Predict Prior Authorization

30% of physicians surveyed by the American Medical Association reported that prior authorization (PA) has led to a serious adverse event for a patient in their care (1). 94% state it delays access to necessary care, and an overwhelming majority perceive the process to have a negative impact on patient health outcomes. PA requests are also extremely time-consuming; physicians and their staff report spending an average of two business days each week completing requests (1).

Payers contend that PA is extremely beneficial despite provider complaints. In a recent survey conducted by America’s Health Insurance Plans, 98% of plans surveyed reported that PA improves the quality of care and provides support for evidence-based treatment (2). 91% also reported using PA to ensure patient safety, and 79% report that it lowers healthcare spending. However, payers also grapple with the associated administrative burdens.

Size of the Problem

• 21% of physicians report that PA delays have led to an avoidable hospitalization (1).
• 12% of 182 million prior authorizations were fully electronic in 2018 (3).
• Roughly 50% of requests in 2018 were conducted entirely via faxes and phone calls (3).
• $25 billion in costs has been attributed to administrative PA processes (3).

‍

PA was designed to act as a patient-safety and cost-saving measure for payers to ensure appropriate provider utilization management, but the massive administrative burden is taxing for both parties. Administrative PA processes have been estimated to contribute as much as $25 billion annually to total healthcare costs and can significantly delay care for patients (3).

The PA request process is complex, laborious, and not standardized. Payer requirements are varied, change frequently, and may even differ across health plans offered by the same payer. Because clinical workflows and billing systems are rarely integrated, providers must manually retrieve pertinent information from a variety of data sources and transfer it into authorization requests. This process is prone to human error, lacks coding consistency, requires clinical staff to manually review individual plan specifications, and frequently takes place over multiple phone calls and faxes that strip away the structure of a patient’s medical record.

1. Data Entry Automation with AI: Implement AI systems to automate the collection and entry of data necessary for PA requests. This reduces the time and errors associated with manual data entry, allowing doctors and their staff to spend more time on patient care rather than administrative tasks.‍
2. Predictive Analytics for Quick Identification of Required PA: Use AI models to analyze the characteristics of PA requests and quickly predict whether a request will require prior authorization. This can facilitate more efficient preparation and submission of necessary requests, reducing delays in treatment.‍
3. Optimization of the Approval Process with AI: Develop and deploy AI models that can review PA requests against payer approval criteria and predict the likelihood of approval. This allows for the pre-screening of requests, where those with a high probability of approval can be processed and approved more quickly.‍
4. AI Predictive Model for PA Approval: We have developed an artificial intelligence model using a synthetic database that mirrors real-world conditions to predict the approval of prior authorization requests. This model is trained with variables such as patient age, gender, health condition, requested medication, and urgency of the request. The goal of the model is to determine the likelihood of PA approval, thereby facilitating quick and efficient decision-making in handling requests. With this model, we aim to significantly reduce the waiting time for treatment approval and improve patient satisfaction and health outcomes.

• Electronic Health Records: EHR data with comprehensive patient histories of vital signs and symptoms, problem lists and chief complaints, tests results, diagnoses and procedures, and prescriptions.
• e-Prescribing Data: Data from electronic prescriptions detailing key information about medications, dosage, patient instructions for frequency and timing, and available refills.
• Medical Claims: Data extracted from health insurance medical claims with details about dates and place of service, diagnosis codes, key procedures, use of medical equipment, and provider specialties.