Quality Measurement & PDSA Cycles
A board-focused walkthrough of how healthcare quality is measured (Donabedian structure/process/outcome, plus the Model for Improvement's balancing measures and the STEEEP aims) and improved (the Model for Improvement and rapid PDSA cycles), drilling the process-vs-outcome and RCA-vs-FMEA distinctions the exam loves to test as next-best-step decisions.
Framework: measure it, then improve it
Quality improvement (QI) on the boards is a systems discipline: fix the process, not the person. Two frameworks dominate.
Donabedian classifies what you measure into Structure → Process → Outcome. (The balancing measure — watching for unintended harm from a change — is part of the Model for Improvement's measurement set, not Donabedian's triad.)
The Model for Improvement (IHI) drives change through rapid, small-scale PDSA (Plan–Do–Study–Act) cycles, guided by three questions:
- What are we trying to accomplish? (the aim)
- How will we know a change is an improvement? (the measures)
- What change can we make that will result in improvement?
Exam instinct: QI seeks local, iterative improvement of a system — distinct from research (generalizable knowledge, fixed hypothesis) and from blaming individuals. When a vignette introduces a new intervention, the next best step is usually a small test of change on one unit before a hospital-wide rollout.
- Structure = attributes of the care setting: nurse:patient ratio, EMR availability, ICU bed count, credentialing
- Process = what is done to/for the patient: % of MI patients given aspirin, hand-hygiene compliance, central-line bundle adherence
- Outcome = effect on the patient: mortality, readmission rate, CLABSI rate, HbA1c control, patient satisfaction
- Balancing measure (from the Model for Improvement) = watches for unintended harm from a change (e.g., shortening length of stay — does readmission rise?)
- IOM quality domains (STEEEP): Safe, Timely, Effective, Efficient, Equitable, Patient-centered
- Process measures give faster, more actionable feedback; outcome measures matter most to patients but lag and require risk adjustment
Types of quality measures
| Measure type | Definition | Classic examples |
|---|---|---|
| Structure | Resources / system attributes | Nurse:patient ratio, EMR, # of ICU beds |
| Process | Actions delivered to patients | Aspirin in MI, hand-hygiene %, bundle adherence |
| Outcome | End effect on the patient | Mortality, readmission, CLABSI, HbA1c |
| Balancing | Unintended consequences of a change | Readmissions rise after shortening length of stay |
- Plan — state the objective, predict what will happen, and plan the test (who/what/when) plus data collection
- Do — carry out the test on a small scale; document problems and observations
- Study — analyze the data and compare to your prediction
- Act — Adopt, Adapt, or Abandon; feed the learning into the next cycle
Key points
- PDSA = rapid, iterative, small tests of change; sequential cycles scale up a successful change
- Small, cheap tests limit risk and speed learning before wide implementation
- Classic image: a ramp of repeating loops climbing toward the aim
- PDSA ≈ PDCA; Deming preferred PDSA, using "Study" (not "Check") to stress learning from the data
Vignette: An ICU aims to cut central-line–associated bloodstream infections (CLABSI). The team introduces an insertion checklist (hand hygiene, chlorhexidine prep, full-barrier drape, avoid the femoral site) and tracks the percentage of insertions in which all bundle steps were followed.
- Which measure are they tracking? → a process measure (bundle adherence). The CLABSI rate itself is the outcome measure.
- Best next step to test the checklist? → run a small-scale PDSA on one unit / one team first — not an immediate system-wide mandate, and not a randomized controlled trial.
- If they also monitor whether the added steps delay emergent line placement → that is a balancing measure.
Vignette A: A patient receives a 10-fold insulin overdose; separately, a wrong-site surgery occurs. After this sentinel event, the hospital convenes a blame-free team that asks "why" repeatedly and maps the contributing system factors. → Root Cause Analysis (RCA) — retrospective/reactive, done after an adverse event or near-miss; systems focus; uses the 5 Whys and a fishbone (Ishikawa) diagram.
Vignette B: Before launching a new chemotherapy ordering system, a team maps every step and scores each potential failure by severity × likelihood × detectability (the Risk Priority Number), fixing the highest-risk steps first. → Failure Mode and Effects Analysis (FMEA) — prospective/proactive, done before implementation.
Hook: RCA looks backward at what did happen; FMEA looks forward at what could happen.
RCA vs FMEA vs PDSA
| Tool | Timing | Purpose | Typical trigger |
|---|---|---|---|
| RCA | Retrospective (reactive) | Find the system cause of an event | After a sentinel/adverse event or near-miss |
| FMEA | Prospective (proactive) | Anticipate & rank failure modes | Before a new high-risk process |
| PDSA | Real-time, iterative | Test & refine a change | Improving an existing process |
- Common-cause variation = inherent random noise in a stable process → don't overreact to a single point ("tampering" makes it worse)
- Special-cause variation = a signal: a point outside control limits, or a run/trend → investigate the specific cause
- Run chart / control chart (SPC) = plot a measure over time to separate signal from noise and see if a change truly helped
- Pareto chart = 80/20 rule → tackle the vital few causes first
- QI vs research: QI improves local care, is iterative (PDSA), and generally does not require IRB oversight; research seeks generalizable knowledge with a fixed protocol → requires IRB review / informed consent
- PDSA = Plan – Do – Study – Act (Act = the 3 A's: Adopt / Adapt / Abandon)
- STEEEP = the 6 IOM aims → Safe, Timely, Effective, Efficient, Equitable, Patient-centered
- 5 Whys = keep asking "why" to reach the root cause in an RCA
- Structure → Process → Outcome = Donabedian, in the order care actually flows
Practice Patient Safety now
Board-style questions, spaced-repetition flashcards, and a Socratic AI tutor — free to start.