Medical coding is a crucial behind-the-scenes process in healthcare, translating clinical diagnoses and procedures into standardized codes for billing, compliance reporting, and analytics. In the United States, the ICD-10-CM, or the International Classification of Diseases, 10th Revision, Clinical Modification system, is used for diagnosis coding and contains tens of thousands of codes to represent every possible health condition. This level of detail brings significant complexity. Coders must account for ICD-10 laterality (whether a condition affects the left side, right side, or both), code to the highest level of specificity (avoiding vague “unspecified” codes), and know when ICD-10 combination codes can capture multiple diagnoses or a complication in one go. Ensuring accuracy in these areas is critical because coding errors can lead to denied insurance claims, compliance issues, or improper patient records.

Traditionally, human coders pore over doctors’ notes to assign codes, but the process is labor-intensive and error-prone. Even well-trained professionals can miss subtle details like which limb was affected or whether a complication was documented, especially under time pressure. The good news is that advances in AI in medical coding are starting to ease this burden. Hospitals and clinics have begun adopting medical coding automation tools that use artificial intelligence to assist with or even automate parts of the coding process. This is part of a broader trend of applying medical artificial intelligence to administrative and billing tasks in healthcare.

AI in medical billing and coding technology can analyze clinical documentation much faster than a person, ensuring the correct codes are captured. For example, innovators like Sully.ai are leveraging AI to address these coding challenges by integrating smart coding assistants into the billing workflow. In this article, we will explore how AI handles diagnosis coding for laterality, specificity, and combination codes and how this AI medical technology is improving accuracy and efficiency in the U.S. healthcare system.

Laterality in ICD-10 Coding

One unique feature of ICD-10-CM is that many diagnosis codes include built-in indicators of laterality. ICD-10 laterality coding ensures that diagnoses precisely reflect which side of the body is affected. For example, there are different codes for a fracture of the left arm versus the right arm, or for an infection in the left ear versus the right ear. If a condition is bilateral and there isn’t a specific bilateral code available, coders should assign both the left and right codes, and if documentation doesn’t specify the side, an “unspecified” side code is used. This level of detail is intended to improve accuracy and patient care, but it also means coders must pay close attention to physician documentation.

Mistakes in laterality coding are a common source of errors and claim denials. If the documentation says a procedure was done on the patient’s left side, but the diagnosis code submitted is for the right side, the claim will likely be rejected by the insurer as inconsistent. Payers and auditors in the U.S. have implemented edits to catch such discrepancies. For instance, if a claim’s procedure code is marked for the right side but the diagnosis code is for a left-side condition, it will be denied as a coding error. Coders must locate any mention of laterality in the clinical notes and ensure the correct code is chosen. When the side isn’t documented, using an unspecified code may technically meet coding guidelines, but many insurance companies now discourage or even deny payment for unspecified laterality if specificity could have been obtained. This puts pressure on providers to document laterality every time.

Specificity and the Push for More Detailed Codes

Beyond laterality, coding to the highest specificity is a guiding principle of ICD-10. The coding system dramatically expanded the number of diagnosis codes (to over 70,000) compared to its predecessor, ICD-9, largely to encourage more precise descriptions of a patient’s condition. Specificity can include details such as the exact location of a condition, severity, the type or cause of a disease, and any associated complications. An example is the range of codes for something like diabetes. In ICD-10, a coder must specify whether the diabetes is Type 1 or Type 2, whether it is controlled or uncontrolled, and whether complications are present. If a patient has Type 2 diabetes with diabetic neuropathy, there is a specific code for that scenario. Using just the code for “Type 2 diabetes” without noting the neuropathy would be considered incomplete coding because it lacks specificity.

Under ICD-10-CM official guidelines, providers and coders are expected to avoid unspecified codes whenever the clinical information allows for a more specific code. Medical coding AI and health information management teams often track the use of unspecified “.9” codes as a quality metric. The reason is not only clinical accuracy but also financial: coding to a higher level of detail can impact reimbursement and billing integrity. Payers have grown increasingly stringent about specificity. Many have implemented policies to deny claims that are not coded to the full level of detail documented. Providers who continue using deficient “unspecified” codes face a higher risk of denied claims, longer reimbursement cycles, and even potential revenue loss, since specificity is now the name of the game under ICD-10.

Of course, there are situations where an unspecified code is appropriate. When a diagnosis is truly unknown or not yet fully worked up, the general push is to limit those cases. Coders often have to query physicians for more information if a note is ambiguous. This back-and-forth slows down the billing process. It’s another area that highlights the need for tools or systems to help catch missing details or prompt for clarifications.

ICD-10 Combination Codes: Capturing Multiple Conditions in One Code

In addition to laterality and detailed specificity, ICD-10 introduced the concept of ICD-10 combination codes for certain common combinations of diagnoses and manifestations. A combination code is a single code that represents two diagnoses all at once. One code “combines” multiple aspects of a patient’s condition. This is meant to simplify coding and ensure related conditions are linked. Without a combination code, a coder might otherwise assign one code for the diabetes and a separate code for the neuropathy. But since a special combined code exists, the guidelines direct the coder to use that one code as it fully captures the diagnostic picture.

The Challenges of Complex Coding for Humans

Modern diagnosis coding demands extreme precision under constant pressure. Human coders must interpret dense clinical narratives, track fine-grained details, and apply thousands of evolving rules correctly every day. The challenge is not a lack of skill or training, but the sheer scale and cognitive load involved. Below are distinct reasons complex coding strains human capacity:

• An Overwhelming Volume of Codes: Coders cannot rely on memory alone and must constantly reference electronic codebooks, search tools, and internal guidelines. This process slows decision-making and increases reliance on imperfect keyword searches. When relevant options are buried among near-identical entries, the risk of selecting an imprecise or incomplete code rises, even for experienced professionals working carefully.

