Leveraging Social Media in the Pharmaceutical Industry - Innovation and Marketing in the Pharmaceutical Industry - page 481

Caution needs to be taken to rule out other factors that could cause physicians in the
same area to act similarly. Conducting a large scale survey to ask physicians to
nominate opinion leaders is expensive for many firms. Such an exercise may also
elicit low response rates and small sample sizes (Iyengar et al. 2011b ). Researchers
need to look for new techniques to identify the connections among doctors. A prom-
ising avenue is the use of data on shared patients that can define the connections
among physicians. Such a method would also allow us to measure the strength of
connection through the total number of patients that any two doctors share, enabling
us to generate weighted networks (Barnett et al. 2011 ; Christakis and Fowler 2011 ;
Li and Shankar 2011 ).
The limited number of studies on WOM and social influence in the pharmaceuti-
cal industry offer contradictory results. In an analysis of the diffusion of tetracycline
by doctors in four Midwestern communities in the 1950s, Coleman et al. ( 1966 )
document the existence of social contagion in physician networks. Van den Bulte
and Lilien ( 2001 ) reanalyze the same medical innovation data supplemented with
journal advertising data and show that contagion effects disappear when journal
advertising effort was controlled for, although a critical component, physician
detailing, was omitted. In contrast, Manchanda et al. ( 2008 ) document the existence
and magnitude of contagion effects even after controlling for physician level detail-
ing, sampling, as well as aggregate DTCA expenditures. The contagious peer effects
among physicians are supported by Nair et al. ( 2010 ) and Iyengar et al. ( 2011b ).
More studies are needed to extend our knowledge in this area and guide pharmaceu-
tical firms' marketing decisions.

16.3.2 Methodologies

The methodologies adopted by researchers to study WOM and social influence in
the pharmaceutical industry can be categorized into two main types: hazard models
and linear models.
WOM and social contagion studies involving new drugs focus on the social
influence on the adoption timing of a new drug. The hazard model is commonly
used for this purpose (Van den Bulte and Lilien 2001 ; Manchanda et al. 2008 ;
Iyengar et al. 2011c ) and is estimated using the maximum likelihood estimation
method. The hazard function is specified as follows:

lim

Pr [ tTt <≤+ ∆ t Tt X > , ]

i i i i

htX ( ) =

i i

∆ t

∆ t → 0

where h i is the hazard of physician i prescribing the drug at time t . X i is a vector
of covariates that includes marketing and social contagion measures, T i is time
elapsed until t , and Δ t is a small change time. Furthermore, a proportional hazard
model is typically used as follows: