Reshaping the Healthcare Industry with AI-driven Deep Learning Model in Medical Coding

1. Overview

Hospitals and healthcare systems have faced unprecedented challenges in recent years, from surging patient volumes to a widespread labour shortage, exacerbated by the coronavirus pandemic, rising inflation, and mass resignations. These disruptions have significantly impacted provider productivity, patient experience, and the healthcare revenue cycle. Amidst unprecedented challenges  and revenue losses exceeding $53 billion in 2021, the healthcare industry is in an urgent need for transformative solutions. With a focus on AI in medical coding, this paper explores how automation and scalable technologies can enhance medical billing and revenue cycle management (RCM) efficiency. By leveraging AI, healthcare organizations mitigate coding errors, streamline administrative processes, and optimize revenue management amidst labour shortages and rising healthcare costs. Adopting AI-driven solutions offer healthcare organizations a promising path to sustainability and efficiency.

2. Abstract

In the complex world of healthcare reimbursement, hospitals face numerous challenges with coding, especially when it comes to denied claims. With approximately 11% of all claims facing denial and some providers seeing these rates soaring as high as 30%, the urgency to address coding inefficiencies cannot be overstated. Alarmingly,42% of denials stem from coding issues alone. Poor coding can result in various revenue challenges, including a surge in underpayments or overpayments, heightened payer audits targeting specific code combinations, and stagnation or decline in per-visit revenue. Moreover, the payer takebacks add financial strain on healthcare providers. Simultaneously, mounting coding backlogs and high turnover rates among coding staff further exacerbate these challenges.

A report from Crowe LLP, a global public accounting, consulting and technology firm revealed a concerning trend in denial rates, which rose from 10.2% to 11% in 2022, resulting in 110,000 unpaid claims for an average health system. Particularly, prior-authorization denials on inpatient accounts have seen a surge, leading to denials representing 2.5% of gross revenue in August 2022—a substantial 67% increase from January 2021. These denials burden providers with prolonged and costly appeals, impacting patient care and financial stability, as payors question the necessity of provided care.

Source: Crowe LLP Report

Moreover, over the past five years, hospitals have confirmed that claims denial rates have surged by over 20%, surpassing the 10% mark. A September 20, 2022, MGMA (Medical Group Management Association) Stat poll has identified staffing (58%) as the primary challenge for medical practices, ahead of Expenses (20%), Revenue (17%), Technology (2%), and Other (2%). Industry averages reveal that nearly 20% of claims face denial, with a staggering 60% of denied claims never being resubmitted. The process of reworking or appealing denials come at a cost of $25 per claim for practices and $181 per claim for hospitals.

With potential annual savings of $122 billion through automation and an additional $16.3 billion through further automation, AI emerges as a transformative force in healthcare administration. This paper is designed to give RCM leaders and healthcare billing managers a complete understanding of how AI in medical coding can improve operational efficiency and financial outcomes. It delves into the necessity of transitioning from manual coding methods to automated systems, due to inherent risks of errors, inefficiencies, and inaccuracies associated with manual processes. Through the integration of AI technologies, particularly machine learning algorithms, healthcare organizations can significantly reduce coding errors, enhance accuracy, and streamline the overall coding workflow.  AI has the ability to decipher and interpret unstructured medical text, transforming it into structured data suitable for coding. Moreover, AI-driven coding solutions can adapt and learn from past coding patterns and decisions, continuously improving their accuracy over time. By automating routine coding tasks and providing intelligent insights, AI empowers healthcare organizations to reduce manual errors, optimize revenue cycle management, and ultimately deliver better patient care.

3. Introduction

Medical coding has its roots in the mid-20th century, emerging as a solution to the complexities faced by the healthcare industry in managing patient records and reimbursement processes. With the advancement in medical practices and technologies, the volume of healthcare data became overwhelming, necessitating a systematic and efficient method of organization. Medical coding involves converting healthcare diagnoses, procedures, and services into standardized alphanumeric codes derived from various medical records. Initially recognized in the early 1960s, the need for a standardized coding system led to the creation of the Current Procedural Terminology (CPT) in 1966, aimed at streamlining communication among healthcare stakeholders. Additionally, the World Health Organization (WHO) introduced the International Classification of Diseases (ICD), notably ICD-10, in the early 1990s, standardizing the classification of diseases globally. Medical coding underwent a digital transformation, transitioning from manual to automated coding systems, which expedited processes and enhanced accuracy.

