Patient Safety & Systems-Based Practice

Plan-do-study-act cycles for quality improvement

Clinical Overview and When to Suspect a PDSA-Amenable Problem

— Contrast with traditional research: PDSA is pragmatic, hypothesis-generating, and adaptive, not powered for statistical significance

— Each cycle answers three Model for Improvement questions: (1) What are we trying to accomplish? (2) How will we know a change is an improvement? (3) What change can we test?

— A process gap with measurable output (hand hygiene rates, door-to-needle time, A1c control, no-show rate, opioid prescribing variance)

— Local variation in practice without clear evidence-based standard

— A recent sentinel or near-miss event prompting workflow redesign

— Quality dashboard showing a metric below benchmark (e.g., DM eye exam rate <50%)

— Root cause analysis (RCA): retrospective, post–sentinel event, asks "why did this happen"

— Failure mode and effects analysis (FMEA): prospective risk scoring before a process is launched

— Lean/Six Sigma: waste reduction and defect rate; PDSA is the engine inside these frameworks

Plan-Do-Study-Act (PDSA) is the foundational rapid-cycle improvement method endorsed by IHI, CMS, and ACGME for testing changes in real clinical microsystems

Core premise: small, iterative tests of change on a limited scale (one patient, one shift, one clinic half-day) before broad implementation

When to suspect a PDSA-amenable problem on Step 3 stems:

Distinguish from other QI tools:

Board pearl: If the stem describes a small team testing a change on 5 patients next Tuesday and measuring results Friday, the answer is PDSA — not RCA, not FMEA, not a randomized trial

Key distinction: PDSA is improvement, not research — it generally does not require IRB approval if the intent is local quality, the change is within standard care, and findings are not generalized for publication without later review

Step 3 framing: expect questions on selecting the right QI method, sequencing cycles, choosing measures, and interpreting run charts rather than on statistical inference

Presentation Patterns and Key History — Recognizing PDSA in Vignettes

— "Only 40% of diabetic patients in your clinic received a foot exam last quarter…"

— "The ICU central-line infection rate has risen above the network benchmark…"

— "Hand-off errors increased after the new EHR rollout…"

— "A pediatric clinic wants to improve the 2-month vaccine on-time rate…"

— Small team, local scope (one unit, one clinic, one ward)

— Short time horizon (days to weeks per cycle)

— Explicit mention of piloting, trying, testing, or "starting with a few patients"

— Plan to measure, then refine before spreading

— "After a patient died from a wrong-site surgery…" → RCA

— "Before launching the new chemotherapy infusion protocol, the team wants to identify failure points…" → FMEA

— "The hospital wants to determine whether bundle A is superior to bundle B across 12 sites…" → cluster RCT, not PDSA

— Who owns the process (frontline nurses, pharmacists, residents)?

— Is there an executive sponsor and a measurement lead?

— Are patients/families embedded as partners (increasingly tested)?

Typical Step 3 vignette opens with a resident, chief, or medical director identifying a performance gap and asking "what is the best next step"

Recurring stem patterns:

Key history elements that signal PDSA as the correct answer:

Contrast cues that point AWAY from PDSA:

Stakeholder history to extract:

Step 3 management: When asked "what is the first step after identifying the gap," the answer is usually forming an interdisciplinary team and defining the aim statement (SMART: Specific, Measurable, Achievable, Relevant, Time-bound) — before designing the intervention

Board pearl: An aim statement like "Increase hand hygiene compliance on 4 West from 55% to 90% by December 31" is the canonical correct-answer phrasing — it contains a baseline, target, population, and deadline

Watch for distractors that propose hospital-wide rollout first — PDSA explicitly rejects this; you test small, then spread

Physical Exam Findings — Mapping the Process (Process Mapping & Driver Diagrams)

— Swimlane diagram: who does what, when (MA rooms patient → RN reviews meds → MD enters order → pharmacy verifies → patient receives)

— Value stream mapping: identifies waiting, rework, defects, motion — the 8 wastes of Lean (DOWNTIME mnemonic: Defects, Overproduction, Waiting, Non-utilized talent, Transportation, Inventory, Motion, Excess processing)

— Driver diagram: links the aim → primary drivers (big levers) → secondary drivers (specific factors) → change ideas (testable interventions)

— Outcome measures: did the patient benefit? (e.g., CLABSI rate, A1c <8%)

