Limitations of on-site dairy farm regulatory debits as milk quality predictors

In the United States, compliance with grade A raw fluid milk regulatory standards is assessed via laboratory milk quality testing and by on-site inspection of producers (farms). This study evaluated the correlation between on-site survey debits being marked and somatic cell count (SCC) or standard plate count (SPC) laboratory results for 1,301 Wisconsin grade A dairy farms in 2012. Debits recorded on the survey form were tested as predictors of laboratory results utilizing ordinary least squares regression to determine if results of the current method for on-site evaluation of grade A dairy farms accurately predict SCC and SPC test results. Such a correlation may indicate that current methods of on-site inspection serve the primary intended purpose of assuring availability of high-quality milk. A model for predicting SCC was estimated using ordinary least squares regression methods. Step-wise selected regressors of grouped debit items were able to predict SCC levels with some degree of accuracy (adjusted R2 = 0.1432). Specific debit items, seasonality, and farm size were the best predictors of SCC levels. The SPC data presented an analytical challenge because over 75% of the SPC observations were at or below a 25,000 cfu/mL threshold but were recorded by testing laboratories as at the threshold value. This classic censoring problem necessitated the use of a Tobit regression approach. Even with this approach, prediction of SPC values based on on-site survey criteria was much less successful (adjusted R2 = 0.034) and provided little support for the on-site survey system as a way to inform farmers about making improvements that would improve SPC. The lower level of correlation with SPC may indicate that factors affecting SPC are more varied and differ from those affecting SCC. Further, unobserved deficiencies in postmilking handling and storage sanitation could enhance bacterial growth and increase SPC, whereas postmilking sanitation will have no effect on SCC because somatic cells do not reproduce in stored milk. Results suggest that close examination, and perhaps redefinition, of survey debits, along with making the survey coincident with SCC and SPC sampling, could make the on-site survey a better tool for ensuring availability of high-quality milk.