In the United States, the landscape of medical coding is intricate and ever-evolving, characterized by a vast array of codes that demand precision and continuous adaptation. With nearly 11,000 Current Procedural Terminology (CPT) codes, the system undergoes significant changes annually. In 2023, 225 new codes were added, 75 deleted, and 93 revised, impacting various sections of the coding guidelines. The International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) system in the U.S. boasts over 70,000 codes, with 1176 new billable codes introduced in 2023. Additionally, 251 codes were deleted, 35 had their descriptors revised, and 36 were converted to parent codes, effective from October 1, 2023.

Some of the top coding errors identified by American Medical Association (AMA) include unbundling, upcoding, incorrect modifiers, and failure to reference National Correct Coding Initiative (NCCI) edits for multiple code reporting. These errors emphasize the critical need for meticulous attention to detail in medical coding practices to ensure accurate reimbursement and compliance. Certified medical coders play a crucial role in ensuring accurate documentation, facilitating proper reimbursement, and supporting data-driven decision-making in healthcare administration.

However, medical coders face numerous challenges in their daily operations, from the time-consuming manual data retrieval processes to the complexities of accurate coding across various specialties. The need to cross-check data for accuracy and the requirement to be well-versed in all specialties, even those unfamiliar to them, add layers of difficulty. Traditional code search methods are inefficient and costly, while billing complexities and the importance of navigating National Correct Coding Initiative (NCCI) edits, cause further problems. Manual coding and the lack of oversight on coder performance result in additional hurdles. Moreover, data entry tasks pose risks of typographical errors with potentially significant consequences. Compliance with Current Procedural Terminology (CPT) guidelines is crucial, with new rules introduced in 2023 adding further complexity.

Medical coding plays a pivotal role in the healthcare sector for effective revenue cycle management and billing. However, these growing challenges drive the pressing need to address them. For instance, the average denial rate for insurance coverage has increased to 23% over the past three years, predominantly due to improper coding practices and upstream challenges, leading to substantial revenue loss and operational inefficiencies. Manual errors, outdated code sets, and miscommunication among stakeholders have underscored the urgency for innovative solutions. Existing solutions, such as Computer Assisted Coding (CAC), have shown promise but lack scalability and production readiness. Deep learning-powered medical coding assistance, offers a transformative solution to these longstanding challenges. Utilizing neural network architectures, explainable AI, and foundational pre-trained models promises faster time to production and explainable outcomes. With improved cash flow, enhanced manageability, and consistent coding, the healthcare industry can revolutionize its billing and revenue cycle management. Deep learning models in medical coding can not only streamline clinical encounters to billable codes but also ensure accuracy, efficiency, and compliance.

4. The Challenges/ Pain Points

The current healthcare system is grappling with substantial challenges, ranging from prolonged processing times to heightened errors and inefficiencies in revenue cycle management. These challenges not only impact the financial health of healthcare facilities but also hinder the efficient delivery of patient care. Some of the significant challenges that impact the healthcare industry are:

1. Prone to Manual Errors

In the medical industry, reliance on manual input for processes like medical coding is susceptible to errors due to factors such as human fatigue, oversight, and limitations in expertise. This situation persists despite advancements in technology and automation. For example, the manual encoding of pathological data using systems like SNOP (Systematized Nomenclature of Pathology) and SNOMED (Systematized Nomenclature of Medicine) in hospitals can lead to error rates and discrepancies in coding accuracy. These errors can result in higher claim denials, revenue loss, and even federal penalties due to inaccuracies in billing documentation.

