Sei sulla pagina 1di 16

HEALTH ECONOMICS ECONOMICS OF THE PHARMACEUTICAL INDUSTRY

Health Econ. 13: 329–344 (2004)


Published online 25 June 2003 in Wiley InterScience (www.interscience.wiley.com). DOI:10.1002/hec.827

Quality safeguards and regulation of online pharmacies


Benito Arrun* ada*
Department of Economics and Business, Universitat Pompeu Fabra, Barcelona, Spain

Summary
Using econometric evidence, this article confirms that distribution of medicines online is split into two market
segments of very diverse quality, and identifies the factors that drive quality and quality assurance in this activity.
Unlike fraudulent, ‘rogue,’ websites, which offer scant guarantees and usually sell just a few medicines without
prescription, online pharmacies offering insurance coverage and linked to conventional pharmacies typically sell a
whole range of drugs, require third-party medical prescriptions and provide abundant information to patients. It is
shown that, where online pharmacies are allowed to act legally, market forces enhance quality, as private insurers
require professional standards, and specialized third parties make a business of certifying them. Furthermore, older
online pharmacies and those running conventional operations offer higher quality, probably because of reputational
investments. Overall, this evidence supports licensing online pharmacies, especially considering that prohibiting
them is ineffective against fraudulent sites. Copyright # 2003 John Wiley & Sons, Ltd.

JEL classification: I11; I18; K23; L15; L84

Keywords e-commerce; pharmacies; quality assurance; regulation; transaction costs

section devises an analytical framework to explain


Distribution of medicines via the which factors determine the level of quality and
internet establishes several hypotheses. According to these,
service quality and information provision are
Online pharmacies challenge existing market struc- positively related to licensing and insurance
ture for the retail sale of medicines. Policy-makers coverage and to the reputation of the website
find it difficult to deal with such a new form of operator. In the penultimate section, these hy-
distribution, because, despite its benefits, its func- potheses are confirmed by a series of empirical
tioning is poorly understood, it poses new risks and tests, run on a purposely built sample of online
it may conflict with vested interests. This article pharmacies. The final section concludes with
clarifies some of these issues by showing how and some policy recommendations and sugges-
why existing online pharmacies offer different levels tions for further research. In particular, results
of quality and suggesting how the regulation could support lifting the prohibitions on online pharma-
be adapted to deal with them more effectively. cies and using public regulators, health insurers
The rest of the article proceeds as follows. The and third-party certifiers to drive this activity to
remainder of this section describes how online the desired level of quality. This might even,
pharmacies work, paying particular attention to indirectly, curb the current underground trade by
their regulation, the diverse quality of their fraudulent online operators, known as ‘rogue
services and the risks they involve. The next sites.’

*Correspondence to: Department of Economics and Business; Universitat Pompeu Fabra; Trias Fargas, 25, Barcelona 08005,
Spain. E-mail: benito.arrunada@upf.edu

Received 30 September 2002


Copyright # 2003 John Wiley & Sons, Ltd. Accepted 26 March 2003
330 B. Arrunada

Different regulations and weak enforcement The simultaneous presence of licensing and
ineffective enforcement leads online distribu-
In the USA, online pharmacies have been allowed
tion of medicines to split into two market
to operate on essentially the same basis as ordinary segments, with very diverse levels of quality: one
‘brick and mortar’ pharmacies [1]. This freedom,
segment of high-quality online pharmacies and
together with greater development of the Internet
another of rogue sites. Their standard processes
and more fragmented and market-driven health can be described by reference to their extreme
care and insurance, has allowed online distribution
types.
of medicines to become relatively more developed
High-quality online pharmacies are licensed and
there.
comply with regulations like any other pharmacy,
In contrast, only three European countries
from which they differ only because they serve
(Denmark, Netherlands and the UK) allow
their customers online. Typically, each user first
distance selling of medicines. According to the e-
opens an account, so must send in the details of his
commerce directive,a the pioneer online pharma-
credit card and medical insurance. After registra-
cies established in these countries are supposedly
tion, each patient is assigned to a specific team of
allowed to sell to buyers from other countries of
pharmacists; then, either the patient or his doctor
the EU. However, national authorities are increas- sends in a valid prescription by conventional mail,
ingly constraining such practice. A prominent
fax or, increasingly, by certified e-mail. The
example can be seen in the Doc.Morris case in
pharmacy then sends the medicines by mail from
Germany, still under litigation.b Paradoxically, a central warehouse or from a conventional
the European Union strives to facilitate e-com-
pharmacy close to the patient’s home from which
merce by all means [7, 8] except, in the case of
the patient may also collect them personally.
e-pharmacy, for lifting the regulatory constraints
Medicines usually take 2 or 3 days to arrive
that make it impossible in most of its member
for which the patient does not incur additional
countries.
mail costs, or can be delivered in 24 hours at a
Given the difficulties for effective enforcement,
special cost for urgent delivery. Most of these
prohibition of online pharmacies has not stopped
pharmacies also have permanent channels of
illegal trading, however. Enforcement is seriously
communication so that customers can consult
hindered in e-commerce because it breaks the
their pharmacists either by a free phone service
territorial correspondence between enforcers and or by e-mail [9, 10].
the regulated. Providers are difficult to control
Low-quality sites, on the contrary, substitute
when their presence in the territory is confined to
independent medical prescriptions by online
the computer screen of users. The end result is that
questionnaires of doubtful effectiveness. Some
prohibition hinders the existence of reliable online
operators even advertise the fact that no pre-
operations and not only protects conventional
scription is required. (‘No prescription? No
pharmacies but may also indirectly benefit rogue
doctor? No problem’). Those that use question-
sites, as an increasing number of customers use
naires ask patients for basic personal details,
them to buy legal medicines that could be provided
such as their age and whether they are following
by reliable online sellers, if only they were allowed
any treatment or have any allergies, or whether
to operate. they have a family or personal background of
certain illnesses, such as heart disease and
cancer. They then provide a text in which they
explain that, in order to make the purchase,
Different standards of practice the customer must be over 18, they recom-
mend patients to consult their doctors about
Two factors, therefore, determine the odd struc- the use of the drugs and, finally, they ask
ture of this market with respect to quality of the patients to accept liability for the con-
service. Firstly, some countries have allowed sequences of the purchase and use of the
licensed online pharmacies to be created. Sec- medicines. Most of these sites state that the
ondly, enforcement of prohibitions to sell medi- questionnaires are checked by a doctor who
cines online from unlicensed sites is weak in all decides on whether or not the medicine is
countries, due to the novelty and difficulties of appropriate and whether the order should go
controlling e-commerce. through or not.

Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
Quality Safeguards and Regulation of Online Pharmacies 331

Advantages 30 per cent over prices in the real pharmacies


belonging to the same companies. Moreover,
The purchase of medicines via the Internet, as with many pharmacies provide additional means of
other types of electronic commerce, can be very comparing prices, offering very substantial savings
beneficial for consumers in terms of service and, for consumers.e Such figures refer to direct
less clearly, prices. discounts for users. In the case of customers
Online pharmacies provide added convenience covered by health insurance schemes, the discount
and new services. Users enjoy greater availability takes the form of a lower co-payment. In addition,
in terms of opening hours, are able to substitute the insurer usually receives an additional, larger
user travel by mail and can carry out comparative discount. Although, in principle, the user does not
shopping, with a wider variety of products and benefit from such discounts to insurers, in the long
enhanced access to product information and term the latter take them into account when setting
privacy [9, 11]. Online pharmacies offer completely the premiums of their policies so that users
new services such as sending electronic messages to purchasing their medicines on-line should end up
remind patients to take their medicines at the right paying less for their policies.
times or in the correct dose.c They also maintain
exhaustive records on their patients, with details of
the medicines they consume, their complaints and The presence and extent of risk
any contraindications, and this allows them to
provide a better-quality service and to collaborate From a public interest viewpoint, the sale of
efficiently in tests on new drugs and in clinical medicines via the Internet represents a threat in
trials. Benefits also extend to other parties, who that it easily escapes the conventional system of
may attain substantial savings in administrative regulation, with dual control by both doctor and
costs [12]. A report to the European Commission pharmacist, whatever its real merits. In fact, in
thus considered them cost-effective [13]. Certainly, many online sales, either the doctor’s prescription
the advantages of Internet distribution vary or the involvement of a pharmacist, or both, are
according to the type of product on offer. Being eliminated. This happens when the patient pur-
small, lightweight and valuable, most drugs are chases the medicines directly, without a prior
ideal for all types of distance selling, especially consultation, with no clinical action by the online
electronic commerce. But e-commerce is most seller or with any action being based on personal
useful for users when their complaints are easy to questionnaires that are difficult to verify and are of
identify and when talking about them to the limited efficacy. Thus, the doctor cannot identify
doctor or requesting them personally in the contra-indications, the patient receives no infor-
pharmacy may be a source of embarrassment. mation and the pharmacist cannot detect harmful
On the other hand, drugs which are well-known by interactions or educate patients [15].
patients and which are not covered by health The existence of doubtful suppliers has been
insurance are easy to buy on-line. Obviously, sales pointed out in many empirical studies. In one
in online pharmacies are mostly for non-urgent study carried out in 1999 on 77 online suppliers of
drugs,d especially when patients can estimate unprescribed Viagra, the price was seen to vary
demand in advance as with chronic illnesses. Such between 5 and 50 dollars, the average price being
factors explain the success of on-line sales of 12.60 dollars. Some suppliers charged an average
Viagra and ‘lifestyle drugs’ which assist in of 64 dollars for carrying out a health question-
problems of obesity, balding and the like. naire. 40 per cent, most of them outside the United
A recent survey of scientific studies concludes States, did not request any medical information
that online markets are more efficient with respect from the purchaser and those that did requested it
to price levels, price updating and price elasticity. in a way that was difficult to understand [15]. Of
They suffer greater price dispersion, however, the 22 pharmacies identified in a more specific
possibly explained by heterogeneity in levels of study that aimed to estimate to what extent Viagra
trust and buyers’ knowledge, and are still in a was being sold via the Internet for patients with
developing stage [14]. Similar results have been contra-indications [16], 2 required a written pre-
observed for online pharmacies, which seem to scription from a doctor, 9 dispensed Viagra
offer lower prices than their conventional counter- without a prescription at all and 11 issued an
parts. Trade journals report discounts from 10 to ‘online prescription’ after a so-called ‘doctor’ had

Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
332 B. Arrunada

examined a clinical questionnaire. The simulated tion and therefore increases its intensity. For
purchaser was a woman aged 69, with difficulty in example, it may be less costly for many patients
experiencing orgasm, obesity, coronary insuffi- to ask for advice by e-mail than in person and it
ciency and high blood pressure and who was may be more convenient for the pharmacist to use
taking captopril, pravachol, athenolol and ery- the same method because it enables him or her to
thromycin. She requested 66 pills from 10 of the 11 schedule his time better. This, together with a
pharmacies requiring a questionnaire, 8 of which wider use of computer databases, may increase the
were based in the United States. Three of these amount and quality of professional advice instead
companies sold the pills, in spite of the serious of reducing it.
contraindications, and seven rejected the order.
The operation of these rogue sites, seemingly
selling drugs of dubious quality without effective
control, has subjected the sector to heavy criticism.
Experienced regulators of online pharmacies have Analytical framework
a more nuanced view, however. First, it is clear
that the main problems of e-commerce medicine Quality assurance and testable hypotheses
sales come from the poor-quality service provided
by fraudulent online operators, who use the new The economics of quality assurance is based on
sales channel to carry out illegal transactions, distinguishing a set of specific problems and
selling unapproved, contaminated, false or expired safeguards. Problems are defined in terms of the
drugs [9]. Furthermore, not only are such practices type of information asymmetry extant between the
far from being general, but their health effects are parties and their unique or repetitive interaction
unclear. Even the FDA accepts that there are very [17, pp. 106–26]. Safeguards are mainly based on
few known cases of real damage being caused by the threat of losing quasi-rents linked to previous
the sale of medicines on line [9]. It seems that such investments in real or reputational assets [18, 19].
rogue sites do not sell narcotics, stimulants, Most medicines present a mix of ‘experience’
depressants and anabolic steroids that are con- attributes [20], whose quality is only known after
trolled substances regulated in the US by the Drug consumption (pain relief, delivery time), and
Enforcement Administration.f To make things ‘credence’ attributes [21], for which quality is
even more ambiguous, allowing users to get round never known (hidden effects). This, together with
regulatory obstacles is not always inefficient. This the need of expertise for the evaluation of
may well be the case when regulators get into the attributes, generates substantial scope for infor-
habit of delaying authorization of new medicines mation asymmetry. This asymmetry is the main
as a silent way of saving pharmacy costs. reason behind normative arguments for licensing
What seems clear, in any case, is that the risks and other conventional regulation. Licensing not
associated with rogue sites are not present when only controls quality directly but it also defers
dealing with high-quality online pharmacies. These compensation, by creating a stream of quasi-rents
may even provide better quality and control than that helps prevent the provision of low-quality
conventional European pharmacies. First, because services to uninformed users.
online pharmacies enjoy more powerful incentives In the case of online pharmacies, identical
to comply due to their larger scale, the role of mechanisms are in place for pharmacies which
reputation (when compared to the typical Eur- are licensed by a public agency or whose sales are
opean pharmacy owned by an individual) may covered by insurers. Both of these decisions,
take on greater importance, and policing their licensing and providing insurance coverage, are
behaviour is easier. Second, their technology also made by third parties with an interest in avoiding
allows them to provide a service of high quality. low quality. Both also create quasi-rents to the
The main alleged defect of online pharmacies in extent that losing a license or the coverage of
this respect is the lack of direct contact between insurers would cause substantial losses.
patient and pharmacist. Personal contact certainly In addition, two other drivers of quality may be
diminishes, but this does not necessarily imply a present as a consequence of pure market forces.
reduction in clinical activity. Personal consultation First, some online pharmacies put great amounts
is always restricted for reasons of cost whereas the of reputational capital in play, the value of which
online pharmacy reduces the cost of communica- would be destroyed if they were caught cheating.

Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
Quality Safeguards and Regulation of Online Pharmacies 333

