Schema Care

Medical Records Data Abstraction Example, Step by Step

By Schema CarePublished Updated

This medical records data abstraction example takes one fictional hospital chart of 24 pages, fills a 14-field abstraction form from it, and records the document and page behind every value. It ends with the four values that needed a judgment call and shows how the same chart would be abstracted differently for a registry or an audit.

The sample chart behind this abstraction example

The sample chart is a fictional, 24-page record for one short hospital stay. The patient, the facility, every date and every medication name are invented for this example. No part of it comes from a real person, and the medication names are made up so they cannot be mistaken for real drugs.

The patient is Dana Example, example medical record number EX-100482, example date of birth 1961-07-14. She was seen in the emergency department late one evening, admitted the next day, and discharged four days later. The chart arrived as one scanned PDF, the way many outside records do.

The documents in the chart

PagesDocumentAuthorWhat it holds
1 to 2Referral letter from an outside clinicExample clinic physicianReason for referral, an outside allergy list
3 to 5Emergency department noteExample ED physicianArrival time, triage vitals, allergies, home medications
6 to 9Admission history and physicalExample hospitalistAdmitting diagnosis, history, medication list
10 to 17Progress notes, one per dayExample hospitalistsDaily plan, medication changes
18 to 21Discharge summaryExample hospitalistDischarge diagnoses, procedures, discharge medications, follow-up
22Pharmacy medication reconciliation formExample pharmacistHome medications checked with the patient
23 to 24Nursing discharge checklistExample nurseDischarge time, disposition, education given

Before abstracting, the pages were split and labelled by document type. A chart that arrives as one long PDF is hard to cite until every page has a document name. Teams that receive charts from other facilities will find more on requesting and sorting outside records in the records retrieval and review category.

The form itself was defined before anyone opened the chart. Defining fields first is step one of the general method in the guide to defining the medical record fields before you start. This example follows that order.

The filled abstraction form for the sample chart

The filled abstraction form below holds 14 fields for the sample chart, each recorded in a fixed format. Dates use year-month-day, times use a 24-hour clock, and medications are written as the chart names them. A value the chart does not contain is recorded as "not documented", never left blank.

The form, filled

#FieldExample value
1Medical record numberEX-100482
2Date of birth1961-07-14
3ED arrival date and time2026-03-01 23:40
4Inpatient admission date2026-03-02
5Discharge date and time2026-03-06 14:15
6Admitting diagnosis, as written"Community-acquired pneumonia"
7Principal discharge diagnosis, as written"Community-acquired pneumonia"
8Procedures during the stayNot documented
9AllergiesCorvaxin (hives); no other allergies documented
10Admission weight72.6 kg
11Number of discharge medications3
12Discharge medicationsPemtrazol 20 mg tablet daily; Lumaferin 10 mg tablet daily; Velorine 5 mg tablet stopped
13Discharge dispositionHome
14Follow-up appointment documentedYes, 2026-03-13

Each value is copied, not interpreted. The diagnosis fields hold the words the physician wrote, because deciding what is wrong with a patient is not the abstractor's job. Field 12 records a stopped drug as stopped rather than leaving it off, so a reader can tell "stopped at discharge" apart from "never listed".

Two formatting choices matter for later use. Weight is converted to kilograms with the original value kept in the source column. And "not documented" in field 8 means the abstractor checked every document that could hold a procedure and found none.

The medication fields follow the method in medication list extraction from records, and the allergy field follows allergy list extraction from records. Both go further into status and source rules than this example needs.

Where each abstracted value came from in the chart

Every abstracted value in this example is traced to one document and one page, so a reviewer can open the chart and confirm it in seconds. The source column below is part of the abstraction, not a note added afterwards. A value without a source cannot be checked or defended.

The source map

#FieldExample valueSource documentPage
1Medical record numberEX-100482Every page header1
2Date of birth1961-07-14ED note header3
3ED arrival date and time2026-03-01 23:40ED note, triage block3
4Inpatient admission date2026-03-02Admission history and physical6
5Discharge date and time2026-03-06 14:15Nursing discharge checklist23
6Admitting diagnosis"Community-acquired pneumonia"Admission history and physical, assessment8
7Principal discharge diagnosis"Community-acquired pneumonia"Discharge summary, diagnoses18
8ProceduresNot documentedDischarge summary, procedures heading reads "None"18
9AllergiesCorvaxin (hives)Referral letter, allergy list2
10Admission weight72.6 kg (160 lb as written)ED note, triage vitals3
11Number of discharge medications3Discharge summary, medications20
12Discharge medicationsSee formDischarge summary, medications20
13Discharge dispositionHomeNursing discharge checklist24
14Follow-up appointment2026-03-13Discharge summary, follow-up21

Why the page matters as much as the document

A document name alone is not enough in a 24-page chart. The discharge summary in this example covers four pages, and the medication list sits on the third of them. Recording page 20 rather than "discharge summary" saves the reviewer a search and makes disagreements easy to settle.

