CC HY

HIGH-YIELD EXAM REVIEWER

Quality Assurance | Quality Control | Lean Six Sigma

THE BIG PICTURE — Know This First

Quality Management (broadest) → contains Quality Assurance → which uses Quality Control + Proficiency Testing + Staff as its 3 pillars.

Think of QA as a 3-legged stool: QC, Proficiency Testing, and Staff. Remove one leg and it falls.

QA vs QC — most testable distinction:

	QA	QC
Phases	ALL 3 (pre, analytic, post)	Analytic ONLY
Orientation	Process → prevents error	Product → detects error

TOTAL TESTING PROCESS (TTP) — 3 Phases

Phase	Also Called	Key Steps
Pre-Examination	Pre-analytical / Pre-test	Sample collection, patient prep, labeling, transport, storage
Examination	Analytical / Test	Actual testing, reagents, result generation, QC, calibration
Post-Examination	Post-analytical / Post-test	Result interpretation, pathologist validation, report dispatch, communication

QUALITY CONTROL — Core Concepts

Definition: Systematic monitoring of analytic processes to detect errors and prevent incorrect results from being reported.

3 Objectives of QC:

Check machine stability
Check reagent quality
Check operator/technical errors

QC Material Requirements

Liquid (ready to use) OR lyophilized (freeze-dried → reconstitute with distilled water)
Same matrix as patient specimens (serum/plasma; human preferred, bovine acceptable)
Minimum 2 control levels (normal + abnormal; or low/normal/high)
Spans the clinically important range of the analyte

Types of Controls

Type	Key Feature
Commercial Assayed	Has given ranges; more accurate; more expensive
Commercial Unassayed	No values given; you calculate mean & SD; cheaper
Specimen Pool	Human serum pooled; addresses artificial matrix concerns; less stable

LEVEY-JENNINGS CHART & WESTGARD MULTI-RULES

Levey-Jennings Chart = most commonly used QC chart; plots control values over time against mean, ±1SD, ±2SD, ±3SD

Westgard Multi-Rules — THE Most Testable Table

Rule	Trigger	Error Type	Action
1₂ₛ	1 value exceeds ±2SD	WARNING only	Do not reject; increase vigilance
1₃ₛ	1 value exceeds ±3SD	Random	REJECT
2₂ₛ	2 consecutive values exceed +2SD or −2SD	Systematic	REJECT
R₄ₛ	Range between 2 controls in same run >4SD	Random	REJECT
4₁ₛ	4 consecutive values exceed +1SD or −1SD	Systematic	REJECT
10ₓ	10 consecutive values on the same side of the mean	Systematic	REJECT

Memory trick for error type:

1₃ₛ + R₄ₛ = Random (spike rules — one sudden jump or a wild range)
2₂ₛ + 4₁ₛ + 10ₓ = Systematic (drift rules — patterns over multiple values)

RANDOM vs. SYSTEMATIC ERROR

	Random Error	Systematic Error
Affects	PRECISION	ACCURACY
Direction	Unpredictable (+/−)	Consistently ONE direction
Measured by	SD / imprecision	—
Detected by	1₃ₛ and R₄ₛ	2₂ₛ, 4₁ₛ, 10ₓ
Causes	Operator technique, instrument, reagent, environment	New reagent lot, new calibration, instrument/operator change

Systematic Error Subtypes:

Constant error — same offset at every concentration
Proportional error — offset grows with concentration

SHIFT vs. TREND

	Shift	Trend
Pattern	Abrupt jump to one side of mean	Slow, gradual movement in one direction
Causes	New reagent lot, new calibration, sudden instrument/operator change	Slow lamp/bulb decay, gradual reagent deterioration, instrument drifting from calibration

ESTABLISHING CONTROL LIMITS

Minimum 20 observations to set initial mean and SD
- Exception: CV < 1% assays (e.g., arterial blood gases) → 5 days is adequate
QC results should ideally fall within ±2SD
Changing lot numbers with similar ranges → use new mean + previous SD
More data over time → average all data for best estimates

INTERNAL vs. EXTERNAL QC

	Internal QC (Intralab)	External QC / Proficiency Testing (Interlab)
Frequency	Daily / per shift	Periodic (monthly, quarterly)
Specimen	Known controls	Unknown specimens from external provider (NRL)
Purpose	Detect errors during routine testing	Compare performance with peer laboratories
Result release	Held until QC within limits	Tested as routine patient specimens

PROFICIENCY TESTING (PT) — High-Yield Rules

Treated as routine patient specimens
Tested by bench staff on the primary analyzer
Secondary analyzers compared twice per year
No sharing of results until after deadline
Abnormalities resolved within 30 days

Failure Consequences

Offense	Consequence
Fail PT for a specific test	Lose authority for that test only
Violate any PT requirement	Lose full accreditation for all patient testing

LEAN SIX SIGMA

	Lean	Six Sigma
Goal	Eliminate waste	Eliminate variation/defects
Core question	"Does this step need to exist?"	"How can this be improved?"

DMAIC Framework

Step	Action
Define	Identify the problem
Measure	Collect data
Analyze	Find root causes using statistical tools
Improve	Re-engineer processes, optimize, standardize
Control	Ongoing tracking with daily QC and Levey-Jennings

Lean Waste Examples: waiting for specimens, searching for reagents, excess inventory, repeat testing, unnecessary paperwork, long walking distances

OUT-OF-CONTROL PROTOCOL (In Order)

Hold all patient results — never release while QC is out of control
Identify the likely error source (reagents, operator, instrument, environment, matrix effects such as hemolysis/lipemia)
Document the root cause and corrective action
Rerun QC to confirm resolution
Release patient results only after QC passes
Review results since last acceptable QC — repeat or recall if needed

RAPID-FIRE MUST-KNOWS

Question	Answer
Most commonly used QC chart?	Levey-Jennings chart
Minimum observations to set control limits?	20 (or 5 for CV < 1%)
Which rule is a WARNING, not rejection?	1₂ₛ
Which rules detect random error?	1₃ₛ and R₄ₛ
Which rules detect systematic error?	2₂ₛ, 4₁ₛ, 10ₓ
Random error affects which?	Precision
Systematic error affects which?	Accuracy
QA covers which phases?	All 3 (pre, analytic, post)
QC covers which phase?	Analytic only
PT external provider in PH?	NRL (National Reference Laboratory)
Lean's core question?	"Does this step need to exist?"
Six Sigma's core question?	"How can this be improved?"
Shift vs. Trend?	Shift = sudden/abrupt; Trend = gradual/slow drift
Unassayed vs. Assayed controls?	Unassayed has no given values; you calculate mean & SD; cheaper