2. High Degree of Complexity and Compliance

Managing manual charge capture involves navigating a multitude of codes and compliance regulations, which can result in using outdated code sets, handling excessive irrelevant documents, and working with highly unstructured content.  This complexity arises from the need to interpret and apply numerous codes accurately, often resulting in the use of outdated codes due to the sheer volume of updates. For instance, outdated code sets can lead to claim denials and delayed payments, impacting revenue flow. Handling excessive irrelevant documents can increase the risk of errors and inefficiencies in the billing process. Working with highly unstructured content can make it challenging to accurately capture charges and comply with billing requirements, potentially leading to compliance issues and financial losses. These factors highlight the complexity and critical need for precise management in manual charge capture processes to ensure accurate billing and compliance with healthcare regulations.

3. Unscalable and Capital Intensive

Manual medical coding is a labor-intensive and capital-intensive process that demands specialized professionals and meticulous regulatory oversight. It demands skilled professionals to have in-depth knowledge of medical terminology and coding guidelines, leading to higher operational costs associated with hiring and training them. Moreover, the manual nature of coding makes it unscalable, particularly when faced with missing data or the need for extensive manual inspection at a large scale. This reliance on manual coding not only increases the risk of errors but also significantly raises the operational costs associated with healthcare services. The inefficiencies and limitations of manual coding underscore the need for more scalable and automated solutions to streamline the coding process, reduce errors, and enhance cost-effectiveness in healthcare operations.

4. Suboptimal Stakeholder Experience

Incorrect coding in healthcare can have far-reaching consequences, impacting various stakeholders such as payers, providers, and patients. When coding errors occur, miscommunication ensues, leading to higher turnaround times and delays in accounts receivable collection. For payers, inaccurate coding can result in claim denials or delays in processing claims, affecting the reimbursement process and potentially leading to financial losses for healthcare providers. Providers, on the other hand, may face challenges in receiving timely payments for services rendered due to coding discrepancies, which can disrupt cash flow and create administrative burdens.

Patients encounter significant challenges due to inaccurate coding in healthcare and the administrative procedures managed by Third-Party Administrators (TPAs). Incorrect coding can trigger billing discrepancies like overbilling, underbilling, or coverage denials, impacting patients financially and eroding their trust in the healthcare system. For example, patients may be charged for services that should have been covered by insurance, leading to unnecessary out-of-pocket expenses. Conversely, TPAs can cause delays and inefficiencies in accessing care by imposing extensive documentation requirements and intricate claims processing. These delays can impede patients from promptly accessing healthcare providers, resulting in frustration, stress, and potential confusion regarding their coverage. Additionally, delays in claims processing due to coding errors can prolong the reimbursement process, affecting the overall revenue cycle efficiency and potentially impacting the quality of care provided to patients. Both inaccurate coding and TPA administrative processes can detrimentally influence patients' healthcare encounters, underscoring the significance of precise coding practices and streamlined administrative protocols to ensure a seamless and dependable healthcare experience for patients.

5. AI and Other Medical Coding Trends

In recent years, several significant improvements in medical coding have helped advance the healthcare revenue cycle. Two of the most important ones are single-source coding and support AI.

1. Single Source Coding

Traditionally, hospitals divided medical coding tasks between entities handling hospital services and physician codes. But now, single-source coding is gaining popularity, particularly for outsourced services. With single-source coding, one vendor manages both sets of data, simplifying communication among healthcare providers, payers, and patients. This approach streamlines administrative tasks, offering a centralized resource and enhancing payment processing and reimbursement efficiency.

2. AI-enabled Medical Support

In the dynamic landscape of modern healthcare, AI heralds a new era of precision and efficiency in medical coding systems. Leading third-party coding providers combine experienced coders with innovative machine-learning algorithms to improve the accuracy and effectiveness of medical coding systems. AI offers additional checks and automates redundant coding tasks, resulting in increased productivity and fewer errors, leading to higher claim approval rates and faster payments. Moreover, within an end-to-end revenue cycle management (RCM) solution, AI technologies can process and interpret large volumes of data, facilitating informed decisions about revenue strategies based on comprehensive knowledge.

