Uploading Patient Labs with an LLM

If your laboratory report is a PDF, image, or another format, you can use an LLM such as ChatGPT to convert the results into the LabNodes CSV format.

Before uploading anything: remove all personally identifying information from the laboratory report. Do not upload the patient's name, date of birth, address, medical record number, phone number, email address, or other identifying information.

Use a patient/formula ID instead—for example, PT-1042.

1. Download the LabNodes CSV template

Use the supplied LabNodes Marker Template CSV as the structure for the data. Do not change the column names or their order.

2. Upload the anonymized laboratory report

Upload the anonymized laboratory report to your LLM along with the LabNodes CSV template. Then use the following prompt:

You are helping prepare laboratory data for LabNodes.
I have attached:
1. An anonymized patient laboratory report.
2. The LabNodes CSV template.
Extract the laboratory results from the patient report and populate the CSV template.

Important instructions:
- Do not invent, estimate, or calculate laboratory values unless explicitly requested.
- Preserve the laboratory's reported value and unit exactly where possible.
- Preserve the laboratory's stated reference range.
- If a low or high value can be directly identified from the reported reference range, enter it in the low and high columns.
- If the laboratory provides a single clinical threshold rather than a conventional reference interval, preserve that information in reference_range and classify it appropriately in range_type.
- Use reference_interval, clinical_threshold, or context_dependent for range_type when appropriate.
- Enter the laboratory test date in the date column when available.
- If a field is not provided in the source report, leave it blank rather than guessing.
- Do not add patient names or other identifying information.
- Do not interpret, diagnose, or alter the laboratory results.
- Do not create additional columns.
- Return the completed data as a CSV using exactly the same column structure as the supplied template.
Before returning the CSV, check that every extracted result corresponds to the correct marker, value, unit, and reference information in the original report.

The output should contain only the completed CSV data.

3. Review the generated CSV

Before importing the file into LabNodes, review the extracted data against the original laboratory report. LLMs can make transcription errors, particularly with units, decimal points, reference ranges, and similarly named markers. The clinician remains responsible for verifying that the uploaded data accurately represents the original laboratory report.

Once verified, the CSV can be uploaded into LabNodes Clinical Mode for analysis and report generation.

The purpose of the LLM step is data extraction and formatting—not diagnosis or clinical interpretation.