Preventing SUT non-compliance in SGK hospital billing

To prevent non-compliance with SUT, the reimbursement rulebook of Türkiye's Social Security Institution (SGK), hospitals need to check each file when a service is recorded, not at the end of the billing period. The check compares the discharge summary, ICD-10 codes, billed procedures and the reports and documents SUT requires, and gaps are closed before the invoice is issued. Deduction reasons are regularly traced to their root causes, and SUT amendments are built into the checking rules. Opinion AI's hospital platform TIS runs these checks without replacing the hospital information system (HBYS); the rule engine makes the SUT decision and people give final approval.

Last updated: 26 September 2026

What SUT non-compliance is and why it shows up at billing

SUT (Sağlık Uygulama Tebliği, the Healthcare Implementation Communiqué) is published by SGK and sets the conditions, documents and prices under which healthcare services covered by General Health Insurance are paid. Medula is SGK's pre-authorization and billing system. For SGK-covered visits, the hospital obtains pre-authorization through Medula, records the services it delivers in the system and invoices SGK in billing periods.

SUT non-compliance means that a billed service does not meet the conditions these rules require. Medula checks some rules at the moment of entry and rejects faulty records. But conditions that depend on clinical content, such as whether a procedure is supported by a diagnosis, whether it is described in the discharge summary, or what a report actually says, are not always caught by these automated checks. Such mismatches often come back to the hospital as deductions or rejections when the invoice is reviewed, at the point where correcting them is hardest.

The scale of the problem is visible across the health system as a whole: 5–10% of claims are denied, and part of these denials stems from missing documents and the documentation burden. Physicians already spend 35% of their working time on documentation. The answer is therefore not more forms for physicians, but a systematic comparison of the clinical documents already written against the invoice.

Common causes of non-compliance

Non-compliance usually falls under one of the following headings.

The diagnosis does not support the procedure. The diagnosis that justifies a billed procedure is not coded, or the discharge summary contains no statement supporting it.
The discharge summary and the invoice do not match. A procedure on the invoice does not appear in the discharge summary or the order record; conversely, a procedure described in the discharge summary never makes it onto the invoice.
A required report or document is missing. For some drugs, medical devices and procedures, SUT requires a specialist physician report, a health board report or other documents; what the report says matters as much as whether it exists.
The physician or specialty condition is not met. Some drugs and procedures are paid only when ordered or performed by physicians in specific specialties.
Package, combined-billing and frequency rules are missed. A service included in a package price is billed separately, or a test limited to certain intervals is repeated before the interval has passed.
Dates and visit type are inconsistent. A procedure date falls outside the inpatient stay, or an inpatient procedure is entered on an outpatient visit.
A rule change has not reached the workflow. SUT is amended frequently through amending communiqués, and entries made the old way fall foul of the new rule.

Steps to take before the invoice is issued

The following steps do not depend on the software in use.

Do not leave checks to the end of the period. Check the file when a service is recorded and when the patient is discharged. Once files pile up at the end of the period, completing a missing report becomes harder.
Link every invoice line to the clinical record. Confirm that every billed procedure has a counterpart in the discharge summary, operative note or order record. Run the comparison in the other direction too, and look for procedures described in the discharge summary but missing from the invoice.
Base diagnosis coding on the text. Code the principal and secondary diagnoses in ICD-10 according to the statements in the discharge summary. Make sure the diagnosis that justifies each billed procedure is coded.
Flag services that need documents at the time of the order. Keep a checklist of drugs, devices and procedures that require a report or other document. Complete missing documents while the service is delivered or before the patient is discharged.
Trace deductions and rejections to their root causes. Record every deduction with its reason and classify it, for example as a diagnosis–procedure mismatch, missing document, physician authorization, package or frequency issue. Share the results regularly with the relevant teams by specialty and department.
Build SUT amendments into the checking rules. After each amending communiqué, identify the affected procedures, drugs and devices, update the checking rules and inform the relevant teams.
Bring teams together on the same record. Have physicians, coders, billing and medical accounting teams work through questions on the same file and the same rationale.
Monitor audit indicators continuously. Track indicators every day, not only when an audit arrives; a deviation seen early can be corrected before files pile up.

The role of the rule engine and AI

SUT checking consists of two different jobs. The first is applying the rule: a service either meets the condition the rule requires or it does not. This assessment should be made in a rule engine that always returns the same result for the same input and shows the rule it relied on, in other words one that can be audited.

The second is reading the clinical content. Discharge summaries and reports are free text; the same finding appears in different words, abbreviations or Latin terms depending on the physician who wrote it. A rule engine cannot read this text on its own. This is where AI comes in: it splits the discharge summary into structured fields, extracts diagnoses and procedures, suggests ICD-10 codes, flags inconsistencies between the discharge summary and the invoice, and explains the rule engine's result with its rationale.