— Process measures: was the change executed? (e.g., % checklists completed)

— Balancing measures: did we cause unintended harm elsewhere? (e.g., increased ED wait time after triage redesign, alert fatigue after new BPA)

In QI, the "exam" of the system is the process map — a visual walkthrough of every step from trigger to outcome

Components to elicit (analogous to ROS + exam):

Hemodynamic-equivalent assessment = measurement plan:

Key distinction: A QI project without a balancing measure is incomplete — Step 3 frequently tests this. Example: a sepsis bundle increases antibiotic use; balancing measure = C. difficile rate or inappropriate antibiotic days

Pareto principle on exam: typically 80% of defects come from 20% of causes — Pareto charts help prioritize which driver to attack first

Fishbone (Ishikawa) diagram organizes contributors into categories: People, Process, Equipment, Environment, Materials, Methods — useful when stem asks for "best tool to identify contributing factors" in a non-sentinel context

CCS pearl: Before "ordering" an intervention, always order the process map and baseline data first — analogous to checking vitals before treating. A common wrong answer is jumping straight to implementation without baseline measurement

Engage frontline staff in mapping — process maps drawn only by leadership consistently miss real workflow workarounds

Diagnostic Workup — Baseline Data, Run Charts, and Initial Metrics

— Define the operational metric precisely: numerator, denominator, inclusion/exclusion, data source, sampling frequency

— Use existing data streams (EHR reports, registry pulls) before building new ones — sustainability favors automation

— Sample size for PDSA is small and iterative (e.g., 5–10 patients per cycle), not powered like a trial

— Plots the metric on the y-axis over time on the x-axis with a median line

— Detects non-random signals via four rules: shift (≥6 consecutive points on one side of median), trend (≥5 consecutive ascending or descending), runs (too few/many crossings), astronomical point (obvious outlier)

— A shift or trend suggests the change produced real signal, not noise

— Adds upper and lower control limits (typically ±3σ)

— Distinguishes common-cause variation (inherent to the process) from special-cause variation (a specific assignable factor)

— Key distinction: Reacting to common-cause variation as if it were special cause is "tampering" and worsens performance — a classic distractor on exams

— Is the process in control (stable) or out of control (erratic)?

— What is the current capability vs the target?

Baseline measurement is the labs and imaging of QI — you cannot demonstrate improvement without a pre-intervention denominator

Initial data collection priorities:

Run chart = the EKG of QI:

Control chart (Shewhart chart) = the advanced study:

Pre-intervention questions to answer:

Board pearl: If a stem shows a metric bouncing within control limits and a leader demands a new initiative after one bad week, the correct action is continue monitoring — that fluctuation is common-cause variation

Avoid before-after means without time-series context; a single pre/post comparison can be misleading due to regression to the mean and secular trends

Diagnostic Workup — The Four Phases of a PDSA Cycle in Detail

— State the objective of this specific cycle (often a sub-question of the larger aim)

— Make predictions ("we think hand-hygiene compliance will rise from 55% to 70%")

— Define who, what, where, when; data to be collected; tools needed

— Identify roles: process owner, data collector, learner

— Execute the plan on a small scale (one room, one shift, one provider)

— Document problems and unexpected observations as they happen — these are gold for the next cycle

— Collect the predefined data — do not change the test mid-cycle

— Compare results to predictions, not just to baseline

— Analyze with run chart, simple percentages, qualitative themes

— Ask: did the change produce the expected effect? What surprised us? What did we learn about the system?

— Crucial: a "failed" PDSA (no improvement) is still successful learning if it clarifies why

— Three options: Adopt (standardize the change), Adapt (modify and retest in next cycle), Abandon (drop this change idea)

— Plan the next cycle — PDSA is a ramp, not a single event

PLAN:

DO:

STUDY:

ACT:

Step 3 management: When a vignette asks "what should the team do next" after a small successful test, the answer is usually adapt and test on a larger scale (more patients, more providers, another unit) — not immediate system-wide rollout

Sequencing of cycles classically: 1 patient → 5 patients → 1 clinic day → 1 week across team → unit-wide → spread

Board pearl: Cycles should be short — days to a couple of weeks. Multi-month "cycles" are a red flag the team is doing a project, not iterative learning

Document each cycle on a PDSA worksheet for institutional memory and to support eventual spread