6. Current Industrial Practices and the Adoption of CAC

In the realm of medical coding, current industrial practices are witnessing a shift from traditional manual coding methods towards the adoption of Computer-Assisted Coding (CAC) solutions. These systems offer automation capabilities that assist in suggesting appropriate codes based on clinical documentation, aiming to streamline the coding process and improve accuracy. This shift is crucial as the global medical coding market size was valued at $15.2 billion in 2019, which is expected to grow at 9.6% CAGR between 2020 and 2030, indicating a significant demand for skilled professionals and advanced software solutions to address challenges like staffing shortages, employee burnout, compliance issues, and high costs/thin operating margins.

Computer-Assisted Coding (CAC) can be of two types: rule-based scenario and NLP (Natural Language Processing). While these methods offer significant advantages in efficiency and automation, they also have inherent limitations that can impact their effectiveness.

Rule-based scenarios in medical coding are based on hardcoded rules and logic created by programmers to automate coding tasks. One major limitation of rule-based systems is their inflexibility; they are only as smart as the rules programmed into them. This means that any changes or updates in coding guidelines or requirements can lead to system failures or inaccuracies. Additionally, developing rule-based systems requires extensive domain expertise to define comprehensive rules, making them less adaptable to evolving coding practices.

On the other hand, NLP, which utilizes machine learning, can automate coding tasks by learning patterns from data. However, NLP is highly dependent on the quality and accuracy of the data provided to it. If the data is incomplete, inconsistent, or incorrect, the results generated by NLP may be inaccurate or misleading. This reliance on data quality poses a significant challenge, especially in healthcare where data integrity is crucial for accurate medical coding.

While rule-based scenarios and NLP with Machine Learning (ML) have revolutionized medical coding by enhancing efficiency and automation, they are not without their loopholes. Rule-based systems can struggle with adaptability to changes, while NLP's accuracy is heavily reliant on the quality of the data it processes. These limitations highlight the need for continuous improvement and innovation in medical coding technologies to address these challenges and ensure accurate and efficient coding practices in the healthcare industry. AI technologies offer intelligent automation, dynamic learning capabilities, and the ability to handle complex coding scenarios with greater accuracy and efficiency.

Source: Limitations of CAC

7. Deep Learning Powered Medical Coding Assist: An Effective Solution

1. Model:

An AI-enabled medical coding assist model utilizes foundational pre-trained models with domain entities, allowing for rapid deployment from day zero. These models facilitate a faster time to production through a continuous learning platform, ensuring efficiency and accuracy in coding processes. Utilizing a neural network architecture and auto AI for hyper tuning of parameters, these focus on discovering medical terms with precision. The feedback loop involving clinicians and subject matter experts (SMEs) ensures continuous model upgrades, enhancing the system's ability to adapt to evolving medical terminology and coding requirements. Such models adopt an explainable AI approach, ensuring transparency and interpretability in prediction outcomes.

This involves providing reasoning behind predictions and tracing them back to their original sources. By avoiding the black box approach, the system explains its predictions and traces back to the original sources, enhancing trust and facilitating the adoption of domain-centric business rules for seamless integration into existing workflows.
An effective model would integrate components like hybrid data ingestion, deep learning model, medical entity recognition and extraction, code assignment and ranking, and workflow orchestrator. These components perform crucial functions to streamline the medical coding process effectively. Hybrid Data Ingestion integrates various data types efficiently for coding purposes. Deep Learning enhances the understanding and utilization of medical data through advanced learning algorithms. Medical Entity Recognition and Extraction play a vital role in identifying and extracting key medical entities from clinical documentation for accurate coding. The Workflow Orchestrator ensures seamless coordination and management of coding tasks, optimizing the overall efficiency and accuracy of the medical coding process.

2. Model Features:

AI-enabled medical coding models address the challenges in medical coding, with a focus on outpatient and inpatient medical billing, pharmacovigilance, and clinical trials. For outpatient billing, it expedites claim submissions by efficiently handling procedure and diagnosis coding using ICD-10-CM, CPT, and HCPCS code sets. Similarly, for inpatient billing, it ensures accurate coding of complex procedures and diagnoses with support for ICD-10-CM and ICD-10-PCS, optimizing reimbursement processes within hospital settings. It also supports pharmacovigilance and clinical trials by identifying and reporting adverse events (AEs) and serious adverse events (SAEs) by extracting relevant information from medical records and interfacing with standardized dictionaries like MedDRA and WHO-DDE. These features aim to improve accuracy, streamline processes, and ensure regulatory compliance, showcasing its versatility and effectiveness in optimizing medical coding practices for diverse patient care scenarios.

