Reducing denials in health insurance pre-authorization

The main causes of pre-authorization denials

In Turkish health insurance, provizyon is the insurer's approval to cover the cost of a treatment under the policy terms; ön provizyon, or pre-authorization, is that approval obtained before a planned procedure. This guide uses "pre-authorization" for both. Across the health system, claim denials run at 5–10%, and some of these denials stem from missing documents and the documentation burden.

The specific reasons differ by institution and by policy, but the ones that recur in practice fall under five headings:

• Missing documents prevent the request from being assessed. The report, imaging or lab result, consultation note or discharge summary that supports the indication may not have been attached.
• Incorrect or inconsistent coding means the ICD-10 diagnosis code does not match the diagnosis in the clinical document, or the procedure code does not match the planned procedure. Even when the document itself is correct, a code mismatch can send the request back.
• Insufficient medical necessity means the document does not clearly show why the procedure is needed. The specialist assessing it then cannot build the rationale on their own.
• Coverage and limit mismatches occur when the procedure falls outside the policy's coverage, the waiting period has not ended, an exclusion applies or the remaining limit does not cover the request.
• Non-compliance with reimbursement rules means the request conflicts with the rules that apply. In Türkiye the main rulebook for public reimbursement is the SUT, the Healthcare Implementation Communiqué published by SGK, the Social Security Institution; private insurers may also check SUT compliance in their assessments.

Some denials reflect a genuine lack of coverage and are the right decision. The goal is not to reduce correct denials but to eliminate the avoidable ones that come from an incomplete or incorrectly prepared file. That is why the distribution of denial reasons should be tracked alongside the denial rate.

What hospitals can do before sending a request

In a hospital, the way to reduce denial risk is to put the checks at the start of the process. The following steps can be added to the existing workflow:

1. Check the policy terms before preparing the request; note the procedure's coverage, the waiting period, any exclusions and the remaining limit.
2. Match diagnosis and procedure codes to the clinical documents; confirm that the ICD-10 code matches the diagnosis in the discharge summary or report, and that the procedure code matches the planned procedure.
3. Write the medical necessity into the document itself; attach the findings that support the indication, the relevant test results and any treatments tried earlier.
4. Use a document checklist for each procedure type; define in advance which documents are required for request types such as planned surgery, advanced imaging or admission.
5. Where SGK covers part of the cost, move SUT and Medula checks ahead of billing. Medula is SGK's pre-authorization and billing system, and an error missed there can come back at the payment stage.
6. Record every denial with its reason and classify denials regularly; see which reasons recur by specialty and procedure type, and update the checklist accordingly.
7. When a request for missing documents arrives, add the requested document and resubmit on the same file.

Consistent, reasoned assessment on the insurer's side

The denial rate does not depend on hospital preparation alone. When requests of the same type are assessed by different specialists against different criteria, unnecessary requests for additional documents and appeals both increase. These steps make the process predictable:

1. When a request arrives, bring the policy analysis, limit usage, claim history and clinical documents together in a single view for the specialist.
2. Tie routine checks such as coverage, waiting periods, limits and rule compliance to explicitly defined rules, so that their results are ready when the specialist opens the file.
3. Apply the same set of checks to requests of the same type, so that two specialists looking at the same file see the same information.
4. Communicate a denial or a request for additional documents with its rationale and with the missing document named, so the hospital can complete only what is missing.
5. Reserve specialists' time for complex cases that require a medical necessity assessment, and leave the final decision with the insurer's medical specialist in every case.

The role of AI in pre-authorization

AI can help with the two most time-consuming parts of pre-authorization: reading clinical documents written as free text, and comparing them with policy terms, codes and reimbursement rules. Diagnoses, procedures and findings in discharge summaries, reports and test results can be split into structured fields, and a mismatch between code and document, or a missing document, can be flagged before the request is sent. By checking the requested procedure against current clinical guidelines, AI can also prepare a preliminary assessment of medical necessity.

The limits of this role should be just as clear. A pre-authorization decision directly affects a patient's access to care, so in a sound setup the decision rules run in an auditable rule engine, AI explains the finding and its rationale, and a specialist makes the decision. When evaluating a solution, check whether its recommendations show their rationale and source, whether it separates what the document says from what the model infers, and whether its outputs can be audited later. Health data is special category personal data under both KVKK, Türkiye's personal data protection law, and the GDPR, so whether the system runs inside the institution's own environment and whether the institution remains the data controller are also core criteria.

