OVERVIEW
The prior authorization (PA) process in healthcare, while necessary for ensuring the medical necessity and cost-effectiveness of treatments, often results in significant delays. This study explores the potential of AI to streamline the PA process, based on research conducted by Humeira Badsha and colleagues. The study compares the efficiency of a proprietary AI rule engine, supported by an OpenAI-based machine learning (ML) model, against traditional methods used by insurance companies. The results indicate that AI can significantly reduce the time required for authorization, thereby improving patient care.
INTRODUCTION
The PA process is a critical step in healthcare, ensuring that patients receive necessary treatments and medications. Mandated by insurance companies, this process confirms the medical necessity and cost-effectiveness of prescribed treatments. However, it often results in significant delays, potentially leaving patients at risk of untreated conditions. Recent advancements in AI offer promising solutions to streamline and expedite the PA process.
BACKGROUND AND PURPOSE
Patients with chronic autoimmune conditions often require expensive treatments that necessitate PAs. According to a 2023 American Medical Association (AMA) survey, 94% of physicians reported treatment delays due to PA, with 78% of patients abandoning treatment as a result.1 A 2020 survey of rheumatic patients revealed that 48% required PA for their medications.2 The American College of Rheumatology’s 2020 position statement emphasizes the need to modernize and streamline the PA process.3 This study aims to compare the efficiency of an AI rule engine against traditional methods used by insurance companies in processing PAs for rheumatology investigations and medications.
METHODS
This study involved 50 patients from a rheumatology clinic who required PA for investigations or treatments with biological medications. Anonymized medical reports were uploaded into an AI rule engine and simultaneously sent to insurance companies for approval. The AI engine analyzed the reports to classify the appropriateness of the patients’ diagnoses and requested treatments, and this data was compared to the insurance companies’ responses.
The Algorithm Health system employs a concept called Guided AI. A medical preprocessing engine handles the initial processing of data before it is fed into the ML model. This preprocessing ensures significantly greater accuracy by avoiding many of the common failings of pure AI models. The proprietary data extraction model created by Algorithm Health ensures complete and appropriate extraction of digitized data from medical records, enhancing the preprocessing stage. The integration of the data extraction model with the preprocessing engine and the customized ML model is the unique selling proposition of Algorithm Health’s platform.
RESULTS
The AI rule engine demonstrated remarkable efficiency. Among the 41 investigation approval requests, the AI engine deemed 95% appropriate within a minute, whereas insurance companies approved only 82.9%, with 17.1% still pending and 2.4% rejected outright. Approval times varied, with some investigations taking over 2 weeks. Additionally, 29.2% of the investigations required further queries. For the 43 medication approval requests, the AI engine matched every diagnosis and treatment plan, but only 81.3% were approved by insurance companies, with 18.6% pending. Further queries averaged a delay of 5 days.
CONCLUSION
This study highlights significant delays in the PA process, which can adversely impact patient health. The AI rule engine demonstrated the potential to significantly expedite the process by quickly identifying appropriate requests and reducing patient wait times. AI can also help eliminate unjustified requests, further streamlining the process.
IMPLICATIONS FOR HEALTHCARE
Integrating AI into the PA process offers several benefits, as highlighted by recent studies and expert opinions:
- Efficiency: AI can rapidly process and analyze patient data, reducing approval times significantly. According to a report by McKinsey, AI can increase productivity and efficiency in care delivery, allowing healthcare systems to provide more and better care to more people.4
- Accuracy: AI ensures that only appropriate requests are approved, eliminating unnecessary denials. The National Academy of Medicine (NAM) identified three potential benefits of AI in healthcare: improving outcomes for both patients and clinical teams, lowering healthcare costs, and benefiting population health.5
- Reduced burden: Streamlining the PA process allows healthcare professionals to focus more on patient care. AI can help reduce the paperwork burden that often leads to burnout among healthcare workers.6
- Improved patient outcomes: AI can assist in early diagnosis and treatment, leading to better patient outcomes. For example, AI can help with preventive screenings and risk assessments, identifying potential health issues before they become severe.5
- Cost savings: By reducing delays and improving efficiency, AI can help lower healthcare costs. This is particularly important in resource-poor settings where access to healthcare is limited.4,5,7
- Enhanced patient experience: AI can improve the patient experience by providing faster responses and more personalized care. For instance, AI can help patients better understand their health conditions and treatment options.4,6,7