Risk Stratification — Selecting and Prioritizing Change Ideas

— High impact / low effort = "just do it" quick wins → ideal first PDSA target

— High impact / high effort = strategic projects requiring multiple linked PDSAs

— Low impact / low effort = fill-ins

— Low impact / high effort = avoid

— Use the Change Concepts library (IHI) — examples: eliminate steps, change the order of steps, use reminders, standardize, make the desired action the default, use checklists

— Leverage human factors engineering: forcing functions > standardization > checklists > education alone (education is the weakest intervention)

— Strongest: forcing functions, automation, physical/engineering changes (e.g., removing concentrated KCl from floor stock)

— Intermediate: standardized order sets, checklists, redundancy, simplification

— Weakest: education, policy memos, "double-checks" by humans, signage

Not every problem warrants PDSA; prioritize using a 2×2 impact/effort matrix:

Change concept selection:

Hierarchy of intervention strength (frequently tested):

Key distinction: When a stem offers "in-service training" vs "default order set change," the default order set wins because it leverages choice architecture and persists when people forget

Stakeholder analysis (RACI matrix): Responsible, Accountable, Consulted, Informed — ensures the right people are engaged without bottlenecks

Patient and family involvement increasingly tested — co-design improves uptake and equity

Health equity lens: stratify baseline data by race, ethnicity, language, insurance, ZIP code to ensure improvements don't widen disparities

Board pearl: If a CLABSI rate improves overall but worsens among non-English-speaking patients, the project has failed an equity balancing measure — redesign with interpreters embedded

Step 3 management: First PDSA target should typically be the highest-impact, lowest-effort, frontline-supported change with a clear measurable outcome within 2–4 weeks

Pharmacotherapy Equivalent — The "Active Ingredient" of Effective PDSAs

— Bundled, evidence-based elements (e.g., central line bundle: hand hygiene, max barrier precautions, chlorhexidine prep, optimal site selection, daily necessity review)

— All-or-none measurement: bundle compliance counted only if every element completed — drives reliability

— Standard work: explicit, written, observable steps

— Level 1 (10⁻¹ failures): intent and vigilance — education, awareness — typically the starting but insufficient state

— Level 2 (10⁻²): human factors — checklists, reminders, standardization, redundancy

— Level 3 (10⁻³): automation, forcing functions, root cause redesign

— Frequency: weekly to biweekly cycles in early phases; monthly once stable

— Titration: ramp from 1 patient → small group → unit → multi-unit as evidence accumulates

— Adverse effects: monitor balancing measures every cycle (alert fatigue, workload, equity gaps, staff burnout)

— Alert fatigue from new EHR pop-ups → use tiered alerts, retire low-value ones

— Workaround behavior when a step adds time → redesign with frontline input

— Reversion to old habits when champions leave → embed in onboarding, audit-and-feedback, default settings

In QI, the "drug" is the change package — the bundled set of interventions targeting drivers in the diagram

Characteristics of high-yield change packages:

Reliability design principles (Step 3 testable):

Dosing schedule for PDSA cycles:

Common implementation side effects and how to mitigate:

CCS pearl: Audit-and-feedback works best when individualized, timely, comes from a respected source, and includes specific actionable goals — generic monthly emails to the whole department are a classic weak intervention

Key distinction: A "bundle" is not just a list — it requires all elements every time with measurement of all-or-none compliance, which is what drives outcome improvement vs single interventions

Implementation — Spread, Scale, and Sustainability

— Frameworks: IHI Framework for Spread, Diffusion of Innovations (Rogers) — innovators → early adopters → early majority → late majority → laggards

— Identify early adopters as champions; do not start with skeptics

— Build a change package + how-to guide + measurement specs + data infrastructure

— Embed in workflow: order sets, EHR defaults, hardwired forcing functions

— Policy and credentialing: include in onboarding, competency assessments

— Ongoing measurement: dashboards visible to the team, not buried in admin reports

— Local ownership: process owner identified, with monthly review

— Context mismatch: what worked on one unit may not translate without local adaptation cycles

— Lack of executive sponsorship and resources

— No data infrastructure for the new measure

— Initial adopters not engaged in adapting the model

— Don't skip adaptation PDSAs in each new unit — copy/paste rarely works

— Avoid declaring victory on process measures only without confirming outcome improvement

Once PDSA cycles demonstrate reliable improvement in the pilot microsystem, transition to spread:

Strategies for sustainability:

