Clinical document extraction pulls named fields out of the documents in a patient chart, and the right fields depend on the document type. An H&P yields history, medications and allergies. An operative report yields diagnoses before and after surgery, the procedure and specimens. A continuity of care document already carries coded sections. Identify the type first, then extract.
Which clinical document types matter for extraction?
The clinical document types that matter most for extraction are the ones that repeat in almost every chart: the history and physical, progress notes, consultation notes, operative and procedure notes, discharge summaries, transfer and referral documents, and continuity of care documents. Each has a typical set of sections, which tells you which fields it can supply.
HL7 publishes a standard list. The HL7 Consolidated CDA implementation guide defines templates for Care Plan, Consultation Note, Continuity of Care Document, Discharge Summary, History and Physical, Operative Note, Procedure Note, Progress Note, Referral Note, Transfer Summary and Unstructured Document. Those templates describe electronic documents, but scanned and printed versions of the same documents follow similar headings, because they were written for the same purposes.
Clinical document types, sections and fields
This table is a working guide, not a standard. Section names follow the HL7 templates where they exist, and the fields are typical examples.
| Document type | Standard sections (examples) | Fields worth extracting |
|---|---|---|
| History and physical | Chief complaint, history of present illness, medications, allergies, family and social history, review of systems, physical exam, vital signs, results, assessment and plan | Encounter date, author, medication list, allergy list, problem list, vital sign values with units |
| Operative note | Preoperative diagnosis, postoperative diagnosis, procedure description, anesthesia, findings, estimated blood loss, specimens, complications | Procedure date, surgeon, diagnoses before and after, procedure name as written, specimens list, implants |
| Discharge summary | Hospital course, discharge diagnosis, allergies, discharge medications, discharge instructions | Admission and discharge dates, diagnoses, medication list with status, follow-up appointments |
| Continuity of care document | Allergies, medications, problems, results, social history, vital signs | Coded entries already in the file, plus the document date and source organization |
| Transfer summary or referral note | Reason for transfer or referral, current problems, medications, allergies, pending items | Sending and receiving facility, transfer date, reason, current medication and allergy lists |
| Progress note | Subjective, objective, assessment, plan (when written in SOAP form) | Note date, author, vital signs, plan items |
The discharge summary row is covered field by field in the post on what a discharge summary contains as data. Lab reports are not on this list because they have their own category of posts on lab report fields and formats.
History and physical document structure
A history and physical, or H&P, is the document written when a patient is admitted or before a planned procedure, and its structure moves from the patient's story to the exam to the plan. The HL7 profile describes it as "a medical report that documents the current and past conditions of the patient."
Electronic H&Ps have a defined set of sections. The HL7 C-CDA on FHIR History and Physical profile requires Allergies and Intolerances, Family History, General Status, Medications, Physical Exam, Results, Review of Systems, Social History and Vital Signs. It lists others as optional, including Chief Complaint, History of Present Illness, Past Medical History, Assessment and Plan, Problem List and Procedures.
Timing rules make the H&P a frequent audit target. The hospital rule in 42 CFR 482.24 requires an H&P documented no more than 30 days before or 24 hours after admission or registration, and before surgery or a procedure requiring anesthesia. An audit therefore needs the H&P's date and time as a field, along with the admission time.
For extraction, the H&P is a rich but uneven source. The allergy and medication sections are usually lists and extract well. The history of present illness is narrative and yields few clean fields. The physical exam can be a template with checkboxes or a paragraph of prose, depending on who wrote it. Vital signs often appear as a short line of numbers that needs each value kept with its unit.
Fields worth pulling from a fictional H&P for patient Iris Tamworth, record EX-550812:
- Document date and time: 2026-04-11 07:40 (example)
- Author: Dr. Femi Adeyemi (fictional)
- Allergies: "NKDA" kept as written, recorded as no known drug allergies
- Medications: one entry per line, as written
- Vital signs: each value with its unit and time
The assessment and plan is where clinical reasoning lives. Extraction can copy the list of problems it names. It should not try to decide what the plan means.
Operative report data extraction
Operative report data extraction pulls the facts of a surgery from the surgeon's note: the date, the surgeon, the diagnoses before and after, the procedure performed, anesthesia type, findings, blood loss, specimens and complications. Most of these sit under predictable headings, which makes the operative report one of the more structured clinical documents.
The federal rule sets the core content. The hospital surgical services rule, 42 CFR 482.51, requires an operative report "describing techniques, findings, and tissues removed or altered", written or dictated immediately after surgery and signed by the surgeon.
