
188 Bayesian Methods in Health Economics
As for the benefits, we follow Briggs et al. (2003) and consider the total
number of weeks that the patients spend in the STW state, which is computed
as
e
t
=
J
j=1
m
(t)
j1
.
Notice that because the time horizon is short (only 12 weeks), we do not
apply discounting to the results observed over time in the virtual follow up.
5.4.2 Model description
For each state s, we can arrange the overall observed number of transitions in
a vector r
(t)
s
=
r
(t)
s1
,...,r
(t)
sS
. In general, we can model these vectors using a
Multinomial distribution
r
(t)
s
| λ
(t)
s
∼ Multinomial
λ
(t)
s
,n
(t)
s
=
n
(t)
s
!
r
(t)
s1
! ···r
(t)
sS
!
λ
(t)r
(t)
s1
s1
···λ
(t)r
(t)