Plan-Do-Study-Act → Standardize-Do-Study-Adjust (SDSA) once optimal state achieved — the change becomes the new baseline that you defend

Common reasons spread fails (high-yield):

Scaling pitfalls:

Board pearl: A successful PDSA pilot on one ward should be adapted, not directly mandated, on the next ward — local cycles re-test fit and build buy-in

Step 3 management: When stem asks what ensures sustainability after pilot success, prioritize EHR default/order-set integration + ongoing dashboard + named process owner over a one-time grand rounds or all-staff email

Recognize that culture change, not just process change, often determines long-term success — psychological safety enables continued reporting of defects and ideas

Special Populations — Resource-Limited Settings and Small Practices

— Use paper-based tally sheets when EHR analytics are unavailable

— Sample 5–10 charts weekly rather than full population — adequate for run charts

— Combine roles: one clinician may be aim-setter, doer, and measurer

— Cycles may be even shorter (one half-day clinic session)

— Cancer screening rates (colon, cervical, breast) — track via EHR registry or paper log

— Diabetes care bundles (A1c, BP, statin, ACE/ARB, retinal exam, foot exam, nephropathy screen)

— No-show reduction via reminder calls, overbooking algorithms, transportation support

— Refill workflows to reduce phone-tag and missed chronic disease medications

— Required to participate in HRSA Uniform Data System (UDS) measures — PDSA is the workhorse for moving these

— Align cycles with HEDIS measures that drive payer contracts

— MIPS/MVPs and ACO shared savings reward measurable improvement → PDSA documents the effort and provides reportable improvement activities

— Improvement Activity category in MIPS explicitly credits PDSA-style projects

PDSA scales down as well as up — it is uniquely suited to small clinics, FQHCs, rural hospitals, and resource-limited environments because it requires no large budget or statistical apparatus

Adaptations for small practices:

Common small-practice projects:

Federally Qualified Health Centers (FQHCs):

Value-based care relevance:

Step 3 management: A solo or small-group practice asking how to start improving diabetes care should identify one measure, pull a baseline, pick one change idea, test on 5 patients next week, and measure — not commission a consultant or buy new software

Board pearl: Resource constraints make small tests of change especially valuable — PDSA fails gracefully when wrong because the investment is minimal

Engage medical assistants and front-desk staff — in small practices they are often the highest-leverage change agents

Special Populations — Pediatrics, OB, Mental Health, and Transitions of Care

— High-yield PDSA targets: immunization on-time rates, well-child visit completion, lead screening, developmental screening (ASQ, M-CHAT), asthma action plans

— Family-centered design — parent input critical

— Schools and home environment must be considered in driver diagrams

— AIM bundles (Alliance for Innovation on Maternal Health) for obstetric hemorrhage, severe hypertension, VTE prevention — implemented via PDSA cycles at unit level

— Severe maternal morbidity disparities require equity-stratified measurement

— Postpartum follow-up rates, breastfeeding initiation, postpartum depression screening (EPDS) — frequent targets

— Collaborative care model rollout commonly tested via PDSA

— Measures: PHQ-9 documentation, follow-up within 14 days of psychiatric discharge, suicide risk screening (CSSRS) completion

— Confidentiality and stigma considerations in measurement design

— Discharge bundle elements: medication reconciliation, teach-back, follow-up appointment scheduled before discharge, summary sent to PCP within 48 h, post-discharge phone call within 72 h

— Targets the 30-day readmission penalty (CMS HRRP)

— Project BOOST and RED toolkits are PDSA-friendly change packages

— Falls reduction bundle, deprescribing (Beers criteria), advance care planning documentation rates

— Balancing measure: don't reduce mobility while reducing falls (over-restraint, immobility harm)

Pediatrics:

Obstetrics/perinatal:

Mental health/behavioral:

Transitions of care (very high-yield Step 3):

Geriatrics:

Step 3 management: For a stem about reducing 30-day heart failure readmissions, the best initial PDSA targets discharge teach-back + 7-day follow-up appointment + post-discharge call — bundle, not single interventions

Key distinction: Pediatric and OB QI requires family/patient partners on the team — not optional, increasingly tested

Complications and Adverse Outcomes of QI Projects

— Alert fatigue: excessive EHR pop-ups → providers click through critical alerts → patient harm

— Workarounds: staff bypass cumbersome new steps, creating shadow workflows and hidden risk

