Quick answer
Validate a SNF EHR migration at three levels: delivered objects, preserved relationships and usable destination workflows. Decide what belongs in active structured data versus a historical archive before conversion. Reconcile identifiers and versions, then test ordinary users retrieving representative cases. Equal counts and successful imports can still hide wrong-person links, false-current history or unusable documents.
Decide where the information must work
“Bring everything over” leaves the most important design decisions unresolved. For every object, ask whether staff must act on it, filter or calculate from it, or retrieve it as historical evidence. The answer determines the destination, cost and acceptance test.
The ONC Electronic Health Information Export test method addresses computable output and format documentation for applicable certified functionality. It does not establish compatibility with a receiving system. Obtain the actual export dictionary and a representative package before approving detailed conversion scope.
Use this original decision tree:
- Must the object drive current work? Have qualified owners define its current meaning, then test structured import and downstream behavior.
- Is structured historical analysis essential? Specify the required fields, versions and relationships; compare conversion effort with a documented alternative.
- Is faithful reference retrieval sufficient? Test an indexed archive or linked document through ordinary user permissions.
- Is meaning or retention unresolved? Preserve the source and route the decision to the qualified records, clinical or legal owner. A migration cutoff is not permission to destroy records.
Build the destination register
| Object | Example destination | Meaning that must survive | Consequential failed test |
|---|---|---|---|
| Person identifiers | Structured person plus source crosswalk | Identifier namespace and approved identity | Facility A's 00124 mapped to Facility B's different 00124 |
| Episodes | Separate structured stays | Person versus stay; actual dates | Prior-stay document appears current |
| Historical medication information | Reviewed historical view/archive | Order, administration or received history; source/status/date | Imported history enters an active list without approved review |
| Historical MDS | Versioned history or supported archive | Assessment ID, revision, source status and episode | Local “complete” label becomes apparent current submission acceptance |
| Scanned documents | Indexed document store | Source file, pages, metadata and links | File opens but belongs to the wrong episode |
| Configuration history | Reference definitions | What old fields and reports meant | Earlier “reviewed” value reinterpreted under a new definition |
These are hypothetical designs, not clinical activation rules or an MDS coding guide. The destination can differ by object and source period. Preserving a meaningful historical representation may be better than forcing every field into the new system's current schema.
Use a specimen that defeats count-only validation
Prepare the following synthetic package and save expected results before import:
| Source object | Expected destination |
|---|---|
| A-104 and retired alias A-087 | One approved person R9; both source references retrievable |
| B-104 | Different person R10; namespace preserved |
| A-220 ambiguous candidate | Unresolved review item; no automatic merge |
| E1 and E2 for R9 | Two distinguishable episodes |
| D1 linked to E1; D2 linked to E2 | Two readable documents retaining their respective episode links |
Now deliberately swap D1 and D2's destination episode links. There are still two documents, two episodes and the expected number of people. Every count can match while both document relationships are wrong. The expected result is a failed relationship check and an investigation of the mapping rule, not a passing completion percentage.
Add a second deliberate fault: export D1 twice and omit D2. File count remains two; uniqueness by original object ID detects the problem. These two faults prove that uniqueness and relationship tests answer different questions.
Separate package acceptance from destination acceptance
Package checks cover source scope/cutoff, object IDs, expected counts, file identity, dictionary version, readability and relationships present in the export. Destination checks cover what the receiving software does with those objects. Record the exact package used for each rehearsal so a later corrected extract cannot be mistaken for the one already tested.
For attachments, open multi-page and older-format examples using intended staff permissions. Search by source date and document type, not just the new import timestamp. A file accessible to a technical administrator in storage is not proof that the nurse or health-information reviewer can find it.
For merged records, preserve approved aliases separately from unresolved duplicate candidates. Search using a retired identifier. Its expected result is the approved surviving relationship, not a new empty record. A manual identity decision needs protected evidence and a versioned crosswalk so a later extract does not overwrite it with a generic match rule.
Test historical meaning and date boundaries
The 2025 SAFER System Management guide discusses configuration, mapping and validation across systems. The following original cases make those risks observable:
| Case | Known input | Expected output |
|---|---|---|
| Date-only | Synthetic birth date 1950-01-02 | Same calendar date; no timezone-driven shift |
| Offset-aware event | 2026-11-01T01:30:00-04:00 and 2026-11-01T01:30:00-05:00 | Two instants one hour apart, despite same clock label |
| Late entry | Event 20:00; entry 22:00 | Both meanings retained; import time does not replace either |
| Historical medication object | Prior episode and source type retained | Visible as history under the reviewed design, not newly activated by import |
| Revised assessment | Original and corrected source versions | Reviewer can distinguish both and trace actual source status |
Test display, sorting, reports and re-export. A correct-looking screen can conceal a report that groups events by the wrong time basis. If a source timestamp lacks an offset, obtain its documented convention or preserve the uncertainty; do not invent precision.
Make the archive tradeoff explicit
An independent facility might preserve older assessments in an indexed archive because structured conversion needs specialist time it cannot provide. That choice needs a demonstrated retrieval route, access support and a priced arrangement. It cannot rest on “the old system will probably remain available.”
An enterprise can reuse a common mapping but must test local identifier namespaces and source variants. Suppose 10 sites use one mapping and two have different historical status definitions. Testing only the common source would leave the highest-risk two sites unexamined. Allocate review by variation and consequence, not just by facility count.
The final migration decision should list tested populations, unresolved relationships, approved archive destinations, affected workflow limits and the correction path. Keep ambiguous cases visible. Importing fewer objects with faithful meaning and retrievable history may be a stronger result than importing every row into misleading active fields.
Questions & answers
Should all historical information be imported into active fields?
Only where the intended use and destination meaning have been reviewed. A historical status can become misleading when mapped to a current workflow. An accessible archive may be preferable when transformation would lose context.
How large should the migration sample be?
There is no universal sample size in this guide. Combine full-population automated relationship checks with targeted cases selected for source risks and actual record types. A systematic failure requires investigation of its entire affected population, not repair of the sampled item alone.
Can a correct person count establish correct resident identity?
No. A retired alias, a duplicate candidate and a different facility’s reused local number can produce similar totals with different meanings. Reconcile people, aliases and unresolved candidates separately and inspect their linked episodes.
Primary sources
Use the cited source and its applicable version when making a facility-specific decision. These guides explain operational evaluation methods; they are not individualized clinical, coding, or legal advice.