Mastitis detection on farms with an AMS

What a little tweet can do. A couple of weeks ago we had the annual meeting of the Dutch Mastitis Research workers. Dutch is here: Dutch language, because we meet with the mastitis research workers from the Netherlands as well as from Flanders (Belgium). It had my laptop with me and an internet connection, so from each talk I made a tweet. Personally I presented research entitled: "Consequences of farmers' interpretation of mastitis alerts". This work was done as research internship of a veterinary student (Klaske Buma, she is now working as practitioner) and mostly supervised by my colleague Ruurd Jorritsma from the Faculty of Veterinary Medicine. 

The tweet of this presentation was: Results of internship of Klaske Buma: #AMS farmers check only 3.5% of mastitis alerts. As a result 74% of mastitis cases is missed 

As a result almost all Dutch farmers journals picked this up and wrote something about this item. Amazing how things go in the land of the journals. But given the interest in this little tweet, it is good to give a little more background on this item.

When milking cows with an automatic milking system (AMS), for mastitis detection, farmers have to rely on sensor alerts. It is known that, although the sensitivity and specificity of current mastitis are quite reasonable, the systems still give a relative large number of false positive alerts. In order to find detect all mastitis cases, farmers have to visually check all alerts. We do know that far from all farmers are doing that. The goal of this research was therefore to study farmers behavior related to mastitis alerts and the consequences thereof.

The study was done on 7 farms, located around one village in the north of the Netherlands. The farmers needed to be motivated to enter the study. All farms were milking with a Lely Astronaut (Lely, Rotterdam, the Netherlands) AMS. Each farm was visited five times. During each visit, the farmers were questioned about the alerts on the alert list of that day. After this questioning, the milk of all four quarters of every cow on the alert list was visually checked for clinical mastitis and was checked for subclinical mastitis using the CMT test.

During the first visit and before checking individual quarters, a questionnaire was held on the criteria farmers used for selecting cows on the alert list to be visually checked. Important criteria for farmers were, an alarming change in milk production, flakes and/or clots on the milk filter in combination with high electrical conductivity or a failed milking. Reasons for not checking alerts were: no flakes or clots on the milk filter, no change in milk production, cows that were repeatedly on the alert list or a lack of time.

In total 421 alerts were studied (see Table). Of all mastitis alerts, 60 % were not associated with mastitis. For the repeated alerts (cows that had been on the alert list earlier), 46 % of the alerts were not associated with mastitis.

Studied alets and mastitis
	Clinical mastitis	Subclinical mastitis	No mastitis	Total
Number of alerts	30	47	150	227
Number of repeated alerts	9	81	104	194
Total	39	128	254	421

Only 15 of the 421 quarter alerts were checked by the farmer (3 % of all alerts). From the checked alerts by the farmers, 67 % had clinical mastitis and 13 % had subclinical mastitis. However, the farmers missed quite a large number of mastitis cases. Of the 39 clinical mastitis cases that were found, only 10 were detected by the farmers, which means that 74 % of the clinical mastitis cases were not (yet) detected by the farmers.

In order to detect all clinical mastitis, farmers should check all alerts visually. For the 7 farmers in this study this would mean that, on average, 12 cows need to be checked each day. The question that remains is: how bad is it that so many cases of clinical mastitis are missed. Well for official EU milk quality regulations that is bad, because milk that has physical abnormalities should be withheld from delivery. On the other hand, many farmers that manually check the milk off their cows do detect flakes or little clots in the milk, squirt a few times more and deliver the milk as well, because the remaining milk is free of clots.

Is it bad for mastitis control? Well I do not know. Are the missed cases detected a few days later and does treatment at that time have similar cure probabilities? What is the probability of spontaneous cure of these "missed" mastitis cases. Interestingly, given all the fuss about this very interesting, but relatively small piece of work, we are thinking about future research.

Finally, for those interested, the topic will also be presented at the coming ECPLF conference in Leuven, Belgium.