— Metric fixation (Goodhart's Law): "when a measure becomes a target, it ceases to be a good measure" — e.g., gaming door-to-doc times

— Disparity widening: improvements concentrated in English-speaking, insured patients; balancing equity measures missing

— Pneumonia core measure pushed early antibiotic administration → inappropriate antibiotics in non-pneumonia patients, increased C. difficile

— Tight glycemic control protocols in ICU → hypoglycemia and mortality (NICE-SUGAR lesson) — balancing measure required

— Sepsis 1-hour bundle → over-resuscitation in CHF patients if not individualized

— Initiative fatigue: too many simultaneous projects → none succeed; prioritize ruthlessly

— Lack of psychological safety: staff stop reporting defects → improvement stalls

— Pseudo-improvement: process measures rise without outcome change (e.g., 100% checklists but unchanged CLABSI)

— Hawthorne effect: behavior changes during observation, reverts after

— Selection bias in sampling

— Inadequate denominators

Unintended harms from poorly designed PDSAs are real and testable:

Specific examples:

Project-level failures:

Data integrity threats:

CCS pearl: When a stem describes a checklist with 100% compliance but no change in the outcome, suspect that the checklist is being completed retrospectively or perfunctorily — redesign with real-time embedded prompts or direct observation

Step 3 management: When unintended harm is detected mid-cycle, pause, study the signal, and modify the change before continuing — do not push forward simply because the project was planned

Board pearl: Always include a balancing measure that reflects patient experience, staff burden, equity, and cost — at least one of these — to prevent narrow optimization

When to Escalate — Beyond PDSA: RCA, FMEA, and System Redesign

— Sentinel event (patient death or severe harm not related to natural course of illness): mandates Root Cause Analysis (RCA) within ~45 days per Joint Commission, often paired with corrective action plan

— High-risk new process (new chemotherapy protocol, new EHR module, new procedure introduction): prospective FMEA

— Persistent failure across multiple PDSA cycles: indicates the problem is systemic/cultural → escalate to leadership for structural redesign

— Repeated cycles fail to move the metric → revisit driver diagram, may indicate wrong primary driver

— Multiple units fail to spread → cultural or resource gap; engage executive sponsor

— Staff resistance unmovable despite engagement → may need leadership, HR, or culture intervention

— Patient safety event reporting systems (Patient Safety Organizations, PSOs under PSQIA — confidential, non-discoverable)

— Hospital Quality Council, M&M conferences (peer-protected), Safety Huddles

— External: CMS, state DOH, Joint Commission, FDA MAUDE for device events

— Pharmacy for medication-error PDSAs

— Infection prevention for HAI bundles

— Risk management/legal when potential harm or claims involved

— Biostatistics when planning to publish or generalize findings

PDSA handles iterative process improvement; certain events require escalation to other methodologies:

Escalation triggers within QI work:

Reporting structures:

Consultation logic:

Step 3 management: A wrong-site surgery → immediate disclosure to patient, event report, RCA initiation, FMEA of pre-op timeout process — not a PDSA cycle as the first step

Key distinction: RCA is reactive and individual-event focused; FMEA is proactive and prospective; PDSA is iterative process improvement. Step 3 stems often offer all three — pick based on timing and trigger

Board pearl: Repeat sentinel events of the same type indicate systemic failure of prior RCA recommendations — escalate to board-level safety oversight

Key Differentials — Other QI and Improvement Methodologies (Same Category)

— Focus: eliminate waste (8 wastes), value stream mapping, 5S workplace organization, kaizen (continuous improvement)

— PDSA fits inside Lean as the test mechanism

— Best when flow, waiting times, and waste dominate (clinic throughput, OR turnover, ED LOS)

— Focus: reduce variation and defects to ≤3.4 defects per million opportunities

— Heavy statistical emphasis; longer projects (months)

— Best when precision and defect reduction matter (lab errors, medication compounding)

— PDSA mirrors the "Improve" phase

— Preoccupation with failure, reluctance to simplify, sensitivity to operations, commitment to resilience, deference to expertise

— Cultural framework, not a project method

Lean (Toyota Production System):

Six Sigma (DMAIC: Define, Measure, Analyze, Improve, Control):

Lean Six Sigma: hybrid; combines waste reduction with variation reduction

Model for Improvement (Langley/IHI): the umbrella that contains PDSA + the three questions; most commonly tested framework in US medical education

Clinical Microsystems approach (Dartmouth): focuses on the smallest replicable unit of care delivery

