Determining sample size and a passing criterion for respirator fit-test panels

Area of Science:

• Occupational Health and Safety
• Biostatistics
• Respiratory Protection

Background:

• Limited research exists on determining sample sizes and passing criteria for respirator fit-tests.
• Existing random effects models are complex and depend heavily on specific variance estimates.
• Practical methods are needed for efficient and reliable respirator fit-testing protocols.

Purpose of the Study:

• To propose a simple binomial approach for simultaneously determining sample size and passing criteria in respirator fit-tests.
• To compare the proposed binomial method with existing random effects models.
• To provide a practical and statistically sound method for respirator fit-test sample size determination.

Main Methods:

• A binomial approach was developed to search for optimal sample sizes and passing cutoffs by evaluating type I and type II errors.
• The method identifies the smallest sample size that meets pre-determined error rate limits.
• Benchmark respirator fit-test data from the National Institute for Occupational Safety and Health (NIOSH) was used for illustration.

Main Results:

• The binomial approach suggests a sample size of 35 to 40 respirators provides acceptable error rates.
• Random effects models can yield smaller sample sizes but show substantial variability based on variance component estimates.
• The binomial method demonstrated practical application and reasonable statistical properties compared to the random effects model.

Conclusions:

• The proposed binomial approach offers a practical and statistically sound method for respirator fit-test sample size determination and passing criteria.
• This method simplifies the process compared to complex random effects models, making it more implementable.
• The binomial approach provides reliable error rate control, enhancing the validity of fit-test results.