The HL7 C-CDA on FHIR Operative Note profile turns that into sections. It requires Anesthesia, Complications, Preoperative Diagnosis, Procedure Estimated Blood Loss, Procedure Findings, Procedure Specimens Taken, Postoperative Diagnosis and Procedure Description. Optional sections include Procedure Implants, Procedure Indications, Planned Procedure and Surgical Drains.
A field map for one fictional operative report:
| Field | Example value |
|---|---|
| Procedure date | 2026-05-19 |
| Surgeon | Dr. Ruth Okonkwo (fictional) |
| Preoperative diagnosis (as written) | "Right inguinal hernia" |
| Postoperative diagnosis (as written) | "Right inguinal hernia, indirect" |
| Procedure (as written) | "Open right inguinal hernia repair with mesh" |
| Anesthesia | General |
| Estimated blood loss | 10 mL |
| Specimens | None |
| Implants | Mesh, lot number as printed |
| Complications | "None" |
Two fields cause most trouble. The procedure description is a long narrative, and the procedure name sits in a short line above it; extract the short line and leave the narrative as text. Implant entries may include lot and catalog numbers that are easy to misread, so they are good candidates for review.
What do continuity of care documents contain?
A continuity of care document, or CCD, is an electronic summary of a patient's record built on an HL7 template, and it contains coded sections for allergies, medications, problems, results, social history and vital signs. In the words of the HL7 profile, the CCD "represents a core data set".
The HL7 C-CDA on FHIR Continuity of Care Document profile requires six sections: Allergies and Intolerances, Medications, Problem, Results, Social History and Vital Signs. Optional sections include Procedures, Immunizations, Encounters, Plan of Treatment, Family History, Payers, Goals and Health Concerns.
The big difference from a scanned page is that a CCD file is already data. Its entries carry codes and dates in machine-readable form, so reading it is a parsing job rather than an optical character recognition job. Nobody has to guess where the allergy list starts.
The catch is that a CCD often arrives printed. A clinic may receive a CCD as a rendered PDF or a fax, which strips out the codes and leaves only the human-readable tables. At that point it is a scanned document again, and it needs the same page-by-page extraction as any other. Ask the sender for the original file when you can.
A CCD also shows only what the sending system chose to include. Narrative notes, operative details and much of the hospital course are usually absent. Treat it as a summary that points to the fuller documents, not a substitute for them.
Transitions of care paperwork as a data source
Transitions of care paperwork is the set of documents that travel with a patient when care moves from one setting to another: transfer summaries, referral notes, discharge summaries, medication lists and pending-result notes. As a data source, it gives the receiving clinic a snapshot of current problems, medications, allergies and open items at the moment of handoff.
Federal rules require that information to move. The discharge planning rule, 42 CFR 482.43, says a hospital must send necessary medical information about the patient's current course of illness and treatment, post-discharge goals of care and treatment preferences to the providers responsible for follow-up care, at the time of discharge.
For extraction, handoff documents share a pattern. They are short, they repeat lists from longer documents, and they carry dates and facility names that matter for matching. Useful fields include the sending facility, the receiving facility or clinician, the transfer or referral date, the stated reason, the current medication list, the allergy list and any pending tests with the person responsible for them.
The common mistake is treating a handoff list as more current than it is. A medication list printed on the transfer date reflects that date. A later discharge summary may change it. Keep the document date with every list, so a reviewer can tell which list is the latest.
Clinical document types that need their own extraction rules
A clinical document type needs its own extraction rules when its fields, layout or meaning differ enough that a generic template would put values in the wrong place. Operative reports, medication-heavy documents, flowsheets and free-text notes are the usual cases. One template for every document type works only for the fields every document shares.
Generic rules handle the header well. Patient name, record number, document date and author appear on nearly every page, in similar places. Past the header, a single template breaks down quickly:
- Diagnosis fields differ by type. An operative note has pre- and post-operative diagnoses. A discharge summary has principal and secondary diagnoses. Merging them loses the distinction.
- Medication lists differ in meaning. A home medication list on an H&P and a discharge medication list mean different things, even if they look alike.
- Flowsheets are tables of repeated values. Vital signs charted every four hours need a time on every value.
- Narrative notes have no fixed position for anything. Progress notes and dictations need text-based rules. The guide to SOAP note structure for clinical notes extraction covers the most common layout.
For documents that are mostly prose, the approaches and their limits are set out in unstructured healthcare data extraction, and note-by-note detail is in clinical notes extraction. Teams wiring extracted fields into their own systems can find formats and schemas in the developer integration category. Schema Care reads medical records for encounter dates, problem lists, allergies and vitals, and it accepts no real charts until a business associate agreement is signed, as set out in its compliance status. Every paid plan includes that agreement.