Opinion AI's role in pre-authorization

Opinion AI offers separate platforms for the two sides of pre-authorization. On the hospital side, TIS is added as an intelligent layer on top of the existing HBYS, the hospital information system. Pre-authorization requests are gathered in a single panel, and each request carries a risk score. Operators do not start a file from scratch; they begin with the clinical synthesis prepared by MINA. Missing documents and incorrect ICD-10 codes become visible before the request reaches the other side, and Medula billing errors and SUT non-compliance are flagged before the invoice is issued.

On the insurer's side, SIT-I opens the policy coverage and the clinical documents attached to the request on the same screen at pre-authorization. Policy analysis, limit tracking, claim history, SUT compliance and clinical context reach the specialist ready for review, and the specialist can put a clinical question to the system through the Doctor Assistant and Council Mode. Decision rules run in the rule engine; AI only explains the result and its rationale, and shows what the document says separately from its own inference. The flow is described in detail in a ready file for the medical reviewer at pre-authorization.

The two sides run as separate deployments, and on both sides the data stays inside the institution. What they share is the same clinical intelligence and the same regulations and rules, so both sides assess a case on a shared basis. That clinical intelligence is MINA. Its open-weight base model has been adapted to Turkish and the clinical language of 14 specialties through continued pre-training (CPT) and specialized by specialty, institution and task with LoRA adapters; knowledge retrieval runs through agentic orchestration over a clinical knowledge graph (GraphRAG).

People make the decision; MINA shows the rationale and source of its recommendation, and every output can be audited. Opinion AI does not collect personal data; the institution is the data controller, and the data the model sees passes through KVKK-compliant masking. Our work with insurers and hospital groups is described on the Success Stories page.

Frequently asked questions

Why are health insurance pre-authorization requests denied?

The main causes of pre-authorization denials are missing documents, diagnosis or procedure codes that do not match the clinical documents, medical necessity that is not stated clearly enough, and procedures that fall outside the policy's coverage, waiting period, exclusion or limit terms. In Türkiye, when SGK covers part of the cost, non-compliance with SUT reimbursement rules can also lead to denial. Some denials are correct coverage decisions; the avoidable ones come from an incomplete or incorrectly prepared file.

How can a hospital reduce its pre-authorization denial rate?

A hospital reduces its denial rate by running its checks before a request is sent. The core steps are checking the policy terms and remaining limit, matching ICD-10 and procedure codes to the discharge summary and report, attaching the findings that support medical necessity and using a document checklist for each procedure type. Recording denials with their reasons and classifying them regularly shows which errors recur and keeps the checklist up to date.

What is pre-authorization in Turkish health insurance?

In Turkish health insurance, provizyon is the insurer's approval to cover the cost of a treatment under the policy terms. Ön provizyon, usually translated as pre-authorization, is that approval obtained before a planned procedure, test or admission. When a missing document or coding error is caught at pre-authorization, the problem can be resolved before the day of the procedure.

How should medical necessity be documented in a pre-authorization request?

The rationale should explain why the procedure is needed in a way the specialist reading the document can understand without asking for more information. It should state the diagnosis, the examination and test findings that support it, any treatments tried earlier and why the planned procedure fits this picture. The diagnosis and procedure codes in the request should match this account exactly.

Can AI make the pre-authorization decision on its own?

Not in a responsible setup. AI reads documents, checks codes and rules, flags gaps and shows the rationale and source behind its recommendation, while the decision rules run in an auditable rule engine. On Opinion AI's platforms, the decision to approve, deny or request additional documents is always made by the relevant specialist, and every output can be audited later.

Does using AI in pre-authorization take patient data outside the institution?

It does not have to. Health data is special category personal data under KVKK, Türkiye's personal data protection law, and under the GDPR, so it matters that the system runs on the institution's own servers or in an isolated cloud environment and that the institution remains the data controller. Opinion AI's platforms are set up this way: the data stays inside the institution, the data the model sees passes through KVKK-compliant masking, and Opinion AI does not collect personal data.

To assess with us how pre-authorization checks could work on your own files, use the POC Request form. For the insurer's side, see the guide on AI-assisted pre-authorization and fraud review in health insurance; for billing, see the guide on preventing SUT non-compliance in SGK hospital billing.