Sampling strategies for total bacterial count of unpasteurized bulk milk

• Published in: Journal of dairy science 2012-05, Vol.95 (5), p.2326-2335
• Main Authors: Pantoja, J.C.F., Rosa, G.J.M., Reinemann, D.J., Ruegg, P.L.
• Format: Article
• Language: English
• Subjects: Animal productions; Animals; bacteria; bacterial count; Bacterial Load - methods; Biological and medical sciences; bulk milk; Cattle; consultants; dairies; farms; Food industries; Food Storage; Fundamental and applied biological sciences. Psychology; milk; Milk - microbiology; Milk - standards; Milk and cheese industries. Ice creams; milk quality; monitoring; plate count; Probability; Sample Size; sampling; sampling strategy; Sampling Studies; Terrestrial animal productions; Vertebrates
• ISSN: 0022-0302; 1525-3198
• DOI: 10.3168/jds.2011-5098

The purpose of this study was to assess bulk tank milk sampling strategies for estimating total bacterial count (TBC). Nine large dairies in Wisconsin that produced and shipped at least 1 milk load per day were selected for this study. Total bacteria count was performed for each milk load produced during a 13-mo period. The milk shipment frequency was once (n=3), twice (n=4), or 3 times daily (n=2 farms). A threshold of 8,000cfu/mL was used to define increased TBC. The proportion of increased TBC (TBCref) during the study period was defined as the reference probability of an increased TBC for each farm. The number of milk loads that would need to be tested to estimate TBCref precisely (TBCref ± 0.05) in selected time periods (month, quarter, 6mo, or a year) was calculated assuming independence among TBC measurements. Sampling simulations (systematic or simple random sampling) were used to assess the validity of the independence assumption and compare different sampling schedules (every second, every third, or every seventh milk load) used for estimating TBCref in a 13-mo or 30-d TBC series. The number of milk loads tested to estimate TBCref depended on the time period of interest. For farms with daily milk shipments, at least 94% of all milk loads produced would need to be tested to estimate TBCref during a 30-d period. In contrast, when the period of interest was a year, reductions of up to 88% in the number of milk loads tested could be achieved. As the probability of an increased TBC departed from 0.50 toward 1 or 0, fewer samples were needed to estimate TBCref. A sampling schedule based on TBC performed on every second milk load resulted in 100% of 5,000 random samples (taken from the 13-mo TBC series) within the range of TBCref ± 0.05, indicating that sampling half of the milk loads would precisely estimate TBCref. Results of this study suggest that dairy consultants and processors can adjust the frequency of testing of milk loads depending on the goal of the milk quality monitoring program.