AI Medical Document Classification for a Healthcare Practice

Statera Health is a medical practice based in Melbourne, Australia. The practice manages a high volume of incoming clinical correspondence and test results from hospitals, specialists, and pathology providers. Accurate and timely document processing is essential for maintaining complete patient records in their Xestro practice management system and supporting effective clinical decisions.

The administrative team at Statera Health spent considerable time each day handling incoming medical documents manually. Hospital and specialist reports frequently arrived as multi-page PDFs containing results for multiple patients in a single file. Staff had to:

• Determine where one patient’s report ended and the next began
• Classify each section (for example, ECG, blood test, discharge summary)
• Extract relevant details such as patient name, test date, and report type
• Rename each file to match the precise naming convention required by Xestro
• Upload the documents to the correct patient records

This process typically took 10–20 minutes per batch and occurred multiple times daily. Single-document files still required manual classification, renaming, and filing. The cumulative effect was hours of repetitive work each day, occasional misfiled documents, inconsistent naming that complicated future searches, and staff frustration from the tedious nature of the task. These issues sometimes delayed doctors’ access to test results and added to the overall administrative burden on the practice.

Osher Digital implemented an automated document classification and organisation workflow specifically configured for Statera Health’s document types and processes.

The solution handles documents from email inboxes, OneDrive folders, and a secure upload portal. It performs the following steps automatically:

• Filters out irrelevant content such as email signatures, logos, and non-medical attachments
• Processes password-protected PDFs and routes them for appropriate handling
• Uses a custom AI agent to analyse multi-page documents and identify boundaries between individual reports
• Classifies each document according to type (ECG reports, echocardiograms, Holter monitor results, pathology reports, referral letters, discharge summaries, patient questionnaires, and others)
• Extracts metadata including patient name, date of birth, test date, report type, and referring provider where available
• Renames each file to precisely match Xestro’s required naming convention
• Places the processed and renamed files into a designated OneDrive folder structured for easy bulk import into the practice management system

A human verification step allows staff to review classifications and splits for unusual or complex documents before final processing. The workflow integrates directly with Microsoft 365 and uses Azure Document Intelligence for reliable extraction and classification. Implementation included detailed mapping of the practice’s document formats, testing with historical files, team training, and post-go-live tuning based on real usage.