8. Benefits of AI-Driven Medical Coding

1. Improved Cash Flow/Revenue Growth:

AI in medical coding optimizes healthcare cash flow by reducing claim denials, expediting reimbursement process, and minimizing revenue leakage caused by coding inaccuracies. It automates processes, reduces errors, and accelerates reimbursement by swiftly analyzing clinical documentation. Its advanced algorithms accurately assign codes and provide valuable insights into revenue cycle performance. This automation improves cash flow by streamlining coding workflows and minimizing revenue cycle inefficiencies. It thus ensures precise billing, timely claim submissions, and optimized revenue cycle management.

2. Increased Productivity:

Efficient medical coding enhances productivity by streamlining coding processes, reducing manual errors, and automating repetitive tasks. By leveraging advanced technologies like Deep Learning and AI, healthcare providers can increase coding efficiency, handle larger volumes of data, and allocate resources more effectively, leading to improved operational productivity.

3. Enhanced Manageability:

Advanced medical coding solutions offer enhanced manageability by providing a centralized platform for coding tasks, facilitating seamless data integration, and offering real-time insights into coding processes. This improves the overall management of coding operations, ensures compliance with regulatory standards, and enables better decision-making based on accurate and up-to-date coding data.

4. Real-time coding updates:

Autonomous medical coding solutions offer the ability to continuously update coding algorithms and guidelines in real-time. This allows healthcare professionals to access the latest and most relevant coding information, ensuring they remain abreast of the current coding requirements. By providing real-time coding updates, these solutions mitigate coding errors, minimize claim rejections, and enhance revenue capture efficiency.

5. Consistent Coding:

Consistency in medical coding is crucial for accurate billing, compliance with coding standards, and effective communication among stakeholders. Modern coding solutions, powered by Deep Learning and AI, ensure consistent coding practices by standardizing code assignment, reducing variability in coding outcomes, and maintaining coding accuracy across different healthcare scenarios.

6. Accuracy:

Accuracy is a fundamental aspect of medical coding that directly impacts revenue, compliance, and patient care. By utilizing AI-enabled medical coding models, healthcare organizations can achieve higher levels of accuracy in code assignment, documentation, and reporting.

9. Use of AI Medical Coding and RCM (Revenue Cycle Management) Services

AI Medical Coding and RCM (Revenue Cycle Management) Services have enhanced various aspects of the healthcare industry. In Accounts Receivable, AI streamlines payment processes, reduces errors, and accelerates revenue collection by automating tasks like claims processing and payment posting. In Medical Billing, AI enhances accuracy by analyzing vast amounts of data to generate precise billing codes, leading to faster reimbursements and reduced claim denials. AI-driven tools elevate patient satisfaction by personalizing interactions, refining communication, and enhancing overall experience. Additionally, AI also aids in health information management by managing and analyzing health data, facilitating real-time coding, and ensuring compliance with coding guidelines.

According to the data shared by Statista, in 2021, the artificial intelligence (AI) in the healthcare market was worth around 11 billion U.S. dollars worldwide. It was forecast that the global healthcare AI market would be worth almost 188 billion U.S. dollars by 2030, increasing at a compound annual growth rate of 37 percent from 2022 to 2030. AI will transform the future of healthcare by optimizing workflow, maximizing revenue generation, and elevating the quality of patient care through its integration into medical coding and RCM services.

10. Value Outcome/ Outcome Delivered/ Conclusion

In the present scenario, the integration of AI in medical coding and Revenue Cycle Management (RCM) services delivers a multitude of value outcomes that significantly enhance the healthcare industry. By leveraging AI technologies, healthcare providers can effectively reduce Account Receivable (AR) days by improving collection rates through agile billing automation and gaining a comprehensive view of aged and high-value accounts. This leads to a more efficient revenue cycle, ultimately lowering costs to collect and increasing margins by centralizing billing processes and scaling RCM review processes without substantial staff cost increases. Operational efficiency is optimized through reduced manual intervention, freeing up bandwidth, and reallocating resources from insurance follow-up to value-added revenue cycle functions. Moreover, you can limit write-offs by minimizing claims reprocessing, eliminating data silos, and enhancing denial management with complete traceability. The transformative impact of AI in medical coding and RCM services not only drives financial performance but also enhances overall operational effectiveness, positioning healthcare organizations for sustained success in an increasingly complex healthcare landscape.

