FHIR-Based Prior Authorization Solution by Smile Digital Health

The Client

Major US healthcare payer—serving over 20 million Americans across multiple states—is evaluating a number of vendors for a CMS solution. The solution they require needs to meet the technical mandate of the CMS-0057-F solution, as well as offer automated and integrated prior authorization workflows.

The Challenge

To automatically generate real-time prior authorization responses for two complex patient scenarios created by the client. As part of the initial specifications, Smile Digital Health received two comprehensive documents: one, a 36-page PDF which outlines the medical guidelines and policies related to Cardiac PET scans; and two, an equally lengthy PDF of National Coverage Policy for Skilled Nursing Facilities. The final demonstration to the client was due within 5 days. Transforming narrative medical policies from PDFs into re-usable computable guidelines has traditionally required significant manual effort (up to 6 weeks).

The Scenarios

Scenario 1

A 58-year-old man with a history of high blood pressure, obesity, and heart disease. The patient has recently reported occasional chest pain during physical activity. A request has been made for an outpatient PET Scan (heart imaging test) to help diagnose his condition and determine how his heart disease is affecting blood flow.

Scenario 2

A 65-year-old woman was recently hospitalized for a highly resistant bacterial skin infection, which was healing very slowly. The patient is currently on day six of a 21-day course of intravenous antibiotics.

A request has been made for a seven-day stay at the skilled nursing facility where she will continue to receive intravenous antibiotics every 12 hours, specialized wound care, and assistance with managing her central IV line.

The Solution

Smile’s CMS Suite coupled with our Knowledge Authoring Lifecycle Management (KALM) Services streamlined and sped up the process of authoring the policy, which in turn, automated the prior authorization requests. Smile’s KALM Services include an AI-accelerated tool (KALM AI) that generates intermediary FHIR and CQL resources, which reduces the time spent codifying complex policies from 6 weeks to under 4 days. The policies are then reviewed and refined by a Knowledge Expert (KE) ensuring that they pass a human inspection before being deployed in the prior authorization workflow.

These solutions run on our FHIR®-native Health Data Platform and Clinical Quality Intelligence (CQI) Suite to provide real-time, automated prior authorization workflows and ensure compliance, while still enabling data modernization.

The Results

Smile delivered a successful demonstration of real-time automated prior authorization for both scenarios for the US payer well within the 5-day allocated timeframe. Smile’s solution ecosystem tightly integrates the Health Data Platform, the Clinical Quality Intelligence Suite and the KALM Services, which allows real-time accurate PriorAuth adjudication and responses. The authoring of the clinical content—which is the most tedious step in the process—was significantly accelerated by the use of the novel KALM AI tool which enabled a record-breaking level of efficiency never before seen with Prior Authorization workflows.

The Objective

Smile delivered a successful demonstration of real-time automated prior authorization for both scenarios for the US payer well within the 5-day allocated timeframe. The tight integration of the Smile Health Data Platform with the CQI Suite and Content Library allowed for seamless integrative workflow and delivered the immediate response. The authoring of the clinical content—which is the most tedious step in the process—was significantly accelerated by the use of the novel KALM AI tool which enabled a record-breaking level of efficiency never before seen with Prior Authorization workflows.

More About Smile's Solution

Up to 40% of inquiries do not even require Prior Authorization. Automating Prior Authorization requests reduce the heavy burden on Payer intake teams, saving valuable time for both physicians and patients. For cases that do require approval, computable medical guidelines are required to enable real-time responses. Translating clinical knowledge from source materials like medical policies, CPGs (clinical practice guidelines) and measure specifications is traditionally a heavily manual and administrative process, taking weeks if not months in certain cases. Additionally, this effort often needs to be duplicated for other use-cases and scenarios, making this process un-scalable. Smile’s KALM AI module streamlines the process of transforming narrative medical policies from PDF into validated FHIR and CQL resources. Traditionally, this process takes over 6 weeks and requires a highly specialized individual who is an expert at both clinical context and coding. Smile’s ground-breaking innovation delivers a highly accurate, physician-validated, faithful representation of the policies in under 4 days. Additional benefits include:

Better relationships and reduced burdens across both payers and providers,
Enhanced patient care
Positioning Payers to seamlessly manage and reuse policies across multiple use cases, including digital quality measures and cognitive support.

How it Actually Works

The PDF medical policies for both scenarios were processed through Smile’s content development life cycle. The process outlined below offers examples from the first scenario.

The content development life cycle has three steps:

ClinicalReasoningArchitecture-ForAarti

Intermediary Primitive Documents

The medical policy in the first scenario was a 36-page narrative PDF outlining the use of PET scans in heart-related conditions. Leveraging Smile's internal AI tooling, this PDF was converted to intermediary documents to aid in the final codification to CQL.

Without these tools, this first step requires significant time and effort from a highly specialized Knowledge Engineer (KE). The KE must first review the narrative guideline document and understand clinical concepts within; then determine appropriate case features and decision flow; and finally code these into an intermediary document. Manually, this process would take weeks.