This is clearer for those which also run conven- prescription (FrgnSales); if it sells or not other
tional operations and cover their online operations products in addition to medicines (OtherPrd); the
under the same reputational ‘umbrella’. Second, number of different products it sells (Products);
older online pharmacies can be expected to and if it sells or not generic drugs (Generics). In
provide greater quality because of reputational addition, the cost of express delivery service
and learning effects. On the one hand, given the (UrgntFee) is also considered. All these variables
presence of experience attributes in the purchase of (with the exception of the urgency fee, because of
medicines, reputation will tend to accumulate with the lower number of observations available) were
each instance that an online pharmacy delivers on used to build a global index of service quality
its promises so, the longer a pharmacy is in (QualIndex) using principal components analysis,
business, the better known and more valuable its as reported in Table 2.
trade name will be.g On the other hand, substantial Second, the information supplied to users is
learning economies can be expected in a new proxied through five dummy variables. Three of
activity like this, and these economies will also be them represent the different kinds of information
greater in older operators. channels, consisting of a phone line or e-mail
In summary, the following four hypotheses will address to deal with problems related to purchases
be tested in this article: and deliveries (PhoneLine and E-mail variables
respectively), or a service that allows buyers to
1. Pharmacy type: Two types of online pharmacies consult about the use of medicines (AskPharm),
deliver widely different quality. plus the total number of information channels
2. Insurance coverage: Quality is higher for phar- available (InfoIndex). Two other dummies con-
macies with insurance coverage. sider the provision of pricing information,
3. Reputation extension: Online quality is posi- either through lists of prices (PriceLis) and/or
tively related to reputational capital. price comparisons with competitive pharmacies
4. Age effect: Quality is positively related to the (PriceCom). An index of information quality
age of the pharmacy. (Info-Index) has also been built using principal
components analysis (weights are given in Table 3).
Third, external certification of quality is mea-
Data and methodology sured considering the availability of each different
accreditation via specific dummies (VIPPS, HON
A specific sample was built to test these hypoth- and VeriSign) that take value one when the
eses. It was drawn up by compiling the most pharmacy holds the corresponding certification,
outstanding characteristics of all online pharma- zero otherwise. The total number of certifications
cies listed by the Yahoo search engine.h The held by each site (Certifcs) is also measured,
information was obtained by direct visits and adding five other certificates that are less common
examination of their web sites during the months to these three.j
of February and March 2001. Visits were repeated The first hypothesis – on pharmacy type – will
for consistency in June 2002 and changes observed be tested using the variables in the previous sets of
were not statistically significant. attributes, while the three quality factors that
Business practices of online pharmacies were define hypotheses two to four will be proxied as
measured based on 17 attributes, each represented follows. In the second hypothesis, licensing and
by a dependent variable, and three additional insurance will be measured by third-party insur-
indices built from them. Table 1 defines the ance coverage, InsurCov, a dummy that takes the
variables considered and summarizes the descrip- value of one if insurers cover purchases of
tive statistics. Attributes considered are of three medicines in this pharmacy, and zero otherwise.
kinds: service quality, information and third-party This assumes that third-party insurers would not
certification. be willing to pay for purchases in unlicensed
First, the quality of services provided takes into pharmacies. This is somehow stricter than mere
account the following: if the pharmacy requires an licensing, but gets rid of potential difficulties
independent prescription (PrescRqd), without re- caused by the use of variable standards of rigor
lying fully on an internal questionnaire or not in licensing decisions, mainly where these decisions
requiring any prescription at all;i if it is willing or are made by regional or state authorities. The
not to sell in foreign markets products requiring a assumption is also supported by the evolution of

Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
334 B. Arrunada

Table 1. Definition and descriptive statistics of all variables


Variable Explanationa Observa- Mean Standard Maximum Minimum
names tions deviation
Independent variables
InsurCov Insurers cover purchases of 113 0.195 0.398 1 0
medicines in this pharmacy
RealAssets Also runs conventional 113 0.133 0.341 1 0
pharmacies
PharmAge Years since the pharmacy is 113 1.142 1.231 7 0
openedb

Dependent variables
Quality of service
QualIndex Scores of principal 104 0.00 1.846 1:315 4.014
components analysis
of standardized PrescRqd,
FrgnSales, OtherPrd,
Products and Genericsb
PrescRqd Requires the buyer to 113 0.301 0.461 1 0
present a prescription,
without relying fully on an
internal questionnaire
FrgnSales Sells in foreign markets 104 0.702 0.460 1 0
products requiring
prescription
OtherPrd Sells other products in 113 0.142 0.350 1 0
addition to medicines
Products Number of products soldb 113 34.699 45.030 100 1
Generics Sells generic medicines 113 0.142 0.350 1 0
UrgntFee Additional cost of express 65 17.246 4.131 26 7
or overnight delivery
service (in US dollars)b

Information to users
InfoIndex Scores of principal compo- 113 0.000 1.258 3:123 3.549
nents analysis of
standardized PhoneLine,
E-mail, AskPharm,
PriceLis and PriceComb
PhoneLine Offers a telephone line to 113 0.522 0.502 1 0
help in problems related to
purchases and deliveries
E-mail Offers an E-mail address for 113 0.991 0.094 1 0
buyers to contact about
purchases and deliveries
AskPharm Offers a service for buyers to 113 0.345 0.478 1 0
consult about the use of
medicines, doses, etc.
PriceLis Presents lists of its prices 113 0.956 0.207 1 0
PriceCom Offers price comparisons 113 0.062 0.242 1 0
with competitive
pharmacies

Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
Quality Safeguards and Regulation of Online Pharmacies 335

Table 1 (continued)
Variable Explanationa Observa- Mean Standard Maximum Minimum
names tions deviation
Third party quality certification
Certifcs Sum of certifications 113 0.283 0.661 3 0
(including VIPPS, HON,
VeriSign and others; could
take integer values zero to
eight)b,c
VIPPS Holds the VIPPS quality 113 0.088 0.285 1 0
certification, created by the
National Association of
Boards of Pharmacy
HON Follows the code of conduct 113 0.053 0.225 1 0
of the foundation Health on
the Net
VeriSign Site registered as a VeriSign 113 0.080 0.272 1 0
Secure Site Program
a
All variables are dummies that take values one when true, zero otherwise, unless those marked.
b
Not a dummy variable.
c
In addition to VIPPS, HON and VeriSign, five other certifications are represented in the sample: NCPA, Hi-Ethics, TRUSTe,
BBBOnLine and ePublicEye.

online pharmacies in the US, in whose success paying special attention to the existence of one-
insurers played an essential role [26,12]. In the way causations by dummy variables (structural
third hypothesis, which tests whether service zeroes). Thirdly, a system of structural equations
quality increases with reputation, reputation will will be estimated to check that the results are not
be proxied by RealAssets, another binary variable substantially affected by potential endogeneity of
that takes value one when the firm also runs a any of the independent variables.
conventional pharmacy operation, and zero if it
operates only through the Internet. This variable,
therefore, captures reputational spillovers from or
to ‘brick and mortar’ activities.k In addition, the Results
existence of such spillovers will also be examined
by checking the prediction that pure online The diversity of online pharmacies
pharmacies rely more on third-party certification.
Finally, the joint effect of accumulated reputa- The descriptive statistics presented in Table 1
tional investments and learning economies, posited already show the diversity of business practices
in the fourth hypothesis, will be measured by the followed in the sale of medicines through the
time the seller has been operating online, measured Internet. Insurers cover purchases in 19.47% of all
in years (PharmAge). the online pharmacies in the sample. A minority
The empirical analysis will be based on several (13.27%) also run conventional operations. On
methods. Firstly, a cluster analysis will be used to average, they had been open for slightly more than
identify well-separated groups in the sample. These 1 year (13.7 months), with a maximum of 7 years.
groups will then be tested by comparing their A doctor’s prescription is required by 30.09% of
means in order to confirm the typological hypoth- them. Only a few (14.16%) sell other products in
esis. Secondly, quality attributes of online phar- addition to drugs but most of them (70.19%) are
macies will be regressed on three independent happy to sell in foreign markets. Around a third
variables (InsurCov, RealAssets and PharmAge) (31.86%) carry a full range of drugs,l in contrast to
which are provisionally considered as exogenous, another third (33.63%) who carry only one

Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
336 B. Arrunada

Table 2. Principal components analysis used to build the service quality index
Component Eigenvalue Difference Proportion Cumulative
1 3.40920 2.66557 0.6818 0.6818
2 0.74364 0.32318 0.1487 0.8306
3 0.42046 0.07426 0.0841 0.9147
4 0.34620 0.26570 0.0692 0.9839
5 0.08050 0.0161 1.0000

Eigenvectors
Variable 1 2 3 4 5
PrescRqd 0.51322 0.06843 0.07530 0.37672 0.76442
FrgnSale 0.43557 0.31256 0.71138 0.43878 0.11829
OtherPrd 0.40149 0.56575 0.66609 0.27334 0.01859
Products 0.49684 0.09502 0.00484 0.58785 0.63130
Generics 0.37262 0.75400 0.21111 0.49525 0.05280

Scoring Coefficients
Variable 1
PrescRqd 0.51322
FrgnSale 0.43557
OtherPrd 0.40149
Products 0.49684
Generics 0.37262
Note: Score based on unrotated principal components, 4 scorings not used.

product, with the remaining sites carrying between Determinants of quality


2 and 29 drugs with an average of 7.26. It is
possible to buy generic drugs in 14.16%. The The regression models also confirm the hypotheses
average fee for express delivery was $17.25, with a about the impact of insurance coverage and
range between $7 and $26. More than half the reputation on quality. In particular, results pre-
online pharmacies in the sample (52.21%) main- sented in Tables 5 and 6 show that the three
tain telephone assistance lines and 34.51% of them independent variables explain a substantial share of
provide access to a pharmacist. Price lists were the variability observed in the quality of service, the
provided by almost all sites (95.58%), while availability of information channels and the supply
only 6.19% offered price comparisons. VIPPS of third-party quality certification. Most strongly,
certification was held by 8.85%, with lower all three-dependent variables are positively related
numbers for HON (5.31%) and VeriSign (7.83%). to insurance coverage. Without inferring the direc-
To explore the data in a more systematic tion of causality (more on this below), those
way, a cluster analysis was performed using pharmacies with insurance coverage are much more
hierarchical agglomerative average linkage clus- likely to require a prescription, not to sell in foreign
tering. The resulting dendrogram (Figure 1) markets, sell other products in addition to drugs, sell
shows that pharmacies in the sample gather a wider variety of drugs and charge less for urgent
neatly into two groups, which are well sepa- delivery. They also tend to provide more informa-
rated by a large dissimilarity gap. Table 4 tion. In particular, they tend to offer information on
shows the descriptive statistics of these two the use of drugs (‘Ask the pharmacist’). Moreover,
groups. The fact that the high-quality group they are more inclined to have their quality
presents significantly ‘better’ means in all relevant independently certified by a third party.
dimensions strongly supports the typology Insurance coverage shows such an unusually
hypothesis. high explanatory power that, as detailed in Table 5,

Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
Quality Safeguards and Regulation of Online Pharmacies 337

Table 3. Principal components analysis used to build the information index


Component Eigenvalue Difference Proportion Cumulative
1 1.58134 0.53393 0.3163 0.3163
2 1.04741 0.08005 0.2095 0.5258
3 0.96736 0.13214 0.1935 0.7192
4 0.83522 0.26654 0.1670 0.8863
5 0.56867 0.1137 1.0000

Eigenvectors
Variable 1 2 3 4 5
PhoneLine 0.62425 0.07059 0.05812 0.35167 0.69157
E-mail 0.18742 0.49982 0.81065 0.23052 0.06907
AskPharm 0.59815 0.20443 0.22856 0.32514 0.66518
PriceLis 0.45076 0.16375 0.24090 0.84340 0.02553
PriceCom 0.11924 0.82255 0.47875 0.07846 0.27172

Scoring Coefficients
Variable 1
PhoneLine 0.62425
E-mail 0.18742
AskPharm 0.59815
PriceLis 0.45076
PriceCom 0.11924
Note: Score based on unrotated principal components, 4 scorings not used.

95.6

71.7
Dissimilarity measure

47.8

23.9

0
G1 G2 G3 G4 G5 G6 G7 G8 G9 G10 G11 G12 G13 G14 G15 G16 G17 G18 G19 G20

Fig. 1. Cluster dendrogram. Note: View limited to the top 20 branches

Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
338 B. Arrunada

Table 4. Characteristics and comparison of means of the two main clusters of online pharmacies
Group of low quality pharmacies Group of high quality pharmacies Mean comparison
Variable Number Mean Std. Dev. Number Mean Std. Dev. t test P > jtj
InsurCov 77 0 0 36 0.6111111 0.4944132 7.4152 0.0000
RealAssets 77 0 0 36 0.4166667 0.5 5.000 0.0000
PharmAge 77 0.8571429 0.8226127 36 1.75 1.679711 3.0243 0.0042
QualIndex 70 1.171907 0.3741161 34 2.412749 1.198849 17.0367 0.0000
PrescRqd 77 0.012987 0.1139606 36 0.9166667 0.280306 18.6367 0.0000
FrgnSales 70 0.9 0.3021661 34 0.2941176 0.4624973 6.9519 0.0000
OtherPrd 77 0 0 36 0.4444444 0.5039526 5.2915 0.0000
Products 77 4.168831 4.894687 36 100 0 171.8016 0.0000
Generics 77 0 0 36 0.4444444 0.5039526 5.2915 0.0000
UrgntFee 44 18.70455 2.833188 21 14.19048 4.781412 4.0039 0.0004
InfoIndex 77 0.5123301 0.8651883 36 1.095817 1.273691 6.8706 0.0000
PhoneLine 77 0.3766234 0.4877165 36 0.8333333 0.3779645 5.4365 0.0000
Email 77 0.987013 0.1139606 36 1 0 1.0000 0.3205
AskPharm 77 0.1428571 0.3522217 36 0.7777778 0.421637 7.8454 0.0000
PriceCom 77 0.0649351 0.2480271 36 0.0555556 0.2323107 0.1957 0.8454
PriceLis 77 0.987013 0.1139606 36 0.8888889 0.3187276 1.7943 0.0804
Certifcs 77 0.025974 0.1601008 36 0.8333333 0.9411239 5.1127 0.0000
VIPPS 77 0 0 36 0.2777778 0.4542568 3.6690 0.0008
HON 77 0 0 36 0.1666667 0.3779645 2.6458 0.0121
VeriSign 77 0.012987 0.1139606 36 0.2222222 0.421637 2.9279 0.0058
Note: Means are compared assuming unequal variances and using Welch formula [27].

Table 5. Dependent variables which are fully determined by some independent variables (‘structural zeroes’)
Dependent variables Independent dummy variables
Variable name Observations Values InsurCov ¼ RealAssets ¼
¼0 ¼1 ¼0 ¼1
PrescReq 113 0 79 0 79 0
1 12 22a 19 15a
PhoneLine 113 0 54 0 52 2
1 37 22a 46 13
E-mail 113 0 1 0 1 0
1 90 22a 97 15a
AskPharm 113 0 74 0 72 2
1 17 22a 26 13
VIPPS 113 0 91a 12 90 13
1 0 10 8 2
HON 113 0 91a 16 93 14
1 0 6 5 1
Observations 113 91 22 98 15
a
According to the sample, the value of the independent variable fully determines the dependent variable.

one of the values of some of the dependent macies in which purchases can be charged to
variables (PrescReq, PhoneLine, E-mail, Ask- insurers require an independent prescription,
pharm, VIPPS, HON) is fully determined by the provide an ‘Ask the Pharmacist’ information
InsurCov variable. In particular, all online phar- service and receive queries via e-mail. Similarly,

Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
Table 6. Determinants of quality attributes of online pharmaciesa
Dependent variables

Service quality Information channels Third party certification


and price information provided

Service Quality dimensions included Express Information Main specific Price information Total certificates Main specific
Quality in online pharmacies’ delivery index information channels provided online number of certifications
Index service quality index

Copyright # 2003 John Wiley & Sons, Ltd.


