Review article Statistical modelling for clinical mastitis in the dairy cow: problems and solutions

part of the between herd observed hetero-
geneity depends on the index calculation
and the characterisation of the CMAST
case in a herd [50]. In most studies, signif-
icant efforts have been made to homoge-
nise mastitis definition criteria (at least the
observation of lumps in milk), standardise
their clinical description, and confirm the
infection through milk quarter bacteriol-
ogy. Another problem is to normalise the
calculation of a synthetic CMAST inci-
dence rate, since at least 12 different defi-
nitions of CMAST incidence rate are given
by the authors [50]. In the literature,
CMAST indices are calculated as percent-
ages of infected lactations by the total
number of lactations or lactations at risk,
or occurrences (more or less censored)
observed for different periods and calcu-
lated as the numbers of days, weeks or
months at risk. According to the calcula-
tion, either the observed lactation duration
is not considered or the time interval while
an animal is not at risk after a CMAST
occurrence is a priori variably chosen.
In other respects, the calculation from
the overall number of lactations (or spe-
cific lactation periods) with at least one
CMAST is then related to the numbers of
cows-lactation at risk [6, 11, 17, 32, 39, 41,
52, 55, 79] or of CMAST-free months [14,
15]. When the calculation is made from the
total number of cases (per udder or per
quarter), it is divided by the total number
of cows-day at risk [3–5, 11, 30, 34, 48, 54,
55, 69, 79], the study period within the lac-
tation [13], or even by the number of cows-
week [9, 10, 56] or cows-month [28, 48,
54]. Sometimes, the calculation procedure
is not mentioned [45, 47]. It is important to
indicate that the calculation of the number
of CMAST cases including a number of
days at risk mainly depends on [79] the a
priori chosen value (from 0 to 90 days
according to the study), which allows to
decide if a new case occurs.