HEDIS measurement pipeline

The problem

Producing HEDIS measurements requires pulling clinical documents and claims from many sources, classifying them, normalizing fields, and emitting numerators and denominators that an auditor can reproduce. Manual triage in spreadsheets failed at scale. Previous automation had race conditions that lost orders, no diagnostics for stuck items, and no Slack-native incident routing for the on-call team. Audit reproducibility was a quarterly fire drill, not a property of the system.

The approach

A fleet of .NET background services on Windows orchestrate the document and claim flow. Fax gateway intake feeds document classification and routing; an insurance portal sync handles claim retrieval. Retry queues isolate transient failures from permanent ones; stuck-order diagnostics surface anomalies before they become outages. A monitoring dashboard reads from MSSQL and New Relic, paging on-call via Slack with the order context attached. Every state transition writes an audit row; replaying from raw inputs reproduces the final measure.

Stack and engineering choices

• .NET Windows background services
• MSSQL transactional core
• New Relic instrumentation
• Slack-native incident routing
• Retry queues with isolation
• Stuck-order diagnostics
• Audit-grade order trail

Outcome

HEDIS submissions reproduce on demand from raw inputs to final measure. Stuck orders surface in minutes, not days. The on-call team handles incidents from Slack without dropping into a separate ticketing system, and audit cycles complete without rebuilding the pipeline state by hand.