# Social Security Disability File OCR: Reading the Exhibits

- Author: Schema Law (https://www.schemalaw.com/)
- Published: 2025-09-01T06:34:08Z
- Updated: 2026-10-11T19:22:20Z
- Category: Practice areas
- Canonical: https://www.schemalaw.com/blog/social-security-disability-file-ocr

> How to OCR a Social Security disability file before a hearing: the exhibit sections, converting medical evidence and agency forms, and searches for a brief.

To OCR a Social Security disability file before a hearing, download the exhibit file, keep that copy untouched, and convert the medical evidence section first, one exhibit per file, into searchable PDF and text. Then convert the agency forms, check dates and measurements against the page images, and search the result for the limitations your brief will argue.

The file is the same record the administrative law judge reads. Searchable text does not change what is in it. It changes how fast you can find the three pages in two thousand that describe how long the client can stand.

Exhibit stamps are how you get from a search hit back to the right page. Before the adjudicatory hearing, search the converted medical exhibits for treatment notes and prescription drug lists in the administrative record. If a decision in the file discusses substantial evidence, search that phrase and read the exhibits it cites from the image. Confirm each exhibit date before you cite the page.

## What is in a Social Security disability claim file?

A Social Security disability claim file holds the agency's decisions and notices, the application and earnings material, the questionnaires the claimant filled in, and the medical records. The agency's own manual sorts an electronic folder into lettered sections A through F, and the exhibit list in your copy shows which documents your hearing office placed where.

The agency's operations manual sets out the sections. [POMS DI 81001.005](https://secure.ssa.gov/apps10/poms.nsf/lnx/0481001005) says the certified electronic folder contains "documents organized in the same sections and order" as the paper folder and lists them:

| Section | Name in POMS DI 81001.005 | Examples the manual gives |
| --- | --- | --- |
| A | Payment Documents/Decisions | The SSA-831 determination form, administrative law judge decisions |
| B | Jurisdictional Documents and Notices | Notices, appeal requests, the SSA-1696 appointment of representative |
| C | Current Development/Temporary | Work in progress, such as a letter requesting records |
| D | Non-Disability Related Development | Applications, earnings records, non-medical evidence |
| E | Disability Related Developments | Disability questionnaires such as the SSA-3368 and SSA-3441, the SSA-827 authorization |
| F | Medical Records | Hospital records, laboratory reports, physician's records, outpatient notes |

The examples in the right-hand column come from [POMS DI 70005.005](https://secure.ssa.gov/apps10/poms.nsf/lnx/0470005005), which describes the paper folder the electronic sections mirror. That page says of the divider tabs, "The lists are not all-inclusive." So a document can sit somewhere you do not expect. It also says that when a hearing is requested, "support staff in the hearing office will organize the relevant documents for the hearing." The manual pages cited here do not describe how a hearing office numbers its exhibits, so take the numbering from the exhibit list in your own download and treat that list as your index.

Check the page counts on your exhibit list. A claimant with five years of treatment at two clinics and a hospital can produce a section F many times the size of everything else combined, and when yours is the largest section, the conversion work starts there.

One reading note: the regulations quoted in this post are from 20 CFR part 404, which governs Title II disability insurance claims. If the claim is for Supplemental Security Income, look up the corresponding provision in part 416 and read it there before relying on a quotation.

## Why is a disability exhibit file hard to search as delivered?

A disability exhibit file is hard to search as delivered because records were filed in the order and bundles in which they arrived, and because pages that came in by scan or fax may be pictures with no text behind them. Test your download: try to select a sentence in a few exhibits from each section.

The manual explains the ordering. POMS DI 81001.005 lists the ways material enters the electronic folder, among them "Scan or fax documents" and "Add documents through Electronic Records Express (ERE)". POMS DI 70005.005 then says medical records at the initial and reconsideration levels may be filed "in receipt-date order or treatment-date order", and defines receipt-date order as filing each new set "on top of the evidence already in the file." It adds that hospital records for one stay are kept together and that cover letters stay with the records they came with.

