Skip to main content

Yira.ai

Health Insurance Claims Glossary: 25 Key Terms Defined

Key terms used in health insurance claims processing, medical document AI, TPA operations and fraud detection — defined clearly for operations teams, reviewers and software buyers.

Adjudication

The formal process by which an insurer or TPA evaluates a submitted health insurance claim and decides whether to pay it in full, pay it partially or deny it. Adjudication checks that the claim documents are complete, the treatment is covered under the policy, the amounts claimed are reasonable and consistent with the diagnosis, and that no exclusions or waiting periods apply. Automated adjudication uses rules engines and AI document extraction to handle straightforward claims without manual review.

Related: TPA · Denial Management

Authorization Code

A unique reference number issued by an insurer or TPA when a pre-authorisation request is approved. The code confirms that the proposed procedure or hospital admission is covered under the policyholder’s plan and authorises the network hospital to proceed with cashless treatment. The code must appear on the final claim submission for the claim to be accepted.

Related: Pre-Authorisation · Cashless Claim

Cashless Claim

A health insurance claim settlement mode in which the insurer or TPA pays the hospital directly and the policyholder does not make an out-of-pocket payment at discharge (beyond any deductibles or sub-limits). The hospital must be in the insurer’s network and the admission must be pre-authorised (for planned procedures) or notified within the stipulated timeframe (for emergencies). Documents are exchanged directly between the hospital and the TPA.

Related: Reimbursement Claim · Network Hospital · Pre-Authorisation

Claim Leakage

The difference between what an insurer actually pays on a claim and what should have been paid under the strict terms of the policy. Leakage is caused by errors in manual data entry (missed sub-limits, incorrect co-pay calculations), adjudicator mistakes, or undetected fraud and upcoding. Reducing claim leakage is a primary business case for automated medical document extraction and validation — when every field is accurately read from the source document, keying errors that inflate payments are eliminated.

Related: Upcoding · Sub-Limit · FWA

CPT Code

Current Procedural Terminology code — a five-digit numeric code maintained by the American Medical Association that identifies a specific medical, surgical or diagnostic procedure for billing, documentation and reporting purposes. CPT codes are the standard for hospital billing in the United States and are increasingly used in GCC insurance markets. Medical document intelligence systems extract CPT codes from hospital bills and cross-reference them against the diagnosis codes on the discharge summary to detect billing inconsistencies.

Related: ICD-10 · Upcoding · Unbundling

Denial Management

The operational process of reviewing claims that have been denied by an insurer or TPA, identifying the reason for each denial, correcting the claim or supporting documentation, and resubmitting for payment. Denial management also includes analysis of denial patterns to identify systemic issues — such as a specific hospital consistently submitting incomplete documentation or a particular diagnosis code triggering automatic denials — so that upstream fixes reduce future denials.

Related: Adjudication · TPA

Discharge Summary

A clinical document prepared by the treating physician or medical team at the end of a hospital stay. It records the principal diagnosis (the condition that caused the admission), secondary diagnoses (comorbidities), procedures performed during the stay, length of stay, the patient’s condition at discharge, and follow-up instructions. The discharge summary is the primary clinical evidence document in a health insurance claim; every other claim document — lab reports, operative notes, hospital bill — is cross-referenced against it.

See: Discharge Summary Extraction

Related: ICD-10 · Medical Document Intelligence

Duplicate Claim

A claim submitted more than once for the same episode of care, the same procedure or the same date of service. Duplicates arise accidentally — through resubmission after a system error — or intentionally as a fraud pattern. Automated duplicate detection uses multiple identifiers (member ID, provider ID, diagnosis code, procedure code, service date, billed amount) to flag potential duplicates before payment. See: Duplicate Claim Detection.

Related: FWA

FWA — Fraud, Waste and Abuse

Three overlapping categories of improper payments in health insurance, often grouped together because they all result in overpayment but require different detection and remediation approaches:

  • Fraud is intentional deception for financial gain — submitting claims for services not rendered, upcoding, billing for a non-existent patient, or forging clinical documents.
  • Waste is overutilisation of health services — ordering unnecessary tests or procedures — without intent to deceive, often due to defensive medicine or lack of cost awareness.
  • Abuse is billing that does not meet accepted standards of care or policy terms — charging for services that were provided but were not medically necessary, or applying incorrect billing codes — without necessarily intending to commit fraud.

