How Rhode Island Improved Healthcare Data Quality With Automated Record Matching

Description and use

In 2004, Rhode Island’s KIDSNET registry implemented an automated matching system using the open‑source tool Febrl, processing 95 % of records, adding ~11,000 patients, and reducing error resolution time from 3 min to 40 sec.

Problem

In 2004, Rhode Island’s KIDSNET child health registry had a backlog of 48,685 uncertain patient records that required manual review, delaying data availability and reducing system usefulness.

Outcome

The automated matching system cleared most held records, added ~11,000 patient records, and cut error resolution time from 3 minutes to 40 seconds, improving data availability for providers.

Workflow

• 1. Evaluate matching options and select Febrl.
• 2. Configure Febrl with rules and probability thresholds to classify records as auto‑match, new patient, or manual review.
• 3. Deploy the system in spring 2004.
• 4. Process incoming records: auto‑match where confidence high, flag others for manual review.
• 5. Review flagged records manually and update registry.
• 6. Monitor metrics: unresolved records, duplicate rates, missing fields, manual review workload.

Evidence

By 2004, the backlog of questionable matches had reached 48,685 records. The estimated time required to review them was 17 months, according to the case report.

Source

The new system went live in spring 2004. Its results were significant for the registry’s day‑to‑day work. It cleared most of the held records, added approximately 11,000 patient records and reduced the time needed to resolve an error.

Source

Rhode Island evaluated several matching options before selecting an open‑source tool called Febrl.

Source