Dates need the same care. The ED note, the admission note and the discharge summary each carry several dates, and the one at the top of a page is not always the date of the event. The rules for picking the right one are in the guide to which date on a medical record page is the encounter date.

Which values in the abstraction example needed a judgment call?

Four values in this abstraction example needed a judgment call: the arrival time, the allergy list, the stopped medication and the discharge time. In each case the chart held two or more candidate values, and the abstractor chose one by applying a written rule, then recorded which sources disagreed so a reviewer could follow the reasoning.

Arrival time versus admission date

The ED note shows arrival at 23:40 on 2026-03-01, while the admission order is dated 2026-03-02. They are different fields, so both are recorded. The mistake to avoid is using the admission date as the start of the encounter, which would make the stay look shorter than it was.

An allergy that only one document lists

The referral letter on page 2 lists "Corvaxin: hives". The ED note on page 4 reads "NKDA", short for no known drug allergies. The form's rule in this example is to record any allergy documented in any source and to note conflicts. So field 9 holds Corvaxin, with a note that the ED note disagrees. The abstractor does not decide which document is right about the patient; a clinician resolves that.

A medication that was held, then stopped

Progress notes on page 13 say "hold Velorine". The discharge summary on page 20 says "stop Velorine". A held drug and a stopped drug are different statuses. Because the form asks for discharge medications, the discharge summary's "stop" wins and is recorded as stopped.

Two discharge times

The discharge summary is signed at 11:02, and the nursing checklist records the patient leaving at 14:15. The form's definition asks for the time the patient left, so the nursing checklist on page 23 is the source. The signature time is a document time, not an event time.

How the judgment calls were logged

Each judgment call got a one-line note attached to its field, written so a reviewer who never saw the chart could follow it. The note for field 9 in this example reads: "Corvaxin (hives) from referral letter p. 2; ED note p. 4 says NKDA; recorded per rule 'any documented allergy'; conflict flagged." The note for field 5 reads: "Left 14:15 per nursing checklist p. 23; discharge summary signed 11:02 p. 21 is a document time."

Notes like these do three jobs. They let a second abstractor check the reasoning instead of repeating the search. They show a reviewer which rule was applied, so a wrong rule can be fixed once for every chart. And they keep the conflict visible after the form is submitted, so nobody later mistakes a settled field for an obvious one.

Mistakes this example avoids

Three errors are easy to make on a chart like this one. The first is taking the first value found and stopping, which here would have recorded NKDA and missed the referral letter's allergy. The second is dropping a stopped drug, which would have made field 11 read 2 and hidden the change at discharge. The third is citing a document instead of a page, which turns every later check into a search.

What would change in this example for a registry or audit?

The same chart would be abstracted differently for a registry or an audit, because each purpose defines its own fields, sources and rules. A registry wants coded values chosen by its data dictionary. An audit wants proof that values already submitted match the chart. The chart stays the same; the form and the rules change.

For a registry

A registry would replace the free-text diagnosis fields with codes from its own manual and would add fields this example skips, such as a diagnosis date chosen by a priority rule. Its data dictionary, not the abstractor's habit, would decide whether the outside referral letter counts as a source. How registries write and enforce those definitions is covered in clinical registry abstraction.

For an audit

An audit starts from the other end. The auditor already has the submitted values and asks the chart to prove each one. Here the source map above becomes the main evidence. A value whose page cannot be found is treated as unsupported, however likely it seems.

For a billing or records review

A billing review would care about the dates, the disposition and the documented services, and less about the medication details. A records request might need only the discharge summary pages. The purpose decides which of the 14 fields matter.

For a research study

A study would add its own fields and its own exclusions, and it would usually ask for a de-identified copy of the form. The medical record number and date of birth in fields 1 and 2 would be replaced by a study ID kept in a separate linking file. The source map would still be kept, because anyone checking the study data later will want to trace each value back to its page.

Whatever the purpose, the habit that makes the work reusable is the same: one value, one source page, one note for every judgment call. Schema Care applies that habit to medical records by returning each field with a link to its place on the page, along with a confidence score; the record fields it reads are listed under documents Schema Care reads.

All posts in Clinical Records and Abstraction