QualIndex PrescRqd FrgnSales OtherPrd Products Generics UrgntFee InfoIndex PhoneLine AskPharm PriceCom PriceLis Certifcs VIPPS HON VeriSign
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16)
Quality Safeguards and Regulation of Online Pharmacies

Independent
variables
InsurCov 2.798nnn +b,nnnn 4.408nnn 6.814nnn 60.749nnn 0.874 4.958nnn 1.710nnn +c,nnnn +c,nnnn 0.004 2.006 4.801nnn +d,nnnn +d,e,nnnn 2.780nnn
(0.263) (1.110) (2.298) (7.324) (0.685) (1.199) (0.249) (1.301) (1.350) (0.861) (0.906)
RealAssets 1.782nnn +b,nnnn 2.871nnn 0.842 44.477nnn 2.170nnn 1.430 0.838nnn 1.070 2.336nn 0.320 2.016n 1.196 1.715n 1.567 0.182
(0.294) (0.963) (0.977) (8.313) (0.701) (2.032) (0.249) (0.907) (0.932) (1.251) (1.120) (0.845) (1.004) (1.223) (0.889)
PharmAge 0.311nnn 0.761nn 0.352 1.236nn 6.477nnn 0.311 0.334 0.052 0.192 0.230 0.718 1.160 0.383n 0.015 0.075 0.387
(0.078) (0.333) (0.233) (0.488) (2.180) (0.208) (0.357) (0.074) (0.190) (0.216) (0.574) (0.905) (0.229) (0.367) (0.366) (0.269)
Constant 1.145nnn 3.616nnn 2.586nnn 8.250nnn 9.573nnn 3.018nnn 17.871nnn 0.503nnn 0.641nn 1.953nnn 2.173nnn 3.465nnn 0.440 0.621 4.331nnn
(0.132) (0.707) (0.490) (2.826) (3.638) (0.520) (0.653) (0.124) (0.291) (0.390) (0.522) (0.828) (0.927) (0.939) (0.882)
Observations 104 85 104 113 113 113 65 113 91 91 113 113 113 22 22 113
R2 0.74 0.64 0.22 0.46
Pseudo-R2 0.15 0.50 0.71 0.22 0.02 0.10 0.05 0.24 0.42 0.11 0.09 0.29
a
Logit regressions except OLS in models (1), (5), (7) and (8) and ordered logit (13).
b
The variables InsurCov and RealAssets are dropped due to estimability caused by structural zeroes. As a consequence, only the 85 observations with InsurCov ¼ 0 and
RealAssets ¼ 0 are considered.
c
The variable InsurCov dropped due to estimability – only the 91 observations with InsurCov ¼ 0 are considered.
d
The variable InsurCov dropped due to estimability – only the 22 observations with InsurCov ¼ 1 are considered.
e
Standard errors in parentheses.
n,nn,nnn
Statistically significant at the 99%, 95% and 90% confidence level, respectively. nnnn Structural zeroes.
339

Health Econ. 13: 329–344 (2004)


340 B. Arrunada

Table 7. Basic attributes of services provided by online pharmacies (estimation using interactive age variables)
Dependent variables
Service quality Information provided Third party certification
QualIndex QualIndex InfoIndex InfoIndex Certifcs Certifcs Certifcs
(1) (2) (6) (7) (3) (4) (5)
Independent variables
InsurCov 2.796nnn 2.745nnn 1.988nnn 1.751nnn 7.963nnn 8.095nnn 7.978nnn
(0.423) (0.261) (0.398) (0.242) (1.565) (1.560) (1.523)
RealAssets 2.385nnn 2.386nnn 1.269nnn 0.839nnn 0.082 2.080nn
(0.431) (0.429) (0.415) (0.282) (2.032) (0.933)
PharmAge 0.381nnn 0.374nnn 0.141 0.948nnn 0.908nnn 0.948nnn
(0.096) (0.084) (0.091) (0.300) (0.289) (0.299)
InsurCovnPharmAge 0.029 0.189 1.315nnn 1.375nnn 1.319nnn
(0.191) (0.184) (0.471) (0.468) (0.458)
RealAssetsnPharmAge 0.407n 0.402n 0.327 1.398 1.353nn
(0.214) (0.210) (0.209) (1.259) (0.587)

Constant 1.217 1.210nnn 0.588nnn 0.452nnn


(0.142) (0.135) (0.133) (0.100)

Observations 104 104 113 113 113 113 113


R2 0.75 0.75 0.48 0.46
Pseudo-R2 0.49 0.48 0.49
n,nn,nnn
Note: OLS regressions except ordered logit in models (3)–(5). Standard errors in parentheses. Statistically significant at the
99%, 95% and 90% confidence level, respectively.

