The Successful Launch and Diffusion of New Therapies - Innovation and Marketing in the Pharmaceutical Industry - page 198

when planning the launch and forecasting the sales of a new pharmaceutical drug.
The model by Camacho et al. ( 2011 ) can even be used to adjust predictions down-
wards after taking into account early switch-outs of patients from the new drug to
other drugs in the market. Their model can also be used to predict, using counterfac-
tual experiments, what would happen if a firm could reduce the number of patients
abandoning the new pharmaceutical drug shortly after its launch. In addition, one
can use the estimated parameters of a learning model for a given drug to predict the
speed at which physicians would switch patients to a new, similar drug.

7.1.5 Consideration and Choice Models

In most diffusion models, the diffusion process is viewed as a single-stage, binary-state
process in which at any point in time, individuals are either adopters or non-adopters.
A few diffusion studies consider diffusion as a multistate, macro-flow process and thus
take into account heterogeneity in customers' pre-adoption states, e.g., by incorporat-
ing awareness stages (Dodson and Muller 1978 ; Kalish 1985 ; Mahajan et al. 1984 ) or
consideration stages (Weerahandi and Dalal 1992 ). However, in these models, hetero-
geneity is not reflected at the individual adopter level but rather at the aggregate level.
To address heterogeneity among consumers in pre-adoption states, one can also build
an individual-level model that separates different stages in the adoption process. For
instance, Landsman and Givon ( 2010 ) proposed an individual-level model of a two-
stage process of the diffusion of a service. In the first stage, customers decide whether
to “consider” joining the service. This (Consideration) stage is modeled by a hazard
model. Customers who decide to consider the service move on to the Choice stage,
wherein they choose among the service alternatives and an outside No Choice option.
This stage is modeled by a conditional multinomial logit model.
The model proposed by Landsman and Givon ( 2010 ) was developed for services
or durable goods outside the pharmaceutical industry. Taking into account the
unique features of the pharmaceutical market environment (Camacho et al. 2010 ;
Stremersch and Van Dyck 2009 ), one could also apply such a model to these mar-
kets at the physician level. In this setting, in contrast to the setting of a new service,
once a new drug is introduced, physicians can prescribe either the new drug or one
of the other therapeutic alternatives already existing in the category. Accordingly,
we must distinguish between physicians' initial adoption decision (the decision to
first prescribe the drug) and their consequent process of integrating the new drug
into the choice set until the new drug reaches its ultimate share in the category.
The time-dynamic process of initial adoption can be represented using a propor-
tional hazard model, where the hazard function is decomposed into two multiplica-
tive components:

hh X
pt

=

0 ·( )

y (7.8)

pt pt

The first component, h p 0 t , defines the baseline hazard function. This function
reflects the longitudinal patterns in the duration time dynamics. The second