SC QUALITY GUILDAlert and action levels
SCQUALITYGUILD.COM/TOOLS/ALERT-ACTION-LEVELS.HTML
Paste the history of a monitored count or measurement, such as cleanroom settle plates. Get alert and action levels four ways, then chart new results against them and see the excursions and adverse trends flagged.
ON THE EXAM · CMDA V.A, III.D.6 · FREE · NOTHING YOU TYPE LEAVES YOUR BROWSER
Q-BotClick for an explanation of this tool.For educational use. This page is a study aid. The method behind it has far more depth and detail than one page and one tool can cover, and what you see here is not everything the method offers or everything this site offers. For full understanding and application, use the governing standards and texts, and seek further education, training or qualified advice. Results depend on the data you enter, so check them before relying on them.
Levels below the limit
A specification or regulatory limit says what is unacceptable. Waiting for a result to break it is too late, so monitoring programs set two levels of their own, below the limit, from the history of the area or system:
- Alert level: a result above it is outside normal experience. It calls for attention, a check of the area and more frequent watching, but not necessarily a full investigation.
- Action level: a result above it calls for an investigation, an assessment of any product impact, and corrective action.
They are used for environmental monitoring of cleanrooms (viable counts on settle plates, contact plates and air samples, and particle counts), for water systems, compressed gases and bioburden. Because they come from your own data, they tell you when your process has changed, long before it fails the limit.
Four ways to set them
- Percentile. The 95th and 99th percentiles of the history: about one result in twenty is over the alert level in normal running. It assumes nothing about the distribution, but it needs plenty of data; the 99th percentile of 60 results rests on the top one or two.
- Mean + k standard deviations, usually k = 2 and 3. Fine for measurements that are roughly normal. Counts near zero are skewed, so the normal levels can mislead.
- Poisson. For counts of independent events, with the mean as λ: the alert level is the smallest count c with P(X ≤ c) of at least 95%. Real environmental counts are usually clumped, with a variance well above the mean, and then Poisson levels are too tight.
- Negative binomial. A count model that allows for the clumping. Fitted from the mean m and variance v by moments (r = m²/(v − m), p = m/v), it usually gives the most sensible levels for microbial counts.
Whichever is used, the levels must sit below the limit. If the statistics put the action level above the limit, the limit wins.
Excursions and adverse trends
A single result over a level is an excursion. A pattern can matter more than any single result:
- two or more consecutive alerts at the same location;
- several alerts in a short period across the area;
- more alerts than the level implies. With a 95th percentile alert level about 5% of results should exceed it; the tool gives the binomial chance of seeing as many as you did if nothing had changed;
- a change in the organisms recovered, which only the laboratory can see.
Use the levels like control limits: they separate routine variation from a signal. Unlike control limits they are usually one-sided, and they are reviewed at least yearly.
On the exam
CMDA V.A lists setting alert and action levels among the quality control and problem-solving tools you must identify, interpret and analyze. CMDA III.D.6 covers production and process controls, including monitoring and control of the environment and contamination. Expect to choose an action for a result over an alert or action level, to recognize an adverse trend, or to say why levels must sit below the limit.