REF / AUTOMATION

Toronto Healthcare Network - Discharge Summary and Claims Automation

Automated discharge summary drafting, insurance claim submission, and appointment routing for a 7-clinic Toronto healthcare network.

RoleAutomation Architect
Year2025
OutcomeClaim cycle 14d → 3d, denial rate −58%
DomainAutomation
00
STACK

Tech used.

Claude Opus 4.6Epic FHIR APIsn8nAzureOpenAI Whispersecure storage

The Problem

A 7-clinic multi-specialty network with 240 staff in the Toronto area was struggling with two adjacent problems. Physicians were spending 90+ minutes a day on documentation. particularly discharge summaries that had to satisfy clinical, billing, and provincial reporting requirements. Insurance claims were taking 14 days on average from service to submission, and denial rates were running at 22%. most denials traced back to documentation issues that could have been caught earlier.

The board wanted physicians seeing patients, not typing notes. The CFO wanted denial rates cut in half.

What I Built

A clinically-careful automation stack that integrates with the existing Epic-based EHR via FHIR APIs:

1. Discharge summary draft assistant. After each patient encounter, the physician reviews the AI-drafted summary built from: visit notes, prior history, lab results, medications, and the procedure record. The AI produces a structurally-compliant draft (provincial reporting format, billing codes, plain-language patient version). Physician edits typically take 3-5 minutes vs 25-40 minutes for a from-scratch summary.

2. Claim pre-flight check. Before any claim is submitted, an automation runs the documentation through a denial-prediction layer: missing modifiers, unclear medical necessity language, code combinations that historically denied, signature/sign-off gaps. Issues are flagged to the relevant physician with one-click resolution suggestions.

3. Appointment routing intelligence. New patient referrals are auto-triaged by Claude. extracting symptoms, urgency, and suggested specialty from the referral letter. and routed to the right clinic + practitioner, with suggested visit duration. Reduced 14-day average referral-to-booking to 3 days.

4. Patient-facing automation. Pre-visit instructions, post-visit care summaries, and follow-up appointment scheduling all run through automated patient communications respecting Ontario's PHIPA regulations.

−85%
Discharge summary time
32m → 5m
−79%
Claim cycle time
14d → 3d
−58%
Denial rate
22% → 9%
+40min
Physician daily face time
Operations Before vs After
MetricBeforeAfterΔ
Avg discharge summary draft time32 min5 min−84%
Claim submission cycle14 days3 days−79%
First-pass claim acceptance78%91%+13 pts
Physician daily documentation hours1h 30m20m−1h 10m
Referral-to-booking time14 days3 days−79%

Outcome

Physicians reclaimed roughly 70 minutes a day each. the network is now seeing more patients per day with the same physician roster, and clinical satisfaction scores have climbed. The CFO's denial rate target was beaten (−58% vs target −50%). The provincial reporting compliance rate hit 100% for two consecutive quarters for the first time in the network's history.