For a reader with a hearing date, three things follow from that filing method:

- **One provider can be in several exhibits.** Records requested at the initial level, again at reconsideration, and again by your office each arrive as a separate bundle. A clinic's 2023 notes may be in three places.
- **The same pages can repeat.** Each later request may bring an updated copy of the whole chart, so earlier visits appear again in the later bundle.
- **Order inside a bundle is the sender's.** A provider's records system prints in its own sequence, which may be newest first, grouped by note type, or neither.

Whether a given exhibit has a text layer depends on how it reached the folder and what the agency's systems did with it afterwards, and the manual pages cited here do not say. So test. Open six exhibits spread across sections A, E and F, and in each drag the cursor across a line and paste what you selected into a blank document. The post on why you cannot copy text from a PDF explains [how to tell whether a PDF has real text](https://www.schemalaw.com/blog/why-cant-i-copy-text-from-a-pdf#how-can-you-tell-whether-a-pdf-has-real-text). An exhibit whose text pastes as nonsense needs converting like any scan, for the reasons in the post on [text copied from a PDF that comes out as gibberish](https://www.schemalaw.com/blog/convert-pdf-to-readable-text#why-does-text-copied-from-a-pdf-come-out-as-gibberish).

Record the result per exhibit on a worksheet. Typed agency notices that already carry good text need no work at all, and there is no reason to spend page allowance on them.

## Converting treatment records in the medical evidence section

Convert the treatment records one exhibit at a time, keeping the exhibit number in every file name, and export a searchable PDF for citation plus plain text or Word for your notes. Do not reorder, trim or merge pages in the copies you cite from, because the hearing office's page numbers are how everyone refers to the record.

A procedure for section F:

1. **Save the download as received.** Put it in a folder nobody edits and note the date you pulled it. The folder grows as evidence is added, so the date identifies which version you worked from.
2. **Split by exhibit.** If the file came as one PDF, split it at exhibit boundaries using the exhibit list. Name each piece by its exhibit number and source, for example `F07 Lakeshore Orthopedics 2022-2024.pdf`.
3. **Check sizes.** Schema Law accepts a PDF, PNG or JPEG of up to 20 MB, and the page limit per document depends on the plan, as the [pricing page](https://www.schemalaw.com/pricing) shows. An exhibit over the limit is split at a page boundary, with the exhibit page range kept in each name: `F07 p001-250`, `F07 p251-412`.
4. **Convert in priority order.** Start with the treating sources whose notes carry the limitations: the specialist, the primary care office, the therapist. Emergency room visits and imaging come next. Billing ledgers and pharmacy printouts come last.
5. **Export two things.** The searchable PDF is the original pages with a hidden text layer, so page numbering is unchanged. Plain text or Word is for extracting quotations into an outline. The complete set of outputs is under [export formats on the homepage](https://www.schemalaw.com/#formats).
6. **Verify what you will cite.** For every passage that goes into the brief, read the converted words against the page image and write the exhibit and page beside it.

Step 6 is not optional. Schema Law's homepage says recognition is strongest on clean print and weaker on faint faxes, stamps over text and handwriting, and that nothing marks a passage as uncertain. A range of motion recorded as 40 degrees and read as 10, or a visit dated 2023 and read as 2028, will not announce itself.

### Tables, flowsheets and handwriting

Lab panels, medication lists and vital sign grids are tables, and a table exported to a spreadsheet can be sorted by date across visits. The post on [clinical tables from medical records in Excel](https://www.schemalaw.com/blog/medical-records-pdf-to-excel) covers medication records, flowsheets and lab results in that form.

Handwritten progress notes are the weak point of any medical exhibit. The post on OCR for medical records explains what to expect from [handwritten chart notes and provider signatures](https://www.schemalaw.com/blog/ocr-medical-records-for-attorneys#handwritten-chart-notes-and-provider-signatures). For a hearing file the working rule is simple: where a handwritten note matters, a person transcribes it, and the transcription is checked by a second person.

### New evidence you are submitting

Your own records requests produce more scans. [20 CFR § 404.935(a)](https://www.law.cornell.edu/cfr/text/20/404.935) says each party must "inform us about or submit any written evidence" as required by § 404.1512 "no later than 5 business days before the date of the scheduled hearing", and [§ 404.1512(a)(1)](https://www.law.cornell.edu/cfr/text/20/404.1512) says that when you submit evidence received from another source, "you must submit that evidence in its entirety, unless you previously submitted the same evidence to us or we instruct you otherwise." Convert a working copy to read and search. What goes to the agency is the complete set as the provider sent it. How those rules apply to your hearing is for the representative to decide.

