Direct-to-Consumer Advertising of Pharmaceuticals: An Integrative Review - Innovation and Marketing in the Pharmaceutical Industry - page 625

DTCA effects and their managerial or societal signifi cance. As a consequence of this
emphasis, we rarely if ever fi nd the use of controlled experiments either in the lab or
in the fi eld in DTCA studies. The two most common research approaches we fi nd in
the DTCA literature are surveys and uncontrolled observational designs.
The use of surveys is widespread (Herzenstein et al. 2004 ; Iizuka and Jin 2005 ;
Murray et al. 2003 ; Robinson et al. 2004 ; Wilson and Till 2007 ). There is a signifi cant
challenge in using surveys to reliably measure the effects of DTCA because there is
typically no control group that has not been exposed to DTCA. Some survey-based
studies have tried to creatively overcome this inherent limitation. An example is Mintzes
et al. ( 2003 ) that compares patient and physician survey responses between two sites—
Sacramento, CA, where DTCA is legal, and Vancouver, Canada, where DTCA is ille-
gal, to assess DTCA effects. Despite this limitation, surveys are an important source of
information about patient and physician attitudes, as well as of DTCA effects on vari-
ables such as patient requests, which need to be measured as self-reports.
In addition to customized surveys, there is an opportunity for researchers to use
syndicated survey data. An example is the National Ambulatory Medical Care
Survey (NAMCS). In the survey, nonfederal offi ce-based physicians complete a
one-page questionnaire for each patient visit sampled during a 1-week reporting
period. The survey data include physician characteristics, patient demographics
(age, sex, race, ethnicity), and visit characteristics (patients' symptoms, complaints
or other reasons for the visit, physician's diagnoses, diagnostic and therapeutic services
ordered or provided at the visit including medications, expected sources of payment,
visit disposition, time spent with physician, etc.).
The second common approach to study DTCA effects is uncontrolled observa-
tional designs. These studies rely on comparisons of data between cross-sectional
units, or across time within units, or panel studies that use both cross-sectional and
time-series variation (Calfee et al. 2002 ; Liu and Gupta 2011 ; Narayanan et al.
2004 ; Stremersch et al. 2011 ; Wosinska 2002 ). In order to draw valid inferences
about DTCA effects one needs good observational data and appropriate statistical
analyses that adequately control for potentially confounding covariates. An example
of a panel-based study is Liu and Gupta ( 2011 ) who explain variation in number of
patient visits and number of patient requests across geographic units in the USA and
across months using DTCA expenditures in these same units as explanatory
variables. A hierarchical Bayesian negative binomial model is used to measure the
effects of DTCA expenditures while accounting for alternative explanations.
In recent years the availability of good observational data has grown, both for the
“causal” variables (advertising) and for the “effect” variables (e.g., prescription sales).
Kantar Media ( http://www.kantarmedia.com , previously known as TNS Media) traces
advertising expenditures on all branded drugs since 1995. The data are available
weekly, monthly, and yearly. Further, the data are available at the Designated Media
Area (DMA) level or the US national level. Expenditures in 11 different media, includ-
ing network TV, national newspaper, magazine, internet, and radio, are reported. On
the effects side, IMS Health ( http://www.imshealth.com ) is the major provider of
prescription sales data by brand and market. ImpactRx ( http://www.impactrx.com )
maintains a large physician panel that records prescriptions written by physicians, as
well as details of detailing visits, patient visits, diagnoses, patient requests, and so forth.