Risk factors for bulk milk somatic cell counts and total bacterial counts in smallholder dairy farms in the 10th region of Chile

We investigated the principal management factors that influenced bulk milk somatic cell count (BMSCC) and total bacterial count (TBC) of smallholder dairy farms in the 10th region of Chile. One hundred and fifty smallholder milk producers were selected randomly from 42 milk collection centres (MCCs). In April and May of 2002, all farms were visited and a detailed interview questionnaire on dairy-cow management related to milk quality was conducted. In addition, the BMSCC and TBC results from the previous 2 months' fortnightly tests were obtained from the MCCs. The mean BMSCC and TBC were used as the dependent variables in the analyses and were normalised by a natural-logarithm transformation (LN). All independent management variables were categorised into binary outcomes and present (=1) was compared with absent (=0). Biserial correlations were calculated between the LNBMSCC or LNTBC and the management factors of the smallholder farms. Management factors with correlations with P less than or equal to 0.20 were entered into the linear multiple regression models and backward elimination was used to exclude non-significant (P > 0.05) factors. A random MCC effect was included in the models to investigate the importance of clustering of herds within MCC. In the null model for mean LNTBC, the random effect of MCCs was highly significant. It was explained by: milk collected once a day or less compared with collection twice a day, not cleaning the bucket after milking mastitic cows versus cleaning the bucket and cooling milk in a vat of water versus not cooling milk or using ice or a bulk tank to cool milk. Other factors that increased the LNTBC were a waiting yard with a soil or gravel floor versus concrete, use of plastic buckets for milking instead of metal, not feeding California mastitis test (CMT)-positive milk to calves and cows of dual-purpose breed. The final model explained 35% of the variance. The model predicted that a herd that complied with all the management practices had a mean predicted TBC of 105 colony forming units (cfu)/ml, whereas a herd that did not comply with any of these management factors had a predicted TBC of 59 x 109 cfu/ml. The model of mean LNBMSCC explained 18% of the variance; the random effect of MCC was not significant. Management factors that decreased the mean LNBMSCC were: using the CMT for 1 year versus using the test for more than 1 year or not at all, absence of a concrete waiting yard, not filtering the milk or using filters other than a plastic sieve to filter the milk, milking cows with mastitis last, and sometimes or always examining the udder before milking. A herd that complied with all of these management factors had a BMSCC of approximately 46,166 cells/ml, whereas a herd that did not comply with any of the management practices above had a mean BMSCC of 2 x 10(6) cells/ml. (C) 2004 Elsevier B.V. All rights reserved