## Function reports, work history reports and other agency forms

Function reports, work history reports and disability reports are multi-page agency forms, and the copies in a file may be typed, printed from an online entry, or filled in by hand. Printed answers convert like any typed page. Handwritten answers convert poorly, and a checkbox cannot be trusted from converted text at all.

The agency's manual describes what each form is for. [POMS DI 11005.016](https://secure.ssa.gov/poms.nsf/lnx/0411005016) lists them:

| Form | What POMS DI 11005.016 says it does |
| --- | --- |
| SSA-3368, Disability Report Adult | Collects "basic information on an adult claimant's medical conditions", treatment and medical sources |
| SSA-3369, Work History Report | Is "used to supplement the SSA-3368 with detailed information about jobs that the claimant has performed" |
| SSA-3373, Function Report Adult | It "records information about a claimant's ability to function and to perform the activities of daily living" |
| SSA-3380, Function Report Adult, Third Party | Collects the same kind of evidence "from third parties familiar with the claimant's ability to function" |
| SSA-3441, Disability Report Appeal | Is "used for appeal cases to update information previously obtained or to identify additional impairments and medical sources" |

The same page notes that the SSA-3373 "is not mandatory in all cases", so its absence from a file is not by itself an error.

### Reading a form after conversion

A form page is a set of printed questions with short answers beneath or beside them. The printed question text is clean type repeated on every copy of the form, the kind of print OCR is strongest on. The answer is the part that varies. Three habits help:

- **Search by the question, read the answer on the image.** The printed prompt is a dependable search target. Searching for a phrase from the question about lifting takes you to the right page of every function report in the file, including the third-party one. Then read the answer from the picture of the page.
- **Never take a ticked box from the text.** A checked box can come through as a letter, a bracket, a stray mark or nothing. Open the image for every yes or no you rely on.
- **Transcribe handwritten answers you will quote.** Type them into your outline yourself, mark anything illegible as illegible, and have someone else read the transcription against the form.

### Comparing forms across time

A file can hold an initial disability report, an appeal report and two function reports completed a year apart. Put the answers to the same question side by side in a table with one row per question and one column per form date. Differences are not necessarily inconsistencies, since conditions change, but the judge can see both forms and you should see them first.

A fictional example of such a row:

| Question | Function report, March 2024 (Ex. 4E p. 3) | Function report, February 2025 (Ex. 11E p. 3) |
| --- | --- | --- |
| How far can you walk before needing to stop and rest? | "2 blocks" | "half a block, then I sit" |

The exhibit and page under each date are what make the table usable at the hearing.

## Can OCR read consultative examination reports and physician opinion forms?

A typed consultative examination report is ordinary print, which is where OCR is strongest. A physician opinion form is harder when it is a printed form completed with check marks, circled numbers and handwritten remarks, because OCR is weak on those marks. Test both on your own exhibits.

### Consultative examination reports

The regulation tells you what a complete report should contain, which doubles as a checklist for navigating one. [20 CFR § 404.1519n(c)](https://www.law.cornell.edu/cfr/text/20/404.1519n) lists the elements: the "major or chief complaint(s)", a detailed history of them, a description of the pertinent positive and negative findings, the results of laboratory and other tests, the "diagnosis and prognosis for your impairment(s)", and a medical opinion. The paragraph adds that "the absence of a medical opinion in a consultative examination report will not make the report incomplete."

After conversion, search the report for each element's heading and confirm it is there. Then check the figures by eye: grip strength, range of motion in degrees, visual acuity, test scores. A typed "5/5" and a typed "3/5" differ by one character.

Paragraph (e) of the same section deals with signatures. It says a report annotated "not proofed" or "dictated but not read" is not acceptable, and that a "rubber stamp signature of a medical source or the medical source's signature entered by any other person is not acceptable." Converted text will find the words "dictated but not read" if they were typed. It will not tell you whether a signature is a stamp or ink. That is a look at the image, and what follows from it is the representative's call.