• Hidden Specificity in Clinical Documentation: Physician notes often contain critical coding details embedded deep within long narratives. Information about severity, acuity, or causal relationships may appear once and never be repeated. Coders must read attentively while interpreting clinical intent, which becomes harder when documentation is inconsistent or loosely structured. A single overlooked phrase can shift the correct code entirely. Under time pressure, this creates a structural disadvantage where specificity is sacrificed simply because it is difficult to extract reliably.

• Laterality and Modifier Oversight: Many diagnoses and procedures require exact indicators for body side, encounter type, or additional modifiers. These elements are easy to miss when documentation references them indirectly or inconsistently. Forgetting a laterality marker or modifier does not reflect poor knowledge, but rather how easily small technical requirements slip through during repetitive work. Unfortunately, payers treat these omissions strictly, rejecting claims even when the core diagnosis is otherwise correct.

• Productivity Pressure and Cognitive Fatigue: High daily record volumes force coders to balance speed with accuracy. Sustained concentration across dozens of charts increases mental fatigue, which directly affects attention to detail. As fatigue builds, coders may default to broader codes or avoid time-consuming clarifications. These decisions are often made to keep workflows moving, not due to carelessness.

As coding complexity continues to expand, the limitations of manual workflows become more visible. Technology is not replacing coders. It is addressing structural challenges that human effort alone cannot consistently or efficiently solve.

How AI and Machine Learning Are Transforming Medical Coding

The application of artificial intelligence in medical coding builds on advances in natural language processing (NLP) and machine learning. The goal is to have software that can “read” a doctor’s notes or a discharge summary the way a human would, understand the medical context, and then either suggest the correct codes or automatically assign them. An AI medical coder is trained to do what a human coder does: identify the diagnoses and key details in the text and map those to the appropriate ICD-10 codes. This is no small feat, because medical language is complex, full of synonyms and abbreviations, and patient records can be long and unstructured. However, recent progress in machine learning medical coding algorithms has made automated coding a rapidly improving reality.

Modern AI coding systems typically work like this: First, the clinical documentation is fed into an NLP engine. The NLP component uses techniques like named entity recognition to spot diagnoses, symptoms, and other clinical entities in the text. It also looks at contextual cues. Then, based on these recognized entities and context, the system references a knowledge base of ICD-10 codes to find the best matches. Advanced machine learning models have been trained on large datasets of past coded records, so they have learned patterns. These AI models can effectively interpret medical terminology and assign appropriate codes with a speed and accuracy that surpasses manual methods. It’s important to clarify that AI isn’t replacing human coders outright. Instead, it’s augmenting them. Many healthcare providers are implementing medical coding automation tools, where the AI suggests codes and the human coder reviews and validates them. This setup combines the consistency and speed of AI with the critical thinking and oversight of humans.

Industry Adoption and AI Coding Solutions

Given the clear advantages, it’s no surprise that adoption of AI in coding is growing. Many top AI healthcare companies, as well as specialized startups, have jumped into this space, offering products that range from assisted coding software to fully autonomous coding systems. Hospitals, large physician groups, and billing companies across the U.S. are piloting or implementing these solutions to improve their revenue cycle performance. Various AI medical coding companies offer different flavors of technology. Some focus on inpatient hospital coding, others on outpatient clinics or specific specialties, but all share the goal of reducing manual coding effort and errors.

One driving factor in adoption is the return on investment. Coding-related errors and inefficiencies cost healthcare providers money, whether through denied claims, delayed payments, or the cost of rework and compliance issues. By deploying AI coding tools, organizations aim to recoup that lost revenue and reduce overhead. Additionally, there is a workforce issue: experienced medical coders are in short supply, and turnover can be high due to the job's stressful nature. Automation helps alleviate some pressure on coding staff and can fill in gaps where you might not have enough coders for the volume of charts. It can also help standardize coding across an organization, which is useful for health systems with multiple sites and different coders. The AI will apply the same logic everywhere.

From the regulatory standpoint, there’s growing acceptance of AI-assisted coding. The American Health Information Management Association (AHIMA) and the American Academy of Professional Coders (AAPC) have been educating their members about AI and how it can be a tool to enhance accuracy. There’s an understanding that the coder of the future might be less of a data-entry specialist and more of a reviewer or auditor of AI-suggested codes, focusing on edge cases and ensuring the machine is doing its job correctly. In the U.S., we’re also seeing artificial intelligence medical billing applications that integrate coding AI with claims submission, meaning once the AI assigns codes, the system can automatically populate claim forms and check for any billing rule violations before sending them off to payers. This end-to-end automation can dramatically speed up the billing cycle.

Laterality, specificity, and combination codes no longer have to be the thorny coding issues they once were. With the support of cutting-edge AI coding systems, healthcare providers can code diagnoses more accurately and efficiently than ever before. The collaboration between human coders and artificial intelligence is yielding a more error-resistant coding process. As technology continues to advance, we can look forward to a future where automated ICD-10 coding tools handle the heavy lifting of coding intricacies, while humans ensure that compassionate, contextual care is delivered for every patient.

Sources

• EmblemHealth – Correct Laterality ICD-10-CM Diagnosis Coding Policy

• Wolters Kluwer – ICD-10 unspecified codes: Financial risks and solutions