Reducing Patient Intake Time by 68%

A 200-bed regional hospital network was drowning in manual intake processes. Every new patient required staff to manually transcribe referral documents, verify insurance by phone, and search the EMR for existing records. The process took an average of 45 minutes per patient - and errors were common. Duplicate records were created 12% of the time. Insurance verification backlogs meant patients sometimes waited days for confirmation.

The network had evaluated three off-the-shelf solutions, but none could handle the variety of referral document formats they received from over 400 referring physicians.

We spent the first three days embedded with intake staff, observing the actual workflow and cataloging the 23 distinct referral document formats in circulation. The system we built uses Claude Opus 4.6 for document parsing - extracting patient demographics, diagnosis codes, referring physician details, and insurance information from scanned and digital referral documents.

A custom rules engine handles insurance verification logic, cross-referencing extracted policy numbers against payer databases in real-time. For patient matching, we built a fuzzy matching algorithm that compares against the existing EMR - but critically, the system flags uncertain matches for human review rather than auto-matching. The principle: AI handles the volume, humans handle the judgment calls.

The output is a pre-filled intake form that staff can verify and submit in under 3 minutes, compared to the previous 45-minute manual process.