References:

1. American Health Information Management Association (AHIMA). (n.d.). Evolution of Coding Whitepaper. Retrieved from https://ahima.org/media/ycgattab/evolution_of_coding_whitepaper_agshealth_ada.pdf
2. Becker's Hospital Review. (2024, March 29). The biggest challenges facing revenue cycle departments in 2024. https://www.beckershospitalreview.com/finance/the-biggest-challenges-facing-revenue-cycle-departments-in-2024.html
3. Capline Healthcare Management. (2023). The Crucial Role of Coding and Clinical Documentation in Resolving Claims. Retrieved from https://www.caplinehealthcaremanagement.com/wp-content/uploads/2023/07/W02219-Whitepaper-30_The-Crucial-Role-of-Coding-and-Clinical-Documentation-in-Resolving-Claims-2023_06_22.pdf
4. Capline Healthcare Management. (2022). Whitepaper10. Retrieved from https://www.caplinehealthcaremanagement.com/wp-content/uploads/2022/06/Whitepaper10.pdf
5. GeBBS Healthcare Solutions. (n.d.). Medical Coding: Advancing the Healthcare Revenue Cycle. Retrieved from https://gebbs.com/whitepapers/medical-coding-advancing-the-healthcare-revenue-cycle/
6. Invensis. (2024, February 16). 7 Most Common Medical Coding Challenges and Their Solutions. https://www.invensis.net/blog/medical-coding-challenges-and-solutions
7. LinkedIn. (2023, June 29). Overcoming the Top 7 Healthcare Revenue Cycle Challenges. https://www.linkedin.com/pulse/overcoming-top-7-healthcare-revenue-cycle-challenges-apaana/
8. LinkedIn. (2023, November 2). 4 Challenges Of Medical Coding And Why It’s Worth The Effort. https://www.linkedin.com/pulse/4-challenges-medical-coding-why-its-worth-effort-roa4e/
9. LinkedIn. (2023, December 13). Challenges Faced by Medical Coders: Solutions and ... https://www.linkedin.com/pulse/challenges-faced-medical-coders-solutions-insights-eg0qf
10. Medicodio. (2024). Whitepaper_Medicodio-revised. Retrieved from https://medicodio.com/wp-content/uploads/2024/03/Whitepaper_Medicodio-revised.pdf
11. Medicodio. (2023, August 22). What are the common challenges in Medical Coding? - Medicodio. https://medicodio.com/what-are-the-common-challenges-in-medical-coding/
12. Medical Group Management Association. (2023, October 19). Staffing, uncertainty among top pandemic challenges for medical groups heading into 2022. https://www.mgma.com/mgma-stats/staffing-uncertainty-among-top-pandemic-challenges-for-medical-groups-heading-into-2022
13. Nym Health. (2024, March 30). A Window Into the State of Medical Coding in the US. https://blog.nym.health/a-window-into-the-state-of-medical-coding-in-the-us
14. Rely Services. (2024, April 3). Consequences Of Inaccurate Medical Coding and Billing. https://www.relyservices.com/blog/consequences-incorrect-medical-coding-billing
15. RevCycle Intelligence. (n.d.). Medical Coding Is the Next Stop for Artificial Intelligence in Healthcare. Retrieved from https://revcycleintelligence.com/features/medical-coding-is-the-next-stop-for-artificial-intelligence-in-healthcare
16. Streamline Health. (2024, March 29). Computer-Assisted Coding: What is it and Does it Work? https://streamlinehealth.net/computer-assisted-coding/
17. XpertDox. (2024, March 28). Securing Patient Data in AI Medical Coding - White Paper. https://www.xpertdox.com/whitepaper/data-security-in-ai-medical-coding/