Learning Health System: continuous data → evidence → practice loop at organizational scale; PDSA is one engine

High Reliability Organization (HRO) principles:

Audit and feedback: a frequently tested QI intervention; effective when timely, individualized, from credible source

Key distinction: PDSA is the iterative engine; Lean and Six Sigma are broader strategies; RCA is post-event; FMEA is pre-launch. Step 3 stems test recognition of the right tool for the trigger

Board pearl: When a stem describes "reducing waiting time in clinic and eliminating non-value-added steps," think Lean with PDSA cycles as the testing tool

Choice of methodology rarely matters as much as leadership, measurement, and execution — but the exam wants the named framework that matches the scenario

Key Differentials — Research, Audit, and Accreditation Activities

— QI: goal = local improvement, uses standard-of-care interventions, iterative, low risk, typically IRB-exempt or non-research determination

— Research: goal = generalizable knowledge, hypothesis-driven, often involves protocol-driven interventions, requires IRB and informed consent

— Gray zone: when QI findings are intended for publication or use experimental interventions → submit to IRB for determination; do not assume exemption

— Compares practice to a predefined standard (e.g., guideline compliance)

— Cycle: standard → measure → compare → change → re-measure

— Similar to but narrower than PDSA; lacks the prediction/learning emphasis

— Driven by external standards; documentation-heavy

— Often catalyzes PDSA projects (e.g., medication reconciliation standards)

— MIPS, HRRP, HACRP, ACOs, bundled payments

— Use QI methods to move reported metrics

— ACGME core competency; residents must engage in QI

— PDSA project completion is a common requirement

— "Conduct a randomized trial" when the situation calls for rapid local testing → wrong answer

— "Submit IRB protocol" when project is clearly local QI with standard-of-care changes → usually wrong, unless publication/research intent

— "Wait for national guidelines" → wrong if local data already show a gap and a safe change is testable

Quality improvement vs research — the most common conflation on Step 3:

Clinical audit:

Accreditation activities (Joint Commission, NCQA, AAAHC):

Pay-for-performance / value-based payment programs:

Practice-based learning and improvement (PBLI):

Common Step 3 distractors:

Step 3 management: If a chief resident wants to test a new ward hand-off template to improve local quality without generalizing findings, the correct path is PDSA with QI/non-research determination — not full IRB submission

Key distinction: Intent + risk + generalizability determine QI vs research classification — when in doubt, consult IRB for a determination letter, which protects the team and patients

Secondary Prevention / Sustaining the Gain — Hardwiring Improvement

— Standard work documents: explicit, written, reviewed annually

— Order sets and EHR defaults: highest-leverage intervention for sustainability

— Forcing functions: cannot complete the workflow without the desired step (e.g., mandatory field in note)

— Onboarding and competency: new hires trained on the new standard

— Dashboards with thresholds triggering review

— Monthly review of process and outcome measures by the unit team

— Quarterly at department QI committee

— Annual in institutional quality report

— Pre-defined control limits that trigger investigation if breached

— Audit and feedback with individualized provider data (effective when timely + specific + from credible source + with actionable goals)

— Recognition for high performers; coaching (not punishment) for low performers

— Refresh PDSAs when metrics drift — small re-tests rather than re-launching the whole program

— Charter retained with process owner, measure definitions, escalation triggers, review frequency

— Lessons learned and prior PDSA worksheets archived for institutional memory

Once a PDSA-driven change shows reliable benefit, the focus shifts from improvement to maintenance and prevention of backsliding

Hardwiring strategies (analogous to long-term medication regimens):

Ongoing measurement cadence:

Maintenance interventions:

Long-term plan documentation:

CCS pearl: When a previously improved metric backslides, do not relaunch a new project — first review whether the original hardwiring (defaults, standards, audits) is intact. Often a forcing function was removed during an upgrade

Board pearl: EHR upgrades, leadership turnover, and merger events are predictable threats to sustained improvement — anticipate them with relaunch plans

Step 3 management: Sustainability requires named owner + automated measurement + embedded workflow + regular feedback — missing any one risks decay

Follow-Up, Monitoring Parameters, and Team Learning

— During active PDSA: data reviewed at the end of each cycle (days to weeks)

— Spread phase: weekly to biweekly review by improvement team

— Sustainability phase: monthly dashboard review, quarterly committee

— Outcome measure (the ultimate goal)

