A Potassium of 6.1 Filed After Discharge

The two methods answer different questions. Root cause analysis is retrospective and event-specific: it reconstructs what happened and why, and produces causal statements with an action plan. Failure mode and effects analysis is prospective and process-wide: it enumerates ways a process can fail and ranks them by severity, occurrence, and detectability so effort goes to the highest-risk failure modes before harm occurs.

Why this case needs both An RCA on this event is appropriate because harm occurred and the timeline is reconstructable: the discharge order, the result verification, the routing rule, the inbox state, and the coverage arrangement. But the QI data show the failure mode is already common: 11% of discharge-day labs have no owner, and 38% of results sit unopened beyond 24 hours. That is not one-off; an RCA alone would produce local fixes to a problem occurring many times each week.

Scope of the RCA

• Minute-by-minute timeline from lab draw to readmission.
• Interviews with the nurse, hospitalist, covering clinician, lab, and unit clerk, framed to learn the constraints they were working under rather than to assign blame.
• Causal tree with branches for policy, technology, staffing, and culture. "The hospitalist did not check" is an observation, not a cause; the cause is a system that routed a routine-flagged result to a shared pool with no owner while the responsible clinician was off.

Scope of the FMEA

• The full abnormal-result lifecycle: order, collection, verification, routing, acknowledgment, escalation, and post-discharge follow-up.
• Score each failure mode. A result routed to an unowned pool is high severity, high occurrence (11%), and poorly detected; that ranking is where the fixes go.

Sequence. Complete the RCA within the required internal window and share preliminary findings, then run the FMEA so the corrective actions are designed against the whole process rather than patched to one patient.

System-level factors, supported by the data

• Critical threshold set at 6.5. A potassium of 6.1 in a patient with renal risk is clinically significant. The threshold should be reviewed against the laboratory's own validation and medical director sign-off; if it diverges from peer institutions or from the clinicians' expectations, the setting is a configuration failure, not a vigilance failure.
• A "discharge pool" with no assigned clinician (11%). A result without an owner has no one accountable for acting. This is the most consequential finding in the dataset.
• Discharge order and pending results not reconciled. Nothing in the process requires the discharging clinician to review unresulted tests and name a responsible party.
• Routine flag applied to a value that is not routine. Routing logic treated a clinically actionable abnormality as background.
• Inbox design and volume. 43 unread results, 38% older than 24 hours, is a system capacity problem; a pool inbox makes it worse by diffusing responsibility.
• Coverage during absence. Two days off with no documented backup or handoff of pending results is a scheduling and policy gap.

Individual-level factors, framed carefully The hospitalist signed a discharge with pending results and did not arrange coverage. That is a real behavior to address, but only after the system factors, because the same person was working in a system that made the omission easy and invisible. Similarly, the nurse and lab followed the routing rules they were given.

What to resist Concluding "human error" or "failure to check the inbox." Neither is actionable and both predict a repeat. Also resist retraining as the primary fix: the numbers show a design problem that will defeat the next well-intentioned clinician.

One thing to verify. Confirm whether the hospitalist was actually the assigned responsible clinician at 06:40, or whether the discharge order had already transferred responsibility. That determines who owned the result and whether the coverage gap is even more serious than it appears.

Interventions, matched to the failure modes

1. Pending-result reconciliation at discharge. A required step in the discharge workflow that lists unresulted tests and requires a named responsible clinician and a follow-up plan. Make it a hard stop only if the softer version fails; test first on one unit.
2. Ownership rule. Eliminate unowned pools. Every result routes to a named clinician, with automatic delegation to a covering clinician when the primary is scheduled off, and escalation if a critical result is unacknowledged within 30 minutes.
3. Threshold review. The laboratory medical director, with clinical input, reviews the potassium and INR thresholds and documents the rationale. Treat this as a clinical decision requiring validation, not an informatics preference.
4. Post-discharge result rule. A result arriving after discharge for a defined high-risk set (potassium, INR, creatinine, hemoglobin) creates a task assigned to the discharging or covering clinician plus an outreach step for values beyond a set limit.
5. Inbox coverage plan. A documented, visible coverage schedule; a daily huddle review of unopened results older than 24 hours.

Measures

• Outcome: readmissions with hyperkalemia or bleeding, serious safety events, number of results with no documented follow-up.
• Process: percentage of discharge summaries documenting pending results and follow-up owner; percentage of discharge-day results acknowledged within 24 hours; percentage of unowned results (target zero); time from critical result to documented action.
• Balancing: discharge time and discharge-before-noon rate (a reconciliation step can delay throughput and tempt shortcuts); alert volume; inbox size; clinician after-hours workload.

How it could look like success while harming care

• Results get acknowledged quickly because clinicians click through a checklist without reading, so the acknowledgment metric improves while nothing is acted on. Counter this with chart audits that check for documented action, not just acknowledgment.
• Discharge summaries are templated with a boilerplate follow-up line. Audit free text for patient-specific content.
• Discharges get delayed, beds back up, and the ED boards, so the team quietly stops using the reconciliation step. Track adoption as well as outcomes.
• Results get pushed to patients through a portal before a clinician reviews them, causing distress without a plan. Verify the release timing against the information-blocking exceptions and your portal policy.

Method. Run these as PDSA cycles with control charts and a baseline period, and report outcome, process, and balancing measures together.