Duplicate patient handling during EHR data migration is best understood as a sequence of ownership decisions rather than a single deduplication feature. The source materials describe distinct stages: identifying potential duplicates before conversion, executing merges with defined record ownership, and validating that downstream systems receive the corrected identity. A VA patient merge manual specifies that the name and Social Security number from the FROM record move into an ALIAS subfile in the TO record, which shows that merge mechanics preserve traceability instead of simply deleting one record [2]. A hospital merger case study describes merging confirmed duplicates in live systems before go-live to reduce the number of duplicate medical records [4]. These examples indicate that duplicate handling requires explicit rules for which record survives, what happens to identifiers, and how linked data follows the surviving record. The practical challenge is coordinating those rules across an enterprise master patient index, the EHR, and ancillary systems.
Record Ownership and the Merge Decision
A merge operation requires a clear answer to which record is the TO record and which is the FROM record. The VA Patient Merge manual describes moving the Name and Social Security Number fields from the FROM record into the ALIAS subfile of the TO record [2]. This preserves the discarded identifier as an alias rather than erasing it, which matters when historical documents or lab results still reference the old number. The manual also lists status codes that track the merge process, including POTENTIAL DUPLICATE, UNVERIFIED, VERIFICATION IN PROCESS, VERIFIED DUPLICATE, and VERIFIED, NOT A DUPLICATE [2]. These statuses indicate that ownership is not assigned automatically; a reviewer must confirm the relationship before a merge proceeds.
The AHIMA practice brief frames duplicate handling as part of data integrity policies that should include registration processes, standards for handling duplicate records, and processes for addressing overlays [7]. An overlay occurs when one person's demographic information is written over another person's data, which is a different failure mode than a simple duplicate. The SAFER guide similarly distinguishes duplicate records from commingled or overlay records and recommends using an enterprise master patient index to identify patients with existing records [1]. The ownership question therefore extends beyond choosing a surviving record: the organization must also decide who is authorized to initiate a merge, what evidence is required, and how the decision is documented for later audit.
Pre-Migration Cleanup and Source of Truth
The Northeast Alabama Regional Medical Center case study describes a phased project that began with analysis and design in May 2017 and ended with go-live in July 2018 [4]. During the action phase, staff established a subset of EMPI data including medical record number, Social Security number, last name, first name, middle name, date of birth, address, age, and sometimes maiden name from both hospital systems [4]. They determined which data subset contained the latest demographic data for each patient and used a scripting tool to verify data comparisons against a first algorithm report [4]. The case study recommends doing as much cleanup work on the EMPI as possible before merging any data [4].
The same case study introduces the concept of a single source of truth, stating that any and all changes to EMPI data should be made at the SSOT point [4]. This is a governance rule, not a technical feature: it means demographic corrections should not be entered independently in multiple systems. The CDC de-duplication best practices document describes a related objective for immunization information systems, where the goal is to correctly match all records related to the same patient even when there are variances in the data used to establish identity [3]. The document notes that discrepancies can arise from different data entry conventions, typing or spelling errors, extra or omitted data, non-standard abbreviations, and misreported or deliberately corrupted data [3]. Pre-migration cleanup therefore involves both algorithmic matching and human review of the exceptions that algorithms flag.
A Suggested Validation Scenario for a Two-Hospital Migration
A useful hypothetical validation scenario would begin with a pre-migration extract of both facilities' EMPI subsets, including medical record number, name, date of birth, and Social Security number where available [4]. The team would run an algorithm report to identify potential duplicate pairs, then have a primary reviewer or HIM staff confirm each pair using the status workflow described in the VA manual [2]. After confirmed duplicates are merged in the live systems before go-live [4], the team would re-extract the EMPI and compare the count of remaining potential duplicates against the pre-cleanup baseline. The validation would also check that the FROM record's identifiers appear in the TO record's alias subfile [2] and that downstream systems show the surviving record rather than the discarded one [8]. This scenario is a suggestion for planning purposes; it does not describe an independently tested procedure.
Merge Mechanics and Downstream Systems
A merge in the EHR is not complete until downstream systems reflect the same decision. The Epic migration account from e4.health describes a New Jersey health system migrating Epic into 14 hospitals and 1,000 providers, with nine different EHR systems and approximately 650 downstream systems including labs, radiology, and respiratory [8]. The project involved a detailed decision flow process to ensure that all downstream systems were accounted for and appropriately matched with the EHR and EMPI [8]. Project leaders considered each facility, each ancillary system within that facility, and whether the downstream system overlapped across multiple facilities within the enterprise [8]. The reported project results include reconciled enterprise-level overlaps, resolved same-source facility-level duplicates, and merged or validated 650 downstream systems [8].
The VA manual provides a more granular view of what happens inside a merge. The FROM record's name and Social Security number move into the ALIAS subfile of the TO record [2]. The manual also indicates that users can no longer select PATIENT records once a merge process is active, and that the system does not add records to File #15 during certain operations [2]. These constraints suggest that merge mechanics include locking records to prevent concurrent edits while the merge is in progress. The manual's status codes, including REQUIRES RESOLUTION and VERIFIED DUPLICATE, imply a workflow where a potential duplicate pair moves through verification before the actual merge executes [2]. Organizations planning a migration should ask whether their target system exposes equivalent status tracking or whether that workflow must be managed outside the EHR.
Validation and the Limits of Automation
The AHIMA practice brief states that testing of data migration with validation needs to occur in a test environment before migration should occur [7]. It also recommends that prior to retiring a legacy system, a thorough assessment of stored data must be undertaken and a data mapping plan to migrate required data elements must be developed [7]. This positions validation as a prerequisite, not a post-implementation afterthought. The brief further notes that when an error in data is discovered, the error should be corrected in the source system when possible [7]. That guidance has direct implications for duplicate handling: if a duplicate is found after migration, the correction should propagate back to the source of truth rather than existing only in the target EHR.
The CDC document distinguishes between automated matching and human review, stating that human review can be complex and that the ability of a system to make these decisions both accurately and on an automated basis determines its efficiency [3]. The document also defines undetected duplicates as the total number of duplicate records that were added as new patients instead of being properly matched to an existing patient [3]. This metric is useful for validation because it measures the failure mode that matters most: a duplicate that bypasses matching entirely. A validation scenario should therefore count resolved duplicates and records that were never flagged as potential duplicates in the first place.
Sources & references
- Patient Identification · www.healthit.gov
- [PDF] PATIENT MERGE USER MANUAL Version 7.3 April 1998 - VA.gov · www.va.gov
- Patient-Level De-duplication Best Practices — February 6, 2025 · www.cdc.gov
- Overcoming Challenges of Merging Multiple Patient Identification and Matching Systems: A Case Study - PMC · pmc.ncbi.nlm.nih.gov
- Electronic Health Records - Health IT Playbook · playbook.healthit.gov
- EHR for HIM: Duplicate Patients and Merge Process · www.ihs.gov
- [PDF] Practice Brief: Ensuring the Integrity of the EHR - Journal of AHIMA · journal.ahima.org
- Preserving Data Integrity During Epic Migration: A Tale of Two Systems · www.e4.health