— Process measure (the change executed)

— Balancing measure (no unintended harm)

— Equity-stratified versions of each, where feasible

— Visible run charts/dashboards on the unit ("daily management boards")

— Tiered huddles: frontline → mid-management → executive, escalating issues quickly

— After-Action Reviews (AAR) at the end of each major project: what was planned, what happened, why, what next

— Storytelling sustains engagement — share patient impact in addition to numbers

— Psychological safety (Edmondson) is the prerequisite for honest reporting and learning

— Rotate roles to build skill across the team

— Celebrate learning from failure as much as success

— PDSA literacy is now a core competency — most programs require completed project for graduation

— Mentor pairing with QI faculty improves project completion

Monitoring cadence by phase:

Metrics to follow continuously:

Visual management:

Reflection and learning:

Team development:

Coaching for residents and staff:

Board pearl: The single best predictor of sustained QI success on a unit is engaged frontline leadership combined with visible, timely data — not external mandates

Step 3 management: When monitoring shows special-cause deterioration, the first action is investigate (talk to frontline, review process) — not punish individuals or relaunch the project

Key distinction: Continuous improvement requires both systems (data, structure) and culture (safety, learning) — neither alone suffices, frequently tested in vignettes that offer only a structural or only a cultural intervention

Ethical, Legal, and Patient Safety Considerations

— Federal regulations (Common Rule, 45 CFR 46) protect human subjects; QI often exempt but determination should be made by IRB

— Red flags requiring IRB review: randomization, non–standard-of-care interventions, intent to publish as generalizable knowledge, collection of identifiable data beyond clinical need

— Even exempt QI requires ethical conduct, data security (HIPAA), and minimization of patient risk

— Most PDSAs do not require individual patient consent because they test changes within standard care

— Patients should still be informed when changes materially affect their experience (e.g., a new visit-flow pilot)

— Edge case: a PDSA that introduces a non–evidence-based intervention or alters disclosure to patients does require consent and IRB review

— Encourage non-punitive reporting (Just Culture model — distinguish human error, at-risk behavior, reckless behavior)

— Reports to Patient Safety Organizations (PSOs) under PSQIA are confidential and non-discoverable in litigation

— Mandatory reporting still applies for events such as wrong-site surgery, infant abductions, certain device failures (state DOH, Joint Commission Sentinel Event database, FDA MAUDE)

— Discharge is the highest-risk transition; PDSA projects targeting med rec, follow-up scheduling, and patient teach-back directly address it

— Failure to send discharge summary to PCP within recommended window is a known driver of readmissions and a frequent legal exposure

— Errors causing harm must be disclosed promptly and honestly to patients/families; CANDOR program supports this

— QI data should never be weaponized against individuals; aggregate use only

QI vs human subjects research:

Informed consent considerations:

Patient safety event reporting:

Transition-of-care risk (high-yield Step 3):

Disclosure of harm (ethical duty):

Equity as an ethical imperative: stratify measures by demographics; failing to do so risks codifying disparities

Board pearl: A QI project that improves outcomes overall but worsens disparities is ethically deficient — redesign with equity in mind

Step 3 management: When uncertain whether a project is QI vs research, the correct action is to request an IRB determination letter before starting — protects patients, team, and institution

High-Yield Associations and Rapid-Fire Clinical Facts

Model for Improvement = three questions + PDSA cycle (Langley, Nolan, IHI)

Aim statement = SMART (Specific, Measurable, Achievable, Relevant, Time-bound)

Three measure types = outcome, process, balancing — every project needs all three

Run chart rules = shift (6), trend (5), runs, astronomical point

Control chart distinguishes common-cause (inherent) vs special-cause (assignable) variation

Tampering = reacting to common-cause variation as if special-cause; worsens performance

Goodhart's Law = when a measure becomes a target, it ceases to be a good measure

Hierarchy of interventions: forcing functions > standardization/checklists > education (education alone is weakest)

PDSA cycle scale: start with 1 patient/shift, then ramp

DOWNTIME mnemonic = Lean's 8 wastes (Defects, Overproduction, Waiting, Non-utilized talent, Transportation, Inventory, Motion, Excess processing)

DMAIC = Six Sigma cycle (Define, Measure, Analyze, Improve, Control)

RCA ≈ post–sentinel event; FMEA ≈ pre-launch risk assessment; PDSA ≈ iterative improvement