### Opinion forms

An opinion form puts questions to a medical source, such as how long the claimant can sit, stand and walk and how much the claimant can lift. The regulation's definition of a medical opinion covers those subjects: [20 CFR § 404.1513(a)(2)](https://www.law.cornell.edu/cfr/text/20/404.1513) describes it as a statement about what you can still do despite your impairments, and lists abilities including "sitting, standing, walking, lifting, carrying, pushing, pulling", and "understanding; remembering; maintaining concentration, persistence, or pace". The same paragraph points to § 404.1527(a) for the definition that applies to claims filed before March 27, 2017.

On the page, the answer to each question is a mark. For every opinion form in the file:

1. Convert it so the printed questions are searchable and the form can be found.
2. Build a small table by hand from the image: question, the option marked, any handwritten remark, exhibit and page.
3. Have a second person check the table against the image.

That table is short, a row for each question on the form, and a brief that relies on the source's opinion will cite from it line by line. It is worth making by hand.

Prior administrative medical findings from the state agency's consultants are a third kind of document. Where they are typed pages, they convert like any other typed page. Find them through the exhibit list.

## Which searches find the limitations a hearing brief needs?

The searches that find limitations are the regulation's own words for work abilities, the clinical words providers use for the same things, and the names of the client's conditions and medications. Run each across the converted medical exhibits, log every hit with its exhibit and page, and remember that a search only finds words that were recognized correctly.

Start from the ability list in 20 CFR § 404.1513(a)(2) and add the vocabulary of the chart beside each term:

| Ability in the regulation | Also search for |
| --- | --- |
| Sitting, standing, walking | "ambulat", "gait", "cane", "walker", "antalgic", "tolerat" |
| Lifting, carrying | "lbs", "pounds", "lift", "no heavy" |
| Reaching, handling | "grip", "fine motor", "numbness", "tingling", "overhead" |
| Stooping, crouching | "bend", "squat", "kneel", "range of motion", "ROM" |
| Concentration, persistence, or pace | "attention", "focus", "distract", "memory", "fatigue" |
| Responding to supervision, co-workers | "irritab", "panic", "isolat", "crowds", "conflict" |

Partial words such as "ambulat" catch ambulate, ambulating and ambulatory in one pass, if your search tool matches part of a word. Test that on a page you know before relying on it.

### Searches beyond the ability list

- **Medication side effects.** Search each drug name from the medication list, then "drows", "dizz", "sedat" and "nausea" near those hits.
- **Frequency of treatment.** Search the provider's name or letterhead text and count dated visits. Missed and cancelled appointments have their own words: "no show", "cancel", "resched".
- **Assistive devices and orders.** "prescribed", "DME", "brace", "TENS", "injection", "referral".
- **Normal findings.** Search for the terms the other side of the record uses: "normal gait", "full strength", "5/5", "no acute distress", "within normal limits". A brief that ignores them has not dealt with the file.
- **The alleged onset date.** Search the month and year in each format the records use, to find the visits closest to it.

### Logging what you find

Keep a hit log as a spreadsheet: search term, exhibit, page, date of treatment, provider, the sentence quoted exactly, and whether you verified it against the image. Sort by date of treatment and the log becomes the skeleton of the medical summary. Sort by term and it becomes the support for each limitation.

Two limits apply. A term misread during recognition is invisible to search, so a clean result is "no hits in the text", never "not in the record". And search cannot be relied on for handwritten notes, so read the handwritten exhibits yourself.

A client with military service may have a second agency file covering the same conditions. The post on [OCR for a VA claims file](https://www.schemalaw.com/blog/ocr-va-claims-file) describes that file and how to split it into dated documents.

Disability files are dense with health information and identifiers. Before sending one to any outside service, read the vendor's terms against your duties to the client. Schema Law's commitments are stated under [security on the homepage](https://www.schemalaw.com/#security).

## Try it on a real document

Upload a scanned PDF and export it as Word, a searchable PDF or plain text on the free plan. Your documents are never used to train an AI model. [Start free](https://www.schemalaw.com/api/auth/google)