Automate Caregiver Scheduling and Visit Reporting

Many home health agencies try using off-the-shelf scheduling software like WhenIWork or Homebase. These tools are designed for shift work and treat every employee as an interchangeable resource. They can show who is available but cannot intelligently match a patient needing bilingual catheter care with a certified, Spanish-speaking caregiver who is 10 minutes away.

A typical failure scenario involves a scheduler at a 12-caregiver agency. A new patient intake form arrives as a scanned PDF via email. The scheduler must manually read the PDF, identify the patient's specific needs, check their address on Google Maps, then cross-reference a separate spreadsheet of caregiver skills, certifications, and current locations. This multi-tool, manual process takes 20 minutes per patient and results in a 15% error rate from misread forms or outdated spreadsheets.

The fundamental problem is that generic tools cannot model the complex constraints of home health care. They lack a data structure that connects patient needs, caregiver qualifications, real-time location, and continuity of care. This forces schedulers to perform the actual matching logic in their heads, rendering the software little more than a digital calendar.

Syntora's approach to workflow automation for home health agencies would start with an assessment of your patient intake sources. This often involves processing email inboxes with PDF attachments. We would develop a Python script using the `pdfplumber` library to extract raw text and data from these forms. The extracted text would then be processed by the Claude API to classify patient needs into structured fields, such as `skill_required: 'Hoyer Lift'` and `language: 'Spanish'`. This structured data would be stored in a HIPAA-compliant Supabase Postgres database.

The scheduling engine would be implemented as a FastAPI service deployed on AWS Lambda. When a new visit needs scheduling, this service would query the Supabase database for caregivers based on required skills and location proximity, such as a 15-mile radius from the patient. The engine would use the Google Maps API to calculate drive times and filter out unavailable caregivers, then present suitable matches to a human scheduler for review and confirmation.

For visit note reporting, the system could allow caregivers to submit notes via SMS. An AWS Lambda function would be triggered by this message, sending the note content to the Claude API with a specific prompt to extract key information: visit duration, tasks performed, and any patient status changes. This structured data would be written back to Supabase, creating an audit-ready record. This method aims to significantly reduce the time spent on administrative reporting tasks.

The delivered system would be deployed within your own AWS account to ensure full HIPAA compliance and data ownership. Syntora would configure monitoring using CloudWatch for API health and data quality. Custom alerts could be set up to send notifications, for example, via Slack, if API response times exceed defined thresholds or if note-parsing failure rates increase, allowing for proactive issue resolution.