Clinical data abstraction
Clinical data abstraction that starts from the coded record.
Clinical Extract fills every coded field from the EHR first. Your abstractors spend their hours on the chart review that needs a person.
Diagnosis codes, dates, results, medications and procedures, filled from the EHR with the FHIR source next to each value.
1 Definition
What is clinical data abstraction?
Clinical data abstraction is the work of reading a patient's record and recording defined data elements in a structured form, for a registry, a quality measure or a study. Medical chart abstraction, medical record abstraction and chart abstraction name the same work. Every layout has two kinds of element: the coded ones (diagnosis codes, dates, laboratory results, medications, procedures, staging) that the EHR holds as structured data, and the judgment calls (a date that is a reading of a note, an ambiguous stage, a cause) that take a person.
Coded elements make up most of the fields and most of the lookup time. Clinical Extract exports them from the EHR's certified bulk FHIR API, one column per element with its FHIR source. The abstraction form opens filled, and the abstractor starts at the fields that need a person.
2 The split
Coded fields and judgment calls in one breast case.
Take one breast case abstracted to the NAACCR layout. The diagnosis and its date, the histology, the biomarkers with their LOINC codes, the stage group, the first course of treatment with its RxNorm codes and dates, the procedures. Clinical Extract fills all of these from coded EHR data. The six fields that take a reading of the notes (a date of first contact that depends on which note counts, a stage the pathology and the imaging disagree on) stay with the abstractor, who now starts there.
3 Minutes per case
Medical chart abstraction with the coded fields pre-filled.
An abstractor's minutes on a case go to finding values, checking them and deciding the calls. With the coded fields pre-filled, there is nothing to find: the values are in the form, each with the FHIR element it came from, and the abstractor's time goes to checking them and making the calls. Figure 2 draws the minutes for the same breast case, before and after.
Figure 2 as a table
| Task | Manual lookup, min | With Clinical Extract, min |
|---|---|---|
| Judgment calls | 11 | 11 |
| Finding the case records | 9 | 0 |
| Diagnosis and histology lookup | 8 | 0 |
| Biomarker lookup | 6 | 0 |
| Stage lookup | 7 | 0 |
| Treatment lookup | 12 | 0 |
| Follow-up lookup | 5 | 0 |
| Checking the pre-filled fields | 0 | 7 |
| Minutes per case | 58 | 18 |
4 Software, services, or both
Clinical data abstraction software and services.
Clinical data abstraction services send abstractors; clinical data abstraction software gives them a form and a workflow. Clinical Extract feeds both. A service that starts from a filled case bills fewer hours per case; software that imports study.csv opens each case pre-populated, with the dictionary as the field map. Either way the coded fields come from the EHR as recorded, and the abstractor keeps the decisions that need the notes.
5 Quality and registry abstractors
Filling SEP-1, NAACCR and NCDR elements from the EHR.
A sepsis bundle review (SEP-1) turns on times: the lactate result, the antibiotic order, the fluid start, each a timestamped coded element in the EHR, and one judgment call, time zero, which is a reading of the note. A cancer registry case is the NAACCR items; a cardiac registry case is the NCDR elements. For every one, the coded elements arrive as columns with their sources and the abstractor's review starts at the calls. The registry page shows the crosswalk from the EHR to the NAACCR items; running a chart review study shows the same split for a research team.
6 Questions
Questions about abstraction
What does a clinical data abstractor do?
Reads a patient’s record against a defined layout (a registry’s data items, a quality measure’s elements, a study’s variables) and records each element in a structured form: dates, codes, results, and the judgments the layout asks for. With the coded fields filled from the EHR, the abstractor’s hours go to checking values and deciding the ambiguous fields.
What is data abstraction in healthcare?
The step between the record and the report: turning what the chart holds into the defined elements a registry, a measure or a study needs. Most of those elements are already coded in the EHR; the rest take a reading of the notes.
What is medical record abstraction?
Medical record abstraction is reading a patient’s chart and copying the facts a registry or study needs into set fields. Clinical Extract fills the coded fields from the EHR first, so the abstractor reviews them and adds the judgment calls.
Is there an AI tool that reads medical records?
AI tools that read clinical notes exist. Coded data does not need one: diagnosis codes, dates, results, medications and procedures come out of the EHR as they are. Clinical Extract delivers those values as the EHR holds them, with the FHIR source next to each value, and leaves the notes to your abstractors and whatever reads them.
How much time does it save per case?
The record lookups drop out. Figure 2 shows the minutes for one breast case abstracted to the NAACCR layout, before and after; the minutes that remain go to checking values and reading the notes.
Does it replace abstractors?
It fills the coded fields. Abstractors keep the judgment calls, the note review and the sign-off.
Which layouts does it fill?
Any layout whose elements are coded in the EHR: NAACCR data items for cancer registries, NCDR elements for cardiac registries, the timed elements of a quality measure such as SEP-1, a study’s variable list. The data dictionary maps each column to the element it fills.
Next
See Clinical Extract run on one of your studies.
Tell us which EHR you run and what the study or registry needs. We reply within one business day to set a meeting time.