When a general-purpose language model interprets SUT from memory, it can miss recent amendments or describe a rule that does not exist in fluent language. Every warning should therefore show the rule it rests on and the relevant sentence in the document, what the document states should be kept separate from what the system infers, and the rule set should be updatable after SUT amendments. The same division of work applies wherever a payer publishes detailed reimbursement rules; outside Türkiye, the local rulebook takes the place of SUT. We cover these criteria in more detail in the guide on choosing a clinical decision support system.

What Opinion AI does in this process

TIS, Opinion AI's hospital platform, does not replace the hospital's HBYS; it is added on top as an intelligent layer. HBYS and Medula data come together in one live patient profile called the patient twin. Discharge summaries and reports become structured fields in this profile, and the clinical summary, risk flags and SUT checks appear in the same view.

Our clinical model MINA, working as TIS's clinical assistant, reads the discharge summary, suggests ICD-10 codes, checks SUT compliance and shows its rationale and source in every answer. What the document states is kept separate from what the model infers. The rule engine decides whether a service complies with SUT; AI explains the result and the relevant rule. Medula billing errors and SUT mismatches are flagged before the invoice is issued, and collected, pending and rejected amounts are tracked per application. The audit cockpit also monitors BH and ADSH indicators continuously and flags those carrying risk for management.

MINA is built on an open-weight foundation model adapted to Turkish and the clinical language of 14 specialties through continued pre-training; specialty, institution and task expertise is added through LoRA adapters, and knowledge retrieval runs through agentic orchestration over a clinical knowledge graph (GraphRAG). In MINA's latest release evaluation, model results on hospital tasks were 87% for ICD-10 coding, 89% for discharge summary field extraction and 91% for SUT decision accuracy (MINA v3.7 · 8 protocols · ~9,300 cases).

Data stays inside the hospital. TIS is installed on the hospital's own servers or in an isolated cloud environment dedicated to the hospital; personal data is masked in line with KVKK, Türkiye's personal data protection law, and the hospital is the data controller. Opinion AI does not collect personal data. An ICD-10 suggestion is not final until the physician approves it, and the result of the SUT check is presented to the relevant team for review. Our blog covers structuring the discharge summary and the division of work between the rule engine and AI in more detail.

Frequently asked questions

What is SUT non-compliance?

SUT non-compliance means that a healthcare service billed to SGK, Türkiye's Social Security Institution, does not meet the conditions set in SUT, the Healthcare Implementation Communiqué that defines reimbursement rules. Typical examples are a diagnosis that does not support the procedure, a missing required report, an unmet physician or specialty condition, and a breached package or frequency rule. Such mismatches often come back to the hospital as deductions or rejections when the invoice is reviewed.

How can hospitals reduce SGK billing deductions?

The key step is to move checks from the end of the billing period to the moment a service is recorded. Each invoice line should be compared in both directions with the discharge summary and the order record, required reports should be completed while the service is delivered, and SUT amendments should be built into the checking rules. When deductions are classified by reason and fed back to departments, the same errors are less likely to recur.

Why should the discharge summary and the invoice match?

The invoice is the financial counterpart of the care described in the discharge summary and other clinical records, so a billed procedure with no basis in the clinical record creates a deduction risk. The reverse is also a problem: a procedure described in the discharge summary but missing from the invoice is lost revenue for the hospital. The comparison should therefore run in both directions.

Can AI decide whether a service complies with SUT?

It should not. SUT compliance should be assessed in a rule engine that always returns the same result for the same input and shows the rule it relied on. AI reads the free-text discharge summary, splits it into structured fields, suggests ICD-10 codes and explains the rule engine's result with its rationale; final approval rests with the physician and the relevant team.

Can a general-purpose AI chatbot give correct SUT rules?

It can give a general idea, but it should not be the basis of a billing decision. SUT is amended frequently; a general-purpose model may not know the latest changes and can describe a rule that does not exist in fluent language. A check used in billing should be tied to a current rule set and show, for every warning, which rule and which statement in the document it rests on.

Does patient data have to leave the hospital for SUT checks?

Not for the internal check before billing; submissions to Medula are a separate process. Under KVKK, Türkiye's personal data protection law, health data is special category personal data, as it is under GDPR, and a checking system can run on the hospital's own servers or in an isolated cloud environment dedicated to the hospital. Opinion AI's hospital platform TIS is deployed this way: data stays in the hospital, passes through KVKK-compliant masking, and the hospital is the data controller.

To assess SUT checking on samples of your hospital's own visits and invoices, starting with a single module, use the POC Request form. We cover rejections at the pre-authorization stage in the guide on reducing denials in health insurance pre-authorization, and how the discharge summary becomes a single patient profile in the guide on the patient digital twin.