Smile’s custom designed AI tools eliminates the need for human effort in this step.

Deterministic AI - An Approach You Can Trust

A deterministic approach with AI is based on published medical research and established clinical policies. Leveraging Smile’s custom AI tooling, policy documents (such as PDFs) are transformed into FHIR and CQL. Since the workflows are based on established guidelines, the process is transparent, reviewable and auditable.

The alternative—probabilistic AI modules—assign likelihoods to potential outcomes and introduce levels of randomness and uncertainty into decision-making. This type of AI modelling renders decisions opaque, inconsistent and difficult to trust, often leading to systemic denials of Prior Authorization requests. Deterministic AI is like a calculator, which given the same inputs, arrives at the same answer. Probabilistic AI is like a weather forecast, giving a probability of rain, based on patterns and data. They are both useful in different ways but for healthcare and prior authorization, Smile's approach is to leverage deterministic AI along with human intervention to deliver Clinical Intelligence. Read more about it here.

Initial Computable Knowledge

These primitive intermediate documents were then transformed into 80+ FHIR PlanDefinitional resources (sets of defined protocols for clinical actions), and 2,000 lines of CQL code by the KE further leveraging the KALM AI tool. The KE reviewed and refined the results to ensure the policy was faithfully represented in its computable form, generating the Initial Computable Knowledge.

Validated Computable Knowledge

In the final stage of the lifecycle, a clinical SME validated the coded policy and finalized the content library using the KALM IDE module. This is an invaluable step because it serves as a human check point to ensure that the narrative was faithfully translated into computable logic.

Now that the guideline is part of the content library, the effort to author the medical policy that outlines the use of PET scans in heart-related conditions does not need to be duplicated for other use cases. It can be reused for any future Prior-Authorization requests around Cardiac PET scans, as well as use-cases like digital quality measures, value-based care initiatives and cognitive support. Also, if changes are made to the guideline, it can be easily updated within the code and deployed instantly to all of these use cases. This delivers a level of speed and consistency in adopting guideline changes never before seen in the healthcare industry.

These steps were followed for the second patient scenario as well. The medical policies for both scenarios were processed through the Smile content lifecycle, concurrently in under 4 days.

Real-time Automated Prior Authorization Response

Once the codified Cardiac PET policy was published to the Payer’s content library, a request for Prior Authorization of a cardiac PET was submitted through Smile’s Da Vinci FHIR CRD (Coverage Requirements Discovery) API—a component of the Clinical Quality Intelligence (CQI) Suite.

The Smile CQI engine then evaluated the request against the codified policy and returned a “PriorAuth Required” result.

Next, an automated request was made to the Smile Da Vinci FHIR DTR (Documentation Templates and Rules) API. The CQI engine dynamically generated and returned an FHIR Questionnaire from the codified policy. The CQI engine made sure to include CQL (Clinical Quality Language) expressions in the Questionnaire to support pre-population from the EMR. This eliminates the need for the provider to manually supply the needed clinical information to support the prior authorization request.

Finally, a request was sent to the FHIR DaVinci Prior Authorization Support (PAS) API, where the CQI engine evaluated the submitted Questionnaire data against the codified policy and returned an adjudicated response.

This entire process was completed in real-time, successfully demonstrating the entire automated prior authorization process in milliseconds, with minimal administrative effort on both the Payer and Provider side.

The Results

Delivery of real-time automated PA for both scenarios within milli-seconds.

Fast, efficient and accurate delivery of computable medical policies within 5 days.

Accelerated authoring of dense and complex clinical content by KALM AI

Ability to build and re-use the content library for additional use-cases.

The Role of AI

The AI component of KALM AI is designed to handle mega-volume document processing and translation, which is a preparatory step. While AI significantly accelerates this step, Smile’s process ensures that a human KE/ SME always refines, reviews and validates the content before it is used in the care pathway. This greatly reduces the risks associated with AI in healthcare around algorithm-bias, bias with diversity and under-served populations, unintentional misuse of AI and lack of transparency. Clinician and SME time is optimized and only utilized to review and refine AI-generated content, before finalizing the medical policies.

Looking Ahead

Most Payers today are evaluating a solution that supports the technical mandates outlined in the CMS-0057-F rule. Innovative Payers, like our client, are also exploring functionality that supports automated prior authorization workflows. This is essential for Payers who are creating a new industry benchmark that improves not only workflow efficiencies, but also impactful population health and value-based care initiatives, along with improving the payer-provider relationship.

Maintenance of medical policies in the content library is greatly simplified. As innovations occur, changes to the policies can be administered through the KALM tooling where FHIR resources can be modified, and changes can be tracked for auditing purposes.

Updated policies can be promoted into production in near-real time, after being validated. This ensures that PriorAuth evaluations and adjudications continue consistently and without interruption.

As mentioned above, since policies can be repurposed in other use-cases, Payers can run proactive quality measures against the same policies. This largely reduces the reactive and manual process of administering clinical quality measures. It also offers the ability to deliver consistent cognitive support to clinicians throughout the care journey.

Real-Time