Run a Connected Health Record Pipeline Incident Without Guessing at the Data

Short summary

An unexecuted incident runbook for connected health-record pipeline failures, emphasizing privacy-safe debugging that tracks opaque identifiers instead of health values. It covers the first fifteen minutes of response, a diagnostic decision table for source timeouts vs. schema faults vs. authorization faults, and a gradual recovery sequence with consent revalidation. The runbook is a generic artifact motivated by OpenAI's Health in ChatGPT announcement.

• •First-fifteen-minutes protocol: assign roles, pause imports, capture metadata, block dead-letter replay
• •Diagnostic table distinguishes upstream failures, compatibility faults, authorization faults, freshness faults, and containment gaps
• •Recovery sequence validates consent snapshots and canaries synthetic fixtures before gradually expanding access

Generated with AI, which can make mistakes.