using NLP again to extract relevant information from clinical notes.

    Payments — Patient Payments

    Pain point: Many health care organizations still use printed and mailed bills, which alienates much of the population that prefer e-payments.

    Solution: Machine learning enables more customized billing in order to reach members through their preferred communication channels. 

Practical insights on tech, data and analytics innovation in health care.


Data. Analytics. Expertise. Learn how we use OptumIQ to change health care for the better.

Learn More


Technology innovation at Optum

Learn about new and emerging technologies that are pushing health care forward.



What challenge can we help you solve?

Contact us

  1. Institute of Medicine of the National Academies. The Best Care at Lower Cost: The Path to Continuously Learning Health Care in America. Washington, DC: National Academies Press; 2013.
  2. American Medical Association. 8 Ways to Improve Health Care Price Transparency. Published Sept. 20, 2018.
  3. The White House. Executive Order on Improving Price and Quality Transparence in American Healthcare to Put Patients First. Issued on June 24, 2019.
  4. Becker’s Hospital CFO Report. Patients more likely to ditch preferred hospital after a surprise visit. Published March 4, 2019.
  5. Patient Bond Blog. 5 Ways Healthcare Organizations Can Boost Patient Loyalty. Published June 26, 2018.
  6. RevCycle Intelligence. Clinical Documentation and Coding Top Revenue Cycle Vulnerability. Published Feb. 11, 2019.
  7. American Hospital Association (AHA). Study: Social risk factors linked to hospital readmissions, penalties. Published March 12, 2019.
  8. Centers for Medicare & Medicaid Services. Comprehensive Error Rate Testing (CERT). 
  9. O’Neill DP, Scheinker D. Wasted health spending: Who’s picking up the tab? Health Affairs Blog. May 31, 2018.
  10. TransUnion. Patients May be the New Payers, But Two in Three Do Not Pay Their Hospital Bills in Full; Published June 26, 2017."