Health insurance claim form data extraction: CMS-1500 and UB-04
For medical billing companies, providers and claims teams that still receive paper, fax and PDF claims: what each form holds, which fields to extract and check, and where automation fits.
Key takeaways
- Health insurance claim data extraction reads paper, fax and PDF claim forms into structured claim data: patient and insured IDs, diagnosis and procedure codes, dates of service, charges and provider NPIs.
- The CMS-1500 is the professional claim form, used by physicians, other practitioners and suppliers. The UB-04 (CMS-1450) is the institutional form, used by hospitals and other facilities.
- A CMS-1500 holds up to 12 ICD-10-CM diagnosis codes and six service lines, each with a CPT or HCPCS code, up to four modifiers, a diagnosis pointer, a charge, units and the rendering provider's NPI.
- Check every claim before it moves on: NPIs pass the check digit, codes are valid for the date of service and line charges add up to the total, and the signature box is filled.
- Claim forms carry protected health information, so the software that reads them should be HIPAA compliant and covered by a business associate agreement (BAA).
On this page
Health insurance claim data extraction reads claim forms, mainly the CMS-1500 for professional claims and the UB-04 (CMS-1450) for institutional claims, into structured data: who the patient and insured are, what was diagnosed and done, when, by whom and for how much. Billing companies, providers and claims teams use it for the claims that still arrive on paper, by fax or as PDFs rather than as electronic X12 837 transactions.
This guide covers the forms, the fields to extract, the checks to run on every claim and how to automate the work.
What are health insurance claim forms?#
A health insurance claim form is what a provider sends a health plan to get paid for care. Professional and institutional claims each have a standard form, and billing teams handle two related documents alongside them:
CMS-1500
The professional claim, used by physicians, other practitioners and suppliers. It has 33 numbered items, with up to six service lines in item 24.UB-04 (CMS-1450)
The institutional claim, used by hospitals and other facilities. It has 81 form locators and lists charges by revenue code.Itemized bill
The provider's line-by-line list of services and charges. Medicare patients can ask for one in writing, and the provider has 30 days to send it.Explanation of benefits (EOB)
The insurer's statement to the patient, summarizing the costs of the care received. Providers get the payment detail for each claim and line as a remittance advice, electronically as an X12 835 file.
Fields to extract from a CMS-1500 and a UB-04#
The two forms carry much of the same data in different boxes. Box numbers follow the NUCC instruction manual for the 1500 (version 02/12) and CMS's instructions for the UB-04.
| Field | CMS-1500 | UB-04 |
|---|---|---|
| Insured's ID | Item 1a | FL 60 |
| Patient name, birth date and sex | Items 2 and 3 | FL 8, 10 and 11 |
| Diagnosis codes (ICD-10-CM) | Item 21: up to 12, lines A to L | FL 67 principal, 67A to 67Q others |
| Dates of service | 24A, per line | FL 6 statement period, FL 45 line dates |
| Procedure codes | 24D: CPT or HCPCS, up to 4 modifiers | FL 42 revenue code, FL 44 HCPCS, FL 74 inpatient procedures |
| Units | 24G | FL 46 |
| Charges | 24F per line, item 28 total | FL 47 per line, total on revenue code 0001 |
| Prior authorization | Item 23 | FL 63 |
| Provider NPIs | 24J rendering, 32a facility, 33a billing | FL 56 billing, FL 76 attending |
| Federal tax ID | Item 25 | FL 5 |
Each service line's code, modifiers, diagnosis pointer (24E), units and charge belong together, so extraction has to keep every line intact, not just read the boxes.
How claim form extraction works#
- Mailroom scans
- Faxed claims
- PDFs from billing systems
- 01Classify the form
- 02Extract header and service lines
- 03Validate codes and totals
- 04Review exceptions
Classification tells a CMS-1500 from a UB-04, an itemized bill or an EOB, and splits a batch into separate claims. Extraction reads every box, handwritten entries included, and keeps each service line together. Validation runs the checks below; anything uncertain or failing goes to a person with its source line highlighted, and clean claims go on to your claims or billing system.
Checks to run on every claim form#
- NPIsEach NPI has 10 digits and passes its check digit, which CMS calculates with the Luhn formula as if the prefix 80840 came first.
- Diagnosis codesItem 21 carries the ICD indicator 0 and up to 12 ICD-10-CM codes, each 3 to 7 characters with no decimal point, valid on the date of service.
