Med Reconciliation Alerts After CPOE Go-Live
A 220-bed community hospital goes live with a new CPOE and medication reconciliation module. Two weeks in, the pharmacy reports that 38% of admission med recs are completed more than 24 hours after admission, versus a pre-go-live baseline of 9%. Nurses say the med rec screen opens only after they finish a separate triage assessment, and the 'time last dose' field rejects free text, forcing them to guess from patient recall. Prescribers are overriding duplicate-therapy alerts at 91%. The CMIO asks you to diagnose and propose a plan, with a 6-week window before the next scheduled release.
1. Is this primarily a training problem, a workflow design problem, or a clinical decision support problem? Justify your reasoning using the evidence in the setup.
Read the worked answer
The dominant issue is workflow design, with CDS as a secondary contributor.
- The med rec screen's dependency on completing triage first is a sequencing/design constraint, not user knowledge. Training cannot fix a screen that opens at the wrong point in the encounter.
- The 'time last dose' field rejecting free text is a data-entry design mismatch with the real clinical situation (patients often don't know exact times). Forcing guesses produces low-quality data that then feeds downstream logic.
- The 91% duplicate-therapy override rate is the CDS symptom, not the root cause. High override rates typically follow alert fatigue, which is itself often downstream of poor data quality and mistimed prompts.
A useful framing: training addresses 'don't know how'; this is 'know how but blocked by the system.' The fix should be workflow-first (reorder the screen, allow structured approximations with a 'patient unsure' option), then re-evaluate CDS firing before touching alert thresholds.
2. Draft a prioritized 6-week remediation plan. What do you fix before the next release, and what waits?
Read the worked answer
Weeks 1-2 - Diagnose and stabilize
- Pull screen-level analytics: time-in-screen, abandonment, field-level error rates on 'time last dose.'
- Shadow 5-8 nurses across shifts to observe the actual sequence.
- Convene a small workgroup (2 nurses, 1 pharmacist, 1 hospitalist, analyst, you) to confirm root causes.
Weeks 2-4 - High-impact configuration fixes (pre-release)
- Reorder the med rec screen so it can be launched independently of triage completion. This is a configuration change, not a build.
- Add 'patient unsure / approximate' structured options to the time-last-dose field.
- Instrument the duplicate-therapy alert to log which fields triggered it, so you can see whether bad data is driving the overrides.
Weeks 4-6 - Evaluate and plan CDS
- Re-measure med rec timeliness and override rate 2 weeks post-change.
- Only after data quality improves should you tune duplicate-therapy logic (e.g., suppress alerts when last-dose data is 'unsure').
Do not promise alert-threshold changes in this window - you'd be treating a symptom before the data is trustworthy.
3. The CMIO wants to 'just turn off the duplicate-therapy alerts' to reduce override noise. What is your response?
Read the worked answer
Push back, but constructively. A 91% override rate is a real signal, but 'turn it off' trades alert fatigue for missed safety events. Recommend instead:
- Measure first. Break the override rate down by drug pair, prescriber, and trigger field. A 91% aggregate may hide a small set of pairs that fire constantly and a small set that are appropriately overridden.
- Target the top offenders. Suppress or downgrade the highest-volume, lowest-value pairs rather than the whole rule.
- Tie to data quality. If overrides correlate with 'unsure' last-dose entries, the alert is firing on bad inputs - fix the input before the rule.
- Set a re-evaluation date. Agree on a threshold (e.g., target override rate <70% within 60 days) so the CMIO has a measurable commitment, not a refusal.
Document the decision and rationale - this is exactly the kind of change that should go through the pharmacy and therapeutics or safety committee.