Medical coding is one of those things clinicians learn just enough about in training and spend the rest of their careers catching up on. It's the quiet engine behind every claim submitted, dollar collected, and piece of diagnostic data that flows through the healthcare system.

If you've ever wondered why a "headache" on a chart isn't treated the same as an "intractable migraine with complications" by a payer, this is your guide.

This article runs through:

• What medical coding is
• Types of medical codes and their definitions
• The importance of coding accuracy
• Medical coding challenges

What is medical coding?

Medical coding is the process of translating clinical documentation — diagnoses, procedures, symptoms, and encounters — into standardized alphanumeric codes. Those codes are what insurance companies, government payers, and health systems use to process claims, track outcomes, and reimburse care.

Every visit you document generates a set of codes. Those codes determine how much you get paid, whether a claim is approved or denied, and how your patient's care gets categorized for reporting and research. This standardized language — mandated under HIPAA — is what allows hospitals, providers, and payers to communicate consistently across the entire healthcare system. Nearly all private health information rests on the codes being assigned correctly.

The American Health Information Management Association (AHIMA) puts it well: medical coders are "the connection between health information and the success stories that come from great patient care."

In short, if clinical documentation is the story, medical codes are how the rest of the world reads it.

Types of medical codes

Most clinical encounters draw from three main coding systems.

ICD-10-CM codes

International Classification of Diseases, 10th Revision, Clinical Modification covers diagnosis codes.

These are the codes that describe what's wrong — from a sprained ankle to Type 2 diabetes with complications. The ICD-10-CM system contains nearly 75,000 codes, and the FY 2026 guidelines from the Centers for Medicare & Medicaid Services (CMS) are unambiguous about what's expected: providers should "code to the highest level of specificity when supported by the medical record documentation."

A general code will do in a pinch. However, it can leave earned money on the table. A specific code tells a more accurate clinical story, and captures the full finances earned.

CPT codes

Current Procedural Terminology describe what was done during a visit — procedures, tests, screenings, injections, and more. These are maintained by the American Medical Association (AMA) and updated annually.

CPT includes more than 11,000 codes and covers most outpatient services.

E/M codes

Evaluation and Management codes) capture the complexity and nature of the clinical encounter itself — from a brief follow-up (99213) to a complex, high-acuity visit (99215).

E/M coding is where a lot of revenue gets left on the table, often not through fraud or negligence, but through under-documentation of the medical decision-making that actually happened.

Note: In some settings, HCPCS Level II codes (Healthcare Common Procedure Coding System) round out the picture, covering supplies, equipment, and certain services not captured by CPT.

Why does coding accuracy matter?

Accurate coding matters both clinically and financially. It impacts billing, data quality, compliance, and increasingly, patient care.

The CMS FY 2026 ICD-10-CM Coding Guidelines are direct on this point.

"The importance of consistent, complete documentation in the medical record cannot be overemphasized. Without such documentation accurate coding cannot be achieved. Under HIPAA, adherence to ICD-10-CM coding guidelines is required across all healthcare settings. A joint effort between the provider and the coder, CMS notes, "is essential to achieve complete and accurate documentation, code assignment, and reporting of diagnoses and procedures." — CMS FY 2026 ICD-10-CM Coding Guidelines

On the financial side, the gap between a general diagnosis code and a specific one can mean the difference between a claim that clears on the first pass and one that gets denied, downcoded, or audited. For practices managing hundreds or thousands of visits a month, that gap compounds quickly.

Beyond reimbursement, coding data feeds into everything from public health surveillance to value-based care contracts to risk adjustment. The codes you assign today shape what your panel looks like to a payer tomorrow.

What are major coding challenges?

The truth is, most clinicians weren't trained to code. Medical school and residency programs teach diagnosis and treatment — not ICD-10 specificity or CPT hierarchies. And yet, accurate coding is now inseparable from financial sustainability, especially for independent and small practices.

"Unfortunately, in training we just don't learn coding. But this is the lifeline of everyone's livelihood." — Dr. Yun Boylston, Burlington Pediatrics

Even professional coders face structural time pressure. Most work under tight production schedules, expected to keep "lag days" (the time between documentation and claim submission) to just two to five days. Nuance can easily get lost.

A 2026 study published in JMIR Medical Informatics examined this gap.

Researchers found that manual ICD-10 coding of free-text clinical notes is "both time-consuming and resource-intensive," and that even well-trained coders working with routine primary care documentation face real limitations. Notes are short, loosely structured, and disjointed.

The study demonstrated that AI-powered language models can classify clinical text into ICD-10 categories with F1 scores of 0.85–0.86, underscoring both the potential and the practicality of automated coding support in real-world clinical settings.

This research reflects what clinicians already know intuitively. Under time pressure, it's easy to reach for a familiar general code rather than pause to find the most accurate one. It's not a failure of effort. It's a structural gap that no amount of training alone can fully close.

How is AI changing medical coding?

The emergence of large language models trained on clinical text has created a meaningful option: AI that can surface relevant codes directly from the clinical encounter, pulling the nuance and complexity directly from the source conversation.

The JMIR research team demonstrated that a fine-tuned language model trained on tens of thousands of real primary care notes could "reduce the burden of repetitive and error-prone manual handling" while maintaining strong performance across a diverse set of diagnostic categories.

Their conclusion: AI-assisted coding has real potential "to streamline and scale the extraction of diagnostic information for practical applications."

This means that AI can can save clinicians from elying on memory or manual lookup. AI coding tools can surface the codes the visit actually supports — including ones that might otherwise go uncaptured.

AI coding tools in practice

One example of AI here is Freed's coding assistant. The AI scribe incorporates this coding and billing support directly into the note-taking flow.

After a visit, Freed analyzes the transcript and note to suggest:

• ICD-10 diagnosis codes
• CPT procedure codes
• E/M coding recommendations

All suggestions appear in the Freed workspace, with supporting rationale so clinicians can review and select.

The results from practices already using it are telling.

Burlington Pediatrics — an independent North Carolina practice with 15 providers and more than 52,000 visits per year — had already built a tight billing operation: same-day notes, 24-hour claim turnaround, denial rates under 1%. Their system worked. But even that wasn't enough to catch everything.

"You don't know what you don't know, and who has time to go looking for it?" — Dr. Yun Boylston, Burlington Pediatrics

In six weeks with Freed's coding assistant, Burlington saw:

• 1,500+ ICD-10 code improvements
• 71 confirmed E/M upgrades (including shifts from 99213 to 99214, worth ~$40 per visit)
• 83% provider acceptance rate on suggestions actively reviewed
• 94% of providers actively using the tool by month two

The difference showed up in specificity. Right shoulder pain became right shoulder pain and injury incurred on the playground. A headache with prompted context became an intractable migraine — documented accurately, coded correctly, and reimbursed appropriately.

Freed also tightened coding consistency across the team. A newer provider whose E/M split had been running 60/40 (L3/L4) shifted to 40/60 within six weeks — aligning with the practice's more experienced physicians. Real-time support, not retrospective correction.

Freed's coding assistant is included at no additional cost for eligible plans (Premier and Group), and works across practices of all sizes — from solo clinicians to growing organizations.

Coding for independent practices

Medical coding is complex. The guidelines are long and the code sets change every year. Most clinicians are already running at capacity before they get to any of it.

But you don't have to navigate it alone. Freed's coding assistant surfaces what your documentation already supports — so the revenue you've earned doesn't quietly slip away.

Try Freed's coding assistant free or talk to our team to see how it fits your practice.