See: What Is FWA in Health Insurance · How Insurers Detect Health Insurance Fraud

ICD-10 Code

International Classification of Diseases, 10th revision — the global standard diagnostic coding system published by the World Health Organization. ICD-10 codes classify diseases, conditions, injuries and causes of death into alphanumeric codes (e.g. J18.9 for unspecified pneumonia). Health insurers and TPAs use ICD-10 codes to categorise claims, determine coverage eligibility, calculate actuarial risk and detect billing anomalies. A hospital bill listing a procedure code that is not clinically indicated by the ICD-10 diagnosis code on the discharge summary is a common FWA indicator.

Related: CPT Code · SNOMED CT · Discharge Summary

Intelligent Document Processing (IDP)

A technology category that combines optical character recognition (OCR), natural language processing (NLP), machine learning and rules engines to automatically extract structured data from unstructured documents — PDFs, scans, images — without manual keying. IDP platforms are built to handle layout variation across document types. Medical document intelligence is IDP applied specifically to healthcare documents, with additional domain-specific components for medical terminology, clinical code mapping and cross-document validation.

Related: OCR · NLP · Medical Document Intelligence

Loss Ratio

The ratio of claims paid by a health insurer to the premiums collected in the same period, expressed as a percentage. A loss ratio of 80% means the insurer pays Rs.80 (or the equivalent in any currency) in claims for every Rs.100 of premium received. Loss ratios above 100% indicate the insurer is paying out more in claims than it earns in premiums — unsustainable without premium increases or cost controls. Reducing claim leakage and FWA are the primary operational levers for improving the loss ratio without changing premium rates.

Related: Claim Leakage · FWA

Medical Document Intelligence

Domain-specific intelligent document processing applied to healthcare documents. It combines OCR, NLP and named entity recognition (NER) to extract, classify and validate data from health insurance claim documents — lab reports, discharge summaries, hospital bills and pre-authorisation forms — and delivers structured output for claims management systems. Unlike generic IDP, medical document intelligence understands clinical terminology, maps terms to standard codes (ICD-10, CPT, SNOMED CT), validates extracted lab values against clinical reference ranges, and cross-references multiple documents within a single claim to surface anomalies.

See: What Is Medical Document Intelligence · YIRA MedSense

Named Entity Recognition (NER)

A natural language processing technique that identifies and classifies named entities in free text — such as diagnoses, drug names, procedure names, body parts, lab test names, numeric values, dates and provider names — into predefined categories. In health insurance document processing, NER extracts the clinically and financially relevant fields from unstructured clinical text so they can be used in claims adjudication and fraud detection without requiring a human to read the full document.

Related: NLP · Medical Document Intelligence

Network Hospital

A hospital that has a formal contractual agreement with an insurer or TPA to provide treatment to the insurer’s policyholders under cashless claim arrangements at pre-negotiated tariff rates. Network hospitals submit claim documents directly to the TPA or insurer; reimbursement claims can be filed at any hospital but cashless claims are only available at network hospitals. Network size and hospital empanelment are significant competitive differentiators for health insurers.

Related: Cashless Claim · TPA

Natural Language Processing (NLP)

A branch of artificial intelligence that enables computers to understand, interpret and generate human language — both spoken and written. In health insurance operations, NLP is used to extract clinical information from free-text documents (discharge summaries, consultation notes, radiology reports), classify documents by type, and identify sentiment or intent in correspondence. Medical NLP models are trained on clinical corpora and understand the vocabulary, abbreviations and conventions of health records, which differ substantially from general-domain text.

Related: NER · OCR · Medical Document Intelligence

Optical Character Recognition (OCR)

Technology that converts document images — scans, photographs, PDFs — into machine-readable text. Standard OCR handles printed text on clean documents well. Medical-grade OCR is additionally trained on clinical abbreviations, drug names, ICD code strings, and handwritten script common in Indian and GCC hospitals where doctors write notes by hand. OCR is the first stage of a medical document intelligence pipeline; the quality of downstream NLP and data extraction depends on the accuracy of the OCR output.

