Quality Improvement & Root Cause Analysis
Quality improvement is the prospective, systems-based discipline (PDSA to test changes, FMEA to anticipate failures), while root cause analysis is the retrospective, blame-free investigation triggered by a sentinel event. Boards test which tool to reach for and reliably reward the answer that redesigns the system over one that relies on vigilance or re-education.
Why systems, not scapegoats
Modern patient-safety science rests on one idea the boards test relentlessly: most errors come from faulty systems, not bad people (a "systems approach," not a "person approach"). Two complementary disciplines follow from this. Quality Improvement (QI) is prospective and continuous — you design, test, and refine processes before harm occurs (PDSA, FMEA, Six Sigma, Lean). Root Cause Analysis (RCA) is retrospective — a structured, blame-free investigation launched after a serious adverse event to find the underlying system flaws that let it happen. The exam question almost always reduces to: which tool do I reach for, and what is the next best step? — and the credited answer redesigns the system rather than relying on someone to "try harder" or "be more careful."
- RCA = retrospective: multidisciplinary, non-punitive investigation after a sentinel event; asks "why" repeatedly (Five Whys) and maps causes on an Ishikawa (fishbone) diagram. Focus on latent/system causes, not individual blame.
- FMEA = prospective: identifies how a process could fail before anyone is harmed; used when designing or changing a high-risk process (e.g., new chemo or insulin protocol).
- PDSA (Plan-Do-Study-Act): iterative, small-scale test of a change — the engine of continuous QI.
- Sentinel event = safety event that reaches the patient causing death, permanent harm, or severe temporary harm → mandates a comprehensive systematic analysis / RCA (Joint Commission; expected within ~45 business days).
- Never event = egregious, largely preventable error (wrong-site surgery, retained foreign body); CMS will not reimburse the related hospital-acquired condition.
- Near miss = error caught before reaching the patient → response is a non-punitive incident report, never punishment.
- Donabedian quality measures: Structure → Process → Outcome.
- Best answer to "how do we prevent recurrence?" = a system fix (forcing function, checklist, standardization) — not re-education or "be more vigilant" alone.
The three core tools — do not mix these up
| Tool | Timing | Core question | When it's used |
|---|---|---|---|
| RCA (Root Cause Analysis) | Retrospective (after harm) | "Why did this happen?" | After a sentinel/adverse event |
| FMEA (Failure Mode & Effects Analysis) | Prospective (before harm) | "How could this fail?" | Designing/changing a high-risk process |
| PDSA (Plan-Do-Study-Act) | Iterative test cycle | "Did this change work?" | Ongoing, small-scale QI |
Buzzwords. A physician gives a verbal order for "15 units" of insulin; it is heard and charted as "50 units." Both nursing and pharmacy "signed off." The patient becomes profoundly hypoglycemic, has a seizure, and requires ICU transfer.
- This reached the patient and caused severe harm → it is a sentinel event.
- NEXT BEST STEP: convene a multidisciplinary, blame-free Root Cause Analysis (Five Whys → fishbone). Do NOT simply discipline or fire the nurse — the distractor answer.
- Durable fixes are system-level: restrict verbal orders for high-alert medications and require a documented read-back/verification when they are unavoidable, plus forcing functions (standardized concentrations, smart-pump dose limits). Insulin is a classic high-alert medication.
Buzzwords (proactive). Before launching a new chemotherapy e-prescribing pathway, a team maps every step and scores each potential failure by severity × occurrence (frequency) × detectability (the Risk Priority Number) to fix weak points before any patient is harmed.
- NEXT BEST STEP / tool: FMEA — the prospective counterpart of RCA.
Buzzwords (interception). A pharmacist catches a 10-fold dosing error and stops it before the drug leaves the pharmacy.
- This is a near miss → NEXT BEST STEP: file a non-punitive incident report. Encouraging reporting (not punishing it) is what surfaces latent hazards and drives learning.
Safety-event vocabulary (high-yield definitions)
| Term | Definition | Classic example |
|---|---|---|
| Near miss | Error caught before reaching the patient | Pharmacist intercepts a 10× dose |
| Adverse event | Injury caused by medical care, not the disease | Anaphylaxis from a drug given despite a charted allergy |
| Sentinel event | Reaches patient → death, permanent, or severe temporary harm; mandates RCA | Wrong-site surgery |
| Never event | Egregious, largely preventable error that should never occur | Retained surgical sponge |
- PDSA — Plan, Do, Study, Act (the QI test-of-change loop).
- DMAIC — Define, Measure, Analyze, Improve, Control (Six Sigma, which reduces variation/defects; Lean reduces waste).
- Donabedian = "SPO" — Structure → Process → Outcome (three tiers of quality measures).
- "5 Whys" — keep asking why until you hit the system root cause (feeds the fishbone diagram).
- Slip vs Lapse vs Mistake (Reason) — Slip = attention/execution failure (right plan, wrong action); Lapse = memory failure (skipped a step); Mistake = wrong plan (knowledge/rule gap).
- Swiss cheese model (Reason): layered defenses each have "holes"; harm occurs only when the holes momentarily line up.
- Active vs latent errors: active = the sharp end (frontline, immediate — wrong drug pushed); latent = the blunt end (system/design/policy — look-alike packaging, understaffing, bad defaults).
- Run/control charts: common-cause variation is random/inherent — don't overreact; special-cause variation is an assignable signal that something changed — investigate it.
- Just Culture: human error → console; at-risk behavior → coach; reckless behavior → discipline (balances no-blame with accountability).
- Error-proofing hierarchy (strongest → weakest): forcing functions/constraints > automation/computerization > standardization/checklists > reminders/double-checks > education alone (weakest).
The marquee system fixes to recognize
When a stem asks for the best way to prevent recurrence, choose the option that changes the system, not the one that leans on vigilance or memory. Know the flagship interventions: the Universal Protocol prevents wrong-site/wrong-patient/wrong-procedure surgery through pre-procedure verification, surgical-site marking, and a "time-out" immediately before incision; the WHO Surgical Safety Checklist lowers perioperative morbidity and mortality; central-line bundles (hand hygiene, maximal barrier precautions, chlorhexidine skin prep, avoid the femoral site, and daily review of line necessity) cut CLABSI; and forcing functions — connectors or tubing that physically cannot be misconnected — design the error out entirely. These beat "re-educate the staff," which is almost always the trap answer.
Practice Patient Safety now
Board-style questions, spaced-repetition flashcards, and a Socratic AI tutor — free to start.