Data Migration Validation Before Cutover
A clinic is migrating from a legacy EHR to a new system. The migration includes 14 years of patient records, problem lists, medication histories, and allergies. The vendor's migration report shows a 99.2% 'record match' rate. A sample audit of 200 charts by two analysts finds 11 charts where the allergy list in the new system is missing at least one entry present in the legacy system. The vendor says the 99.2% figure 'exceeds industry standard' and that the missing allergies are 'likely inactive or duplicate entries.' Cutover is in 9 days.
1. Evaluate the vendor's 99.2% claim and the sample audit findings. What does the evidence actually tell you?
Read the worked answer
The 99.2% figure and the audit findings measure different things, and the vendor is conflating them.
- 'Record match' likely means a patient record was created and some data mapped - not that every clinical field is complete. A chart can 'match' and still have missing allergies.
- The audit found 11/200 = 5.5% of sampled charts with at least one missing allergy. That is a patient-safety signal, not a rounding error. If the sample is representative, that implies a meaningful number of charts with incomplete allergy data across 14 years of records.
- 'Likely inactive or duplicate' is speculation, not evidence. Allergies should not be assumed inactive without clinical review - a documented reaction matters even if the allergen is no longer prescribed.
Conclusion: The 99.2% claim does not rebut the audit. The audit is the more clinically relevant measure and it shows a defect the vendor has not explained.
2. What validation would you demand before cutover, and what is your go/no-go recommendation?
Read the worked answer
Validation demands:
- Stratified sample by year of record and by record type (allergies, meds, problem list) - 200 charts is too small and possibly unrepresentative for a 14-year migration.
- Root-cause the 11 misses. Are they a mapping rule error, a field-type conversion issue, or a source-data problem? The answer determines whether the fix is a re-run or a manual cleanup.
- Targeted allergy reconciliation for all patients with active medications or upcoming appointments, prioritized by risk.
- Independent verification by the clinical informatics team, not only the vendor.
Go/no-go: Recommend conditional no-go for full cutover in 9 days unless (a) root cause is identified and corrected, (b) a re-migration or reconciliation pass is completed and re-audited with a passing threshold (e.g., <1% missing allergy entries in a stratified sample), and (c) a documented manual reconciliation plan covers any residual gaps.
A 9-day deadline does not override allergy safety. Propose a phased cutover or a delay with a clear remediation date.
3. Leadership is worried about the cost and disruption of a delay. How do you frame the decision?
Read the worked answer
Frame it as risk exposure, not perfectionism:
- Quantify the risk. If 5.5% of charts have a missing allergy, and the clinic sees N patients per week, estimate the number of encounters per week where a clinician could prescribe against an incomplete allergy list. That is the concrete exposure.
- Compare costs. A short delay (2-3 weeks) for reconciliation versus the cost of a serious adverse drug event, a patient-safety investigation, and the trust damage of a public failure. The asymmetry is stark.
- Offer options, not just 'no.' Present: (a) full delay, (b) phased cutover with high-risk patients reconciled first, (c) cutover on schedule with a mandatory allergy-confirmation step at every encounter until reconciliation completes. Let leadership choose with eyes open.
- Document the recommendation and the decision. If leadership overrides, the record shows the risk was surfaced and who accepted it. That protects patients, the project, and you.