A DISTRIBUTION-BASED SPC ALERT RULE FOR LOW-COUNT OVERDISPERSED DATA: A CASE STUDY IN HEPA FILTER MAINTENANCE

In pharmaceutical manufacturing, terminal high-efficiency particulate air (HEPA) filters are a critical barrier against airborne microbiota, yet maintenance is often performed on fixed schedules. This case study evaluates statistical process control (SPC) rules for low-count airborne Colony Forming Unit (CFU) data from Class C cleanrooms and assesses whether they can support HEPA filter maintenance decisions beyond a time-based program. Using 813 quarterly monitoring records collected over three and a half years from 79 HEPA filters, a distribution-based alert rule was developed by fitting a negative binomial model to discrete, overdispersed counts, combining discrete outlier diagnostics, a 95th-percentile control limit, and the Class C specification limit. The attached documents include the CFU data set, a Demo file data set, the Discrete SPC app and spc_core codes in R. Also a Readme file with the instructions on how to use the R code is included.