Information and diffusion of new prescription drugs

Abstract

This article examines the role of different product information flows on the diffusion of new pharmaceuticals. Given the innovative nature of pharmaceutical drugs and their impact on health care expenditure there is a surprisingly small literature devoted to this topic. Some information flow mechanisms have been examined individually in the literature, but very few have captured the simultaneous impact of these mechanisms on up-take and diffusion. This article uses the up-take of statins as an example. Diffusion of this therapeutical group is expressed as a function of four specific informational channels: self-experience, consumption externalities, scientific evidence and marketing. In addition to this, the influence of economic factors is tested to examine whether they have any role in drug diffusion. Prescription data from over 130 General Practitioners (GP) practices in the UK during 1991–2004 are used to test the econometric specification applying dynamic panel data methods. Results suggest individual self-experience and clinical evidence are major factors promoting diffusion, while there is an inverse relationship with GP practice size and diffusion. Having controlled for these factors financial incentives and marketing appear to play little role.

Keywords:

• technology diffusion
• information
• new prescription drugs
• system GMM estimator

JEL Classification::

• I11
• I18
• L65

Acknowledgements

We would like to than Martin Knapp, Matt Sutton and various members of the European Health Economics Association who have commented on earlier drafts of this article. Responsibility for the final product remains our proprietary right of course.

Notes

¹ Azoulay (2002) uses the stock of scientific information as proxy to study the relationship between clinical evidence in the sales pattern in the anti-ulcer drug market.

² Heart disease and cerebrovascular disease are the first two leading causes of death not only in the UK but also worldwide. In 2004 statins represented around 4% of all prescription items dispensed in England and approximately 10% of the overall net ingredient cost. The evidence regarding statins is incontrovertible and NICE has published a technology appraisal promoting the use of statins to prevent cardiovascular disease (NICE, 2006).

³ Within the statins therapeutical group there are six different molecules. The first statin to be marketed in the UK was simvastatin and it was introduced in 1989. Other statins like pravastatin and fluvastatin were introduced early in the 90s and during the second half of the 90s, atorvastatin and cerivastatin emerged in the market. In 2003 rosuvastatin was launched, this is a year before the end of the study period however its prescription is included into the analysis as part of the diffusion process.

⁴ Molecules within therapeutical class share common features and informative inter-molecular spillovers are assumed to exist: once the first molecule within the same therapeutical group is marketed, information will spill over subsequent molecules. Within-class product variation is not an issue with the drug classes studied below where individual products all exhibit the same therapeutic class effects.

⁵ In the context of the UK National Health System (NHS), in which we analyse demand, it is not evident that price will in any case enter the demand function. Under the public health insurance system of the type existing in the UK the prescribing physician has little incentive to incorporate price as a factor affecting demand. On the consumer (patient) side, the cost borne, regardless of the actual drug price, is a relatively low, fixed co-payment and thus it is unlikely that demand is affected.

⁶ Although a reduced form model is estimated endogeneity of price remains of an issue, it is addressed in our empirical specification below.

⁷ It could be argued that externalities could also arise from information sharing among patients generating consumer externalities. However, as we are examining the prescription decision what matters are physician-related externalities. Including potential consumer externalities would introduce elements of the doctor–patient relationship that are beyond the scope of this article. In any case these types of externalities are not observed by the researcher and any potential influence would be captured by the unobserved fixed-effect c_i shown in Equation 1.

⁸ Azoulay (2002) labels the articles as ‘marketing-expanding science’ to the articles that compare the drug with placebo and ‘comparative science’ when they compare two or more drugs within the same group. This distinction is not made here as we are not examining within group patterns but the influence of clinical evidence on statins diffusion as a whole. Regardless of the benchmark drug, clinical articles will be informative in nature.

⁹ Note that in characterizing marketing as an informative source we do not refer to the goals that marketing efforts may pursue. It has been discussed in the literature whether marketing has a pure information (knowledge dissemination) objective or it aims at ensuring prescription persistence over time. Evidence in support of both objectives has been empirically shown (Leffler, 1981; Hurwitz and Caves, 1988; Azoulay, 2002; Windmeijer et al., 2006). In the context of the current model, the interest is not about the discussion regarding the final goal of advertising efforts but rather what is the market mechanism used to inform physicians on availability and product characteristics.

¹⁰ Although each manufacturer will have a different marketing strategy, this variable only intends to account for the total effect of marketing on diffusion. Note that the present analysis is interested in the association between marketing and diffusion as a general trend for different therapeutical groups.

¹¹ Whether it is specifically sales or distribution employment figures will depend upon the information provided by individual manufacturers in their Annual Accounts.

¹² Studies published early after the scheme was introduced showed evidence of prescription cost containment for the first waves of fundholding practices (Bradlow and Coulter, 1993; Maxwell et al., 1993; Wilson et al., 1995). It was suggested that even though there was a general increase in prescribing costs, the growth rate was lower for the practices with fundholding status (Gosden and Torgerson, 1997; Delnoij and Brenner, 2000).

¹³ These practices were selected to be representative of the GPs distribution in the UK. The demographics (age and gender) of the patients covered by the panel of doctors in Disease-Analyzer are similar to UK population demographics when figures are compared to the census population from the Office of National Statistics (ONS).

¹⁴ Note that the available price data covers a shorter period than the prescription data, thus when obtaining estimates of demand equations with prices the study period will be restricted by the price data availability.

¹⁵ Price series and sales were deflated using the Consumer Price Index (CPI).

¹⁶ Alternative specifications with one and two lags of the dependent variable (Pres(t − 1) and Pres(t − 2)) and sales (Sales and Sales(t − 1)) were also considered to explore alternative dynamic specifications but to no great effect.

¹⁷ A large number of instruments included in the estimation may lead to a problem of weak instruments and the Sargan test of the validity of additional moment conditions may not be reliable. In the results presented, all possible instruments were included. However, we also run the model using five lags as instruments. Results did not change and the Sargan test confirmed the validity of the additional moment conditions used in the estimation procedure.

¹⁸ The fundholding and drug dispensing practice characteristics present the peculiarity that are both constant over time. The prescription data collected by IMS Disease-Analyzer recorded at the beginning of the data collection whether the practice was classified as fundholder and/or drug dispenser; however, this information was not updated in the subsequent years. Although practices might have changed status, these characteristics indicate the managerial attitude that the practice might have. In the case of fundholding, in 1999 all GP practices were required to join into Primary Care Groups (PCGs) but this change can be considered to happen in a mature stage where the efficacy of the prescription drugs was better known.

¹⁹ A further possibility was inspected to detect whether the interaction of these two effects may be strong enough to show a significant result that could support the hypothesis of organizational factors. This might indicate that in these cases the combination of having a budget could be counterbalanced by the additional set of incentives that can be derived from having extra revenue arising from selling the drug in-site. Results could not support the effect of the interaction of these two factors and thus corroborates the lack of influence of the managerial strategy that defined the activity of each practice.

²⁰ The contamination between these two variables is further confirmed by the fact that the same specification as in column 3 but excluding sales instead of marketing gives an insignificant and negative coefficient of the variable sales.

²¹ This measurement difficulty is not aided by a general reluctance of the industry to release product-specific marketing data due to commercial sensitivities.