Related: NLP · IDP · Medical Document Intelligence

Pre-Authorisation

Advance approval required from an insurer or TPA before a planned medical procedure, hospital admission, or high-cost treatment will be covered under a health insurance policy. The treating doctor submits a pre-auth request form with the proposed diagnosis, procedure and estimated cost; the TPA’s medical team reviews it against policy terms and clinical guidelines and either approves, partially approves, or denies it. Pre-authorisation reduces claim disputes at discharge and allows insurers to manage utilisation before it occurs. See: Pre-Authorisation Automation.

Related: Authorization Code · Cashless Claim · TPA

Reimbursement Claim

A health insurance claim type in which the policyholder pays the hospital directly at the time of discharge and then submits the original bills, discharge summary, lab reports and other documents to the insurer for repayment. Unlike cashless claims, reimbursement claims can be filed at any hospital — not just network hospitals — but require the policyholder to arrange payment upfront. The claim is then adjudicated based on the submitted documents; payment is made to the policyholder’s bank account.

Related: Cashless Claim · Adjudication

SNOMED CT

Systematized Nomenclature of Medicine — Clinical Terms: a comprehensive, multilingual clinical terminology owned by SNOMED International that covers diseases, clinical findings, procedures, body structures, organisms and other health concepts. SNOMED CT provides more granular clinical detail than ICD-10 and is used in electronic health records (EHRs) for clinical documentation. In AI-based medical document processing, SNOMED CT is used for standardised entity mapping — ensuring that synonymous clinical terms extracted from different documents are recognised as referring to the same concept.

Related: ICD-10 · NER

Sub-Limit

A cap within a health insurance policy that restricts coverage for a specific category of treatment to a fixed amount or a percentage of the total sum insured, regardless of the actual expenses incurred. Common sub-limits in Indian and GCC health policies cover room rent (e.g. 1% of sum insured per day), specific surgeries (e.g. cataract, hernia), maternity benefits, and psychiatric treatment. Sub-limits are a frequent source of claim leakage when manual adjudicators miss or misapply them; automated adjudication enforces sub-limits systematically on every claim.

Related: Claim Leakage · Adjudication

Third-Party Administrator (TPA)

An organisation licensed by the insurance regulator and contracted by one or more health insurers to manage claims processing, pre-authorisation, network hospital empanelment, member services and sometimes policy issuance on the insurer’s behalf. TPAs are the operational backbone of health insurance in India and many GCC markets — the insurer holds the risk and the product, while the TPA handles the day-to-day service delivery. TPA efficiency is directly measured by claim turnaround time, rejection rates, and customer satisfaction scores. See: TPA Claims Processing Software.

Unbundling

A billing fraud pattern in which a provider submits separate claims for individual components of a procedure that should be billed as a single, bundled procedure code. Because the bundled code pays less than the sum of the individual component codes, unbundling increases the total payment to the provider. Medical document intelligence detects unbundling by identifying procedure codes on the hospital bill that are components of a higher-level bundled code present in the same claim.

Related: Upcoding · FWA · CPT Code

Upcoding

A billing fraud pattern in which a provider bills for a more expensive procedure code, a higher level of care, or a more severe diagnosis than was actually delivered or documented, thereby inflating the claim amount. For example, billing a complex cardiac procedure when a standard catheterisation was performed, or billing an ICU stay when the patient was in a general ward. Upcoding detection compares the procedure codes on the hospital bill against the clinical detail in the discharge summary and lab reports; a mismatch is a fraud indicator. See: How Insurers Detect Health Insurance Fraud.

Related: Unbundling · FWA

Waiting Period

A specified duration at the beginning of a health insurance policy during which certain claims will not be accepted. Common waiting periods include: an initial 30-day waiting period for any illness (emergencies excepted), a 2-4 year waiting period for pre-existing conditions, and a 1-2 year waiting period for specific named diseases such as cataract or hernia. Claims submitted during a waiting period are denied; ensuring that admission dates and diagnosis codes are checked against the policy’s waiting period provisions is a standard adjudication step.

Related: Adjudication · Cashless Claim


This glossary covers 25 core terms used in health insurance claims, TPA operations and medical document AI. Additional terms will be added in future updates. For questions about how MedSense applies these concepts, see the MedSense product overview or request a demo.