AI Diagnosis Code: How Diagnosis Coding Works in Veterinary Medical Records
A diagnosis code makes the veterinarian’s assessment searchable and comparable. The code is used for statistical purposes, in communications with insurance companies, and when information needs to be reported to other parties. AI can suggest diagnosis codes based on the medical record, but it is always the veterinarian who determines the diagnosis and approves the code.
This article explains why diagnosis codes matter, how the Swedish diagnosis registry Pyramidion is structured, and how a clinic can use AI-generated code suggestions without compromising the quality of patient records.
Why Diagnosis Codes Matter
According to the Swedish Board of Agriculture, animal health professionals must maintain a medical record for each patient visit or consultation, and the record must include the diagnosis and any differential diagnoses for each consultation. The record must be written immediately following the visit and retained for at least five years from the date of the last entry. These requirements are summarized on the Swedish Board of Agriculture’s page on record-keeping, issuing certificates, confidentiality, delegation, and reporting.
A coded diagnosis does more than just meet medical record requirements. It facilitates communication between veterinarians, makes it easier for pet owners to deal with their insurance companies, and provides a basis for statistics and research. An incorrect or overly general code, on the other hand, follows the patient and can give a false impression of what was actually diagnosed.
Pyramidion – the Swedish Diagnosis Registry
Pyramidion is a diagnostic registry owned by Gröna arbetsgivare. The registry has a hierarchical structure: diagnoses are recorded in categories grouped by level, much like a family tree. By building longer diagnostic chains, veterinarians can record more precise diagnoses than in the previous diagnostic registry.
The structure is flexible, which means the registry can accommodate more diagnosis codes. New diseases can be added as needed, and terminology can be updated as research progresses. Since all clinics that use Pyramidion record data in the same way, it is also possible to compile statistics that can be used in research, both from Sweden and from other countries that use the system.
Pyramidion is designed to work with all standard medical record systems, and the cost of the license is covered by the medical record system provider. More information is available on Gröna’s page about the Pyramidion Diagnosis Registry.
Diagnosis registries and reporting lists are not the same thing
In addition to the diagnosis registry in the medical record, there are lists used for reporting. Veterinarians must submit information about certain drug treatments to the Swedish Board of Agriculture no later than the 15th of the month following the month in which the treatment was prescribed. For this reporting, the Swedish Board of Agriculture’s own lists of treatment reasons (diagnoses), animal categories, and age categories are used. The medical record systems connected to the Swedish Board of Agriculture are configured to use these lists during data transfer.
For the clinic, this means that a single assessment may need to be expressed in more than one way, depending on the purpose. Anyone using AI-generated code suggestions therefore needs to know which coding system the suggestion applies to and where the code will be used.
What does "AI diagnosis code" mean?
An “AI diagnosis code” refers to a situation in which an AI tool reads the medical record and suggests diagnosis codes that match what has been documented. The suggestion is based on the text in the medical record—that is, on the veterinarian’s own assessment—and does not replace it.
This is a different function from AI diagnostic support. Diagnostic support assists the veterinarian in the actual assessment, for example by suggesting differential diagnoses during or after the consultation. Diagnosis coding comes afterward: once the veterinarian has made their assessment, the AI helps link it to the correct code. Diagnostic support is described in more detail in “AI and Diagnostic Support.”
Diagnosis Codes in Vetz
Vetz suggests relevant diagnosis codes based on the medical record notes. This saves the veterinarian from having to look up the code themselves, which saves time and ensures more consistent coding among colleagues and across visits.
The coding suggestions are part of the same workflow as medical record-keeping. Vetz listens during the conversation with the pet owner and creates a structured draft of the medical record; the veterinarian reviews and approves the text; and the coding suggestions are based on the approved notes. How the medical record workflow functions is described in “AI-Based Medical Record-Keeping for Veterinarians.”
Before the clinic begins using the coding suggestions, it should check with Vetz to see how the suggestions fit with the coding system used by the clinic’s medical record system. An overview of all features is available on Vetz’s features page.
How to Review a Proposed Diagnosis Code
A proposed code should be treated as a draft until the veterinarian has approved it. The following checks may be included in the clinic’s routine:
- Does the code match the confirmed diagnosis? The code should reflect what the veterinarian actually observed, not what the pet owner suspected.
- Is it a diagnosis or a suspicion? A differential diagnosis or a symptom should not be coded as a confirmed diagnosis.
- Is the code specific enough? In a hierarchical registry such as Pyramidion, a more precise code can often be selected further down the chain.
- Does the code correspond to the medical record text? The diagnosis, findings, and treatment described in the note should support the selected code.
- Does the code apply to the correct purpose? A code for the medical record and a reason for treatment for reporting to the Swedish Board of Agriculture are not always the same thing.
AI can generate a suggestion with a high degree of certainty even when the supporting information is limited. If the medical record text is vague, the coding suggestion will also be uncertain, and in that case, the medical record entry itself needs to be clarified first. Common shortcomings in the medical record text itself are described in “The Most Common Mistakes in Medical Record Writing.”
Questions Regarding the Implementation
Before the clinic begins using AI-generated suggestions for diagnosis codes, it is a good idea to have answers to the following questions:
- What coding system does the clinic’s medical record system use, and how do the AI suggestions relate to it?
- How is an approved code transferred to the medical records system?
- Who is responsible for reviewing the code, and how is a piece of code handled if it turns out to be incorrect?
- How does the clinic ensure that coding becomes more consistent—and not just faster?
A good way to evaluate the feature is to have a few veterinarians use the coding suggestions for a period of time and compare them with how they would have coded the cases themselves. This will reveal both the time saved and any patterns in incorrect suggestions.
You can try Vetz for free for 14 days. Clinics that want to see how the coding suggestions work within their own workflow can schedule a demo.