- Procedure codesEach 24D code is a CPT or HCPCS code in effect on the date of service, with no more than four two-character modifiers.
- Diagnosis pointersEvery 24E pointer names a diagnosis line (A to L) that is filled in.
- TotalsThe 24F line charges add up to the total charge in item 28. On a UB-04, the lines add up to the revenue code 0001 total.
- Dates and unitsDates of service fall inside the member's coverage and any prior authorization, and grouped dates match the days or units in 24G.
- SignatureItem 31 has a signature, "Signature on File" or "SOF", and a date.
In Docsumo, checks like these run as steps you set up in the workflow, such as an AI step or a few lines of your own Python code (the NPI check digit is a short one). Checking a claim against other documents in the file, like the member card or the prior authorization, is cross-document validation, on the Enterprise plan.
Automating claim intake with Docsumo#
Docsumo is an intelligent document processing (IDP) platform. It reads claim forms like any other document type, along with EOBs, itemized bills and medical records, and sends values below your confidence thresholds to your own reviewers. It's HIPAA compliant and SOC 2 Type 2 audited, and a BAA is available (security details). It doesn't adjudicate or pay claims: the data goes to your claims, billing or practice management system through API and webhooks.
- 99%field-level accuracy on 250+ document types
- 95%+of documents processed straight through, without manual review
- <5 minper document, down from 2+ hours
The bottom line#
Claim forms are standardized, so the work is predictable: know which box holds what, keep each service line together, and check codes, NPIs and totals before a claim moves on. Automate the read and the checks, and your team handles only the claims that fail a rule.
Book a demo with a few sample claim forms, or start a free trial.
Frequently asked questions#
What is health insurance claim data extraction?
Reading claim forms such as the CMS-1500 and UB-04 from paper, fax or PDF and turning them into structured claim data: patient and insured IDs, diagnosis and procedure codes, dates of service, charges and provider NPIs. Software built for it, sometimes called a claim parser or claim data extractor, uses OCR plus models that know which box holds which field, then checks the result before it reaches a claims or billing system.
What is the difference between the CMS-1500 and the UB-04?
The CMS-1500 is the professional claim form, used by physicians, other practitioners and suppliers, with numbered items 1 to 33. The UB-04, also called the CMS-1450, is the institutional form used by hospitals and other facilities; it lists charges by revenue code across 81 form locators.
Are paper claim forms still used?
Yes. Medicare requires electronic claims except from small providers (fewer than 10 full-time equivalent employees for physicians and suppliers, fewer than 25 for institutional providers such as hospitals) and a few other exceptions, and that rule doesn't apply to claims sent to other health plans. The electronic versions are the X12 837 professional and institutional transactions.
Is a claims log with patient ID, code and amount structured data?
Yes. A log that holds the same fields in every row, such as patient ID, procedure code and amount, is structured data. A claim form on paper or PDF is semi-structured: the boxes are standard, but the values have to be read off the page first. See structured vs unstructured vs semi-structured data.
Can explanation of benefits forms be processed automatically?
Yes. Paper EOBs and remittances vary by payer, so extraction reads each claim and service line with the amount paid, the adjustments and their reasons, and what the patient owes, then checks that the amounts add up. Electronic remittance advice arrives as an X12 835 file, which billing systems can post without extraction.
Is claim form data extraction HIPAA compliant?
It can be. Claim forms carry protected health information, so use software that is HIPAA compliant and covered by a business associate agreement (BAA). Docsumo is HIPAA compliant and SOC 2 Type 2 audited, and a BAA is available: see security.
Sources
- NUCC: 1500 Claim Form Reference Instruction Manual, version 13.0 (July 2025)
- CMS: Medicare Claims Processing Manual, Chapter 26 (Form CMS-1500)
- CMS: Medicare Claims Processing Manual, Chapter 25 (Form CMS-1450, the UB-04)
- CMS: Medicare Claims Processing Manual, Chapter 24 (mandatory electronic claims)
- CMS: Requirements for the NPI and its check digit
- CMS: ICD-10-CM Official Guidelines for Coding and Reporting, FY 2026
- CMS: Health care payment and remittance advice
- New Hampshire Insurance Department: Explanation of benefits (EOB)
- 42 U.S.C. 1395b-7: Explanation of Medicare benefits and itemized statements
First published . Last updated .