All-or-none bundles drive reliability (central line bundle, vent bundle, sepsis bundle)

HRO traits: preoccupation with failure, reluctance to simplify, sensitivity to operations, commitment to resilience, deference to expertise

Just Culture = differentiate human error, at-risk behavior, reckless behavior — not punitive for honest mistakes

Joint Commission sentinel event policy mandates RCA within ~45 days

PSQIA protects data shared with PSOs from legal discovery

CMS HRRP penalizes excess 30-day readmissions (HF, AMI, pneumonia, COPD, CABG, THA/TKA)

MIPS rewards QI participation under Improvement Activities

ACGME PBLI requires resident engagement in QI

Strongest sustainability lever: EHR default + standardized order set + named owner + ongoing dashboard

Board pearl: When in doubt on a QI methodology question, anchor to the trigger (event vs gap vs new process) and scope (local vs generalizable) — those two axes pick the right tool

Board Question Stem Patterns

— "A unit identifies that hand-hygiene compliance is 55%. Which methodology should the team use to improve this?" → PDSA/Model for Improvement

— "Following a wrong-site surgery, what is the appropriate next step?" → RCA

— "Before launching a new chemotherapy infusion workflow, the team wants to identify possible failure points." → FMEA

— Choose the option that is specific, measurable, time-bound, and has a target (e.g., "Increase DM foot exam rate from 45% to 80% in clinic A by June 30")

— After successful 5-patient test → adapt and expand to 20 patients/next clinic, not system-wide rollout

— After failed cycle → study the data, modify, retest — not abandon the entire project

— A vignette describes a sepsis bundle improving antibiotic timing (process) and mortality (outcome) but increases C. difficile → identify balancing measure as the missing piece

— 6 consecutive points above the median after intervention → special-cause/improvement signal

— Random scatter within control limits → common-cause variation, do not react

— "Which is most likely to sustain improvement?" → default order set / forcing function, not in-service education

— Local project using standard care, not for publication → QI, IRB-exempt determination

— Multi-site testing with publication intent → IRB review

— Pilot succeeded on one unit → adapt and test on second unit via new PDSAs, not direct mandate

— Overall improvement but worsening disparity → redesign with equity-stratified measurement and patient/community input

Pattern 1 — Methodology selection:

Pattern 2 — Aim statement quality:

Pattern 3 — Right next step in cycle:

Pattern 4 — Measure types:

Pattern 5 — Run chart interpretation:

Pattern 6 — Intervention hierarchy:

Pattern 7 — QI vs research:

Pattern 8 — Sustainability/spread:

Pattern 9 — Equity:

Board pearl: Two best discriminators in QI questions: (1) trigger (event, gap, new process) selects the methodology; (2) scope (local, generalizable) selects QI vs research

Step 3 management: When two answers seem plausible, choose the one that engages frontline, starts small, measures, and iterates — Step 3 rewards the PDSA mindset

One-Line Recap

PDSA cycles are small, rapid, iterative tests of change within the Model for Improvement — defined by a SMART aim, paired outcome/process/balancing measures, and a Plan-Do-Study-Act loop that adapts, adopts, or abandons based on data — used for local quality improvement rather than generalizable research.

— PDSA = iterative improvement of a process gap

— RCA = post–sentinel event analysis (reactive)

— FMEA = prospective risk identification before launch

— Lean = waste/flow; Six Sigma (DMAIC) = variation/defect reduction

— Always have outcome + process + balancing measures

— Use run charts (shift ≥6, trend ≥5) and control charts to distinguish common-cause vs special-cause variation

— Avoid tampering (reacting to common-cause variation)

— Forcing functions and EHR defaults > checklists/standardization > education alone

— Bundle elements with all-or-none compliance

— Stratify by equity to avoid widening disparities

— Local QI usually IRB-exempt; request determination when generalizing or using non-standard interventions

— Start small (1 patient, 1 shift), then ramp, then spread with adaptation cycles

— Hardwire with named process owner, embedded workflow, visible dashboards, audit-and-feedback

Methodology mapping:

Measurement essentials:

Design for reliability:

Ethics, scope, and sustainability:

Step 3 mindset: When a vignette describes a local performance gap and asks for the best next step, choose form an interdisciplinary team, write a SMART aim, collect baseline data, and run a small PDSA cycle — and remember that the strongest sustained improvement comes from changing the system, not exhorting individuals to try harder