Skip to content
All lessons
Cross-cutting · Patient Safety

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.

12 min readHigh yield

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:

  1. What are we trying to accomplish? (the aim)
  2. How will we know a change is an improvement? (the measures)
  3. 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.

Donabedian triad, balancing measures + STEEEP
  • 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 typeDefinitionClassic examples
StructureResources / system attributesNurse:patient ratio, EMR, # of ICU beds
ProcessActions delivered to patientsAspirin in MI, hand-hygiene %, bundle adherence
OutcomeEnd effect on the patientMortality, readmission, CLABSI, HbA1c
BalancingUnintended consequences of a changeReadmissions rise after shortening length of stay
The PDSA cycle
  • 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
  • ActAdopt, 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
Circular Plan-Do-Check-Act (PDCA) improvement cycle diagram
The PDSA/PDCA cycle: rapid, iterative small tests of change. Healthcare QI uses 'Study' in place of 'Check.' · Wikimedia Commons — Karn-b - Karn Bulsuk (http://www.bulsuk.com). Originally published at http://www.bulsuk.com/2009/02/taking-first-step-with-pdca.html — CC BY 4.0, via Wikimedia Commons
Vignette — process vs outcome, and the next step

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 — RCA vs FMEA (backward vs forward)

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.

Ishikawa fishbone cause-and-effect diagram with branching contributing factors
A fishbone (Ishikawa) diagram organizes contributing system factors during a Root Cause Analysis. · Wikimedia Commons — FabianLange at de.wikipedia — CC BY-SA 3.0, via Wikimedia Commons

RCA vs FMEA vs PDSA

ToolTimingPurposeTypical trigger
RCARetrospective (reactive)Find the system cause of an eventAfter a sentinel/adverse event or near-miss
FMEAProspective (proactive)Anticipate & rank failure modesBefore a new high-risk process
PDSAReal-time, iterativeTest & refine a changeImproving an existing process
Variation, charts, and QI vs research
  • 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
Lock it in
  • 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.