Example—outline of the niosh 7400 QA procedure

4.3.1 Precision: Laboratory Uses a Precision of 0.45

Current guide specifications give a precision of 0.45 as acceptable in calculation of the 95% upper confidence level (UCL). Using a precision of 0.45 implies that the standard devi­ation divided by the arithmetic mean gives a value of 0.45. This ratio is variously called the coefficient of variation (CV) or the SR in the NIOSH 7400 method. With the 90% confidence interval of mean count, which includes a subjective component of 0.45 plus the Poisson component, 0.45 precision implies that reproducibility of results is questionable. Thus, the use of the 0.45 value in the calculation of the UCL must be clearly identified. For compli­ance purposes the following equation is acceptable:

Measured + (air quality) (0.45) (1.645) concentration (standard)

The QA procedures given in the NIOSH 7400 method must be referenced in a discus­sion of the 0.45 precision value used. Detailed outlines of the intralaboratory procedures are not necessary. Proof of acceptable PAT participation (i.e., judged proficient for the tar­get parameter in four successive round robins) as administered by the American Board of Industrial Hygiene (ABIH) must be provided.

4.3.2 Precision: Laboratory Uses a Precision SR that is Better Than 0.45

When a precision better than 0.45 is suggested, an outline of the NIOSH 7400 QA pro­cedures used and a current CV curve must be provided in addition to the proof of accept­able PAT participation. The outline must include the following:

• Document the laboratory's precision for each counter for replicate fiber counts by
using this procedure.

• Maintain as part of the QA program a set of reference slides to be used daily.
These slides should consist of filter preparations including a range of loading and
background dust levels from a variety of sources including both field and PAT
samples.

—Have the QA officer maintain custody of the reference slides and supply each counter with a minimum of one reference slide per workday. Change the labels on the reference slides periodically so that the counter does not become famil­iar with the samples.

—From blind repeat counts on the reference slides, estimate the laboratory intra-and intercounter SR.

—Determine separate values of relative standard deviation for each sample matrix analyzed in each of the following ranges. Maintain control charts for each of these data files.

• 15 to 20 fibers in 100 graticule fields

• >20 to 50 fibers in 100 graticule fields

• > 50 to 100 fibers in 100 graticule fields

• 100 fibers in less than 100 graticule fields

Note: Certain sample matrices (e.g., asbestos cement) have been shown to give poor precision. Prepare and count field blanks along with the field samples. Report counts on each field blank.

Note 1: The identity of blank filters should be unknown to the counter until all counts have been completed.

Note 2: If a field blank yields greater than 7 fibers per graticule fields, report possible contamination of the samples.

• Perform blind recounts by the same counter on 10% of filters counted (slides rela­
beled by a person other than the counter). Use the following test to determine
whether a pair of counts by the same counter on the same filter should be rejected
because of possible bias:

• Discard the sample if the absolute value of the difference between the square
roots of the two counts (in fiber/mm²) exceeds 2.8 (x) 2 - 8 X SR, where x = the
average of the square roots of the two fiber counts (in fiber/mm²) and SR = one
half the intracounter relative standard deviation for the appropriate count range
(in fibers).

Note 1: Fiber counting is the measurement of randomly placed fibers that may be described by a Poisson distribution; therefore, a square root transformation of the fiber count data will result in approximately normally distributed data.

Note 2: If a pair of counts is rejected by this test, recount the remaining samples in the set and test the new counts against the first counts. Discard all rejected paired counts. It is not necessary to use this statistic on blank counts.

The analyst is a critical part of this analytical procedure. Care must be taken to provide a nonstressful and comfortable environment for fiber counting. An ergonomically designed chair should be used. With the microscope eyepiece situated at a comfortable height for viewing. External lighting should be set at an intensity level similar to the illu­mination level in the microscope to reduce eye fatigue. Counters should take 10-20 min

breaks from the microscope every 1 to 2 hours to limit fatigue. During these breaks both eye and upper back/neck exercises should be performed to relieve strain.

Calculation of compliance uses the same equation with the substitution of a different value for 0.45. Use of this different value requires prior approval. All filters selected for the 10% recount should be randomly selected, rather than selecting every tenth filter.

4.3.3 Records to Be Kept in a QA/QC System

Records include:

• Sample logbook

• Chain-of-custody record

Дата добавления: 2015-11-16