Cutover Under Pressure: Replacing a Hospital EHR
Northbridge Regional Hospital (420 beds) runs a 14-year-old electronic health record whose vendor ends support in nine months. The replacement is contracted, but acceptance testing found that 2.3% of migrated allergy and medication records mismatch the legacy system, and the laboratory results interface trails the vendor's reference implementation by 11 seconds. The hospital handles 3,200 outpatient appointments and 190 emergency admissions per week, and clinicians state a maximum tolerable downtime of 30 minutes for medication administration. The board is choosing between a 'big-bang' weekend cutover and a phased rollout by department over five months. The phased option costs an extra $410,000 in dual-running licences and temporary staffing, while big-bang risks a single failure affecting every ward at once. The IT director notes that the hospital has never run two EHRs concurrently, and that 70% of nurses surveyed have had only four hours of training on the new system.
1. Recommend a cutover strategy, justifying it against at least three explicit criteria.
Read the worked answer
The relevant criteria are patient safety (30-minute downtime tolerance), data integrity (2.3% mismatch), organisational readiness (four hours of training for 70% of nurses) and cost ($410,000). A big-bang cutover concentrates all risk into one weekend: if the allergy mismatch affects even a fraction of the suspect records, every ward is exposed simultaneously and there is no unaffected system to fall back on. Phased rollout caps the blast radius to one department and lets later departments benefit from earlier lessons, but dual-running two EHRs is itself a safety hazard, because medication reconciliation across two systems is error-prone, and the hospital has no experience of it. The $410,000 is small against the cost of a serious medication error or a prolonged outage. Given that readiness is the weakest link, the strongest answer is a phased rollout beginning with a low-acuity, high-volume area such as outpatient clinics, with emergency and theatres last, plus a written rollback trigger (for example, more than five medication discrepancies per shift). Credit should go to answers that name the trade-off rather than assert that one option is risk-free.
2. Explain how the 2.3% mismatch rate should influence the decision, and describe one control to mitigate it.
Read the worked answer
A percentage alone hides the absolute number. If the migration covers roughly 1.2 million records, 2.3% is about 27,600 records, and mismatches are unlikely to be randomly distributed: allergy and medication fields are exactly where an error can kill. The correct reasoning is that risk is a function of both frequency and severity, so a 2.3% rate in the demographics field might be tolerable while the same rate in allergies is not. It should therefore block go-live until a verified subset is clean, not simply be logged as a known issue. One concrete control is a two-pass reconciliation in which any migrated allergy or medication record is flagged in the new system and must be confirmed by a clinician before prescribing is enabled for that patient; this converts a silent error into a visible prompt. A stronger variant adds an independent 'golden record' comparison against a manually verified sample to estimate the true error rate and its confidence interval.
3. Identify two stakeholder groups and analyse how the chosen strategy affects each.
Read the worked answer
Nurses are the largest operational stakeholder. A phased rollout spreads their training burden and lets them learn in a live but bounded environment, whereas big-bang forces every nurse to be proficient on the same weekend; with only four hours of training, that is a direct safety risk. However, nurses in later phases face months of dual-running, which means duplicate documentation and higher workload, so the strategy shifts risk rather than removing it. Laboratory staff are affected through the 11-second interface lag: in a phased model the lab must feed both the old and new systems, doubling the chance of a misrouted result, so the phased plan needs an explicit interface reconciliation step. A good answer should note that patients bear the residual risk in either model and that their consent to migrated data being reused should be handled transparently. A weaker answer will list stakeholders without tracing how the strategy changes their work.