VIPPS and HON certificates are found only in Internet certificates seem to act as a substitute
pharmacies in which drugs can be charged to for extending pharmacies’ reputation from the
insurers.m As a consequence, logit models for conventional to the online segments of the market.
variables with such one-way causation are esti- (This effect on third-party certification, which is
mated by dropping the variables with zeroes and not statistically significant for the full sample,
excluding the observations fully determined by the becomes significantly negative when considering
dropped variables, in order to avoid biasing the the interactions with age, as shown in Table 7).
remaining coefficients. This affects models (2), (9), Thirdly, the age of the online operations is
(10), (14) and (15) in Table 6, in which the absence strongly related to most of the dependent vari-
of a coefficient, instead of indicating a lack of ables. Older online pharmacies are more likely to
statistical significance, means quite the opposite: a require a prescription, sell products other than
certain relation between the corresponding depen- medicines, carry a full line of medicines, sell
dent and independent variables. generics and be certified. Specific analysis using
There is also a positive connection between most interactive variables shows that the effect of age on
of these quality factors and the fact that firms also quality of service is concentrated amongst phar-
operate conventional pharmacies. What’s more, all macies operating only online (Table 7). Similarly,
online pharmacies running conventional opera- third party certification is more likely for older
tions also require prescriptions, as described in sites without insurance coverage or conventional
Table 5, causing another case of structural zeroes. operations. This is consistent with the idea that
They are also less likely to sell prescription drugs pure online players increase their quality with time
in foreign markets, and they stock more products but conventional firms diversifying online or
and generics, and provide better information to linked to insurers start their operations by
users. As predicted, they are, however, less likely providing a high-quality service.
to be certified, probably because they use the Coefficients in all models are not only statisti-
reputation of their conventional operations to cally significant but also economically significant.
distinguish them effectively from rogue sites. For instance, insurance coverage and running

Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
Quality Safeguards and Regulation of Online Pharmacies 341

conventional operations increase the quality index their insured parties in a specific pharmacy or
level in half and one-third, respectively, of this grant them a quality certificate.
index’s full range. The effect of pharmacies’ age is To check for the possible effects of endogeneity,
also substantial, with only 5 and 9 years being a system of structural equations was estimated
necessary for age to produce a similar impact to using three-stage regression. The equations con-
that of insurance and running conventional sider the three independent variables, the service
operations, respectively. quality index and the certification and information
indices used in the previous econometric analysis.
The equations are:
QualIndex ¼ b1 þ b2 InsurCov þ b3 RealAssets
Endogeneity of quality determinants þ b4 PharmAge þ e1
The previous section assumed that all the inde-
Certifcs ¼ b5 þ b6 InsurCov þ b7 RealAssets
pendent variables were exogenous. This allowed us
to analyse service quality, information and certi- þ b8 QualIndex þ e2
fication attributes at the level of specific variables.
The analysis, however, could be affected by the InsurCov ¼ b9 þ b10 QualIndex þ b11 Certifcs
endogenous nature of two of the independent þ b12 InfoIndex þ e3
variables – insurance coverage and certification.
When pharmacies decide on their strategy and The results, presented in Table 8, confirm the
quality, they are likely to take into considera- analysis in the previous section – in particular, the
tion both insurance coverage and insurers’ and positive effect of insurance coverage, running
certifiers’ requirements. However, insurers and conventional operations and pharmacies’ age on
certifiers also consider pharmacies’ quality when service quality. Also, the propensity of pharmacies
deciding if they are going to cover purchases by with insurance coverage and without conventional

Table 8. Comparison between reduced-form models and structural equations


Individual OLS estimations System of simultaneous equations
Dependent variables (three-stage estimation)
QualIndex Certifcs InsurCov QualIndex Certifcs InsurCov
(1) (2) (3) (4) (5) (6)
InsurCov 2.798nnn 0.827nnn 3.331nnn 1.076n
(0.263) (0.166) (0.502) (0.570)
RealAssets 1.782nnn 0.596nnn 1.563nnn 0.530nn
(0.294) (0.147) (0.343) (0.231)
PharmAge 0.311nnn 0.272nnn
(0.078) (0.084)
QualIndex 0.160nnn 0.089nnn 0.104 0.076nnn
(0.040) (0.017) (0.133) (0.028)
Certifcs 0.222nnn 0.121
(0.043) (0.185)
InfoIndex 0.063nnn 0.099nn
(0.022) (0.050)
Constant 1.145nnn 0.190nnn 0.134nnn 1.173nnn 0.134 0.162nnn
(0.132) (0.057) (0.024) (0.134) (0.137) (0.055)
Observations 104 104 104 104 104 104
‘R2 ’ 0.74 0.62 0.72 0.73 0.61 0.69
Note: Table entries in columns (1)–(3) report OLS coefficients for column variables regressed onto row variables. Entries in columns
(4)–(6) report coefficients of a system of simultaneous structural equations estimated through three-stage regression. Standard
errors in parentheses.
n,nn,nnn
Statistically significant at the 99%, 95% and 90% confidence level, respectively.

Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
342 B. Arrunada

operations to submit themselves to third-party (00/31/EC). It could be similar to that for


certification. Lastly, as expected from the endo- conventional pharmacies, including equivalent
geneity rationale, quality of service and informa- quality assurance, insurance coverage and correc-
tion positively affect the likelihood of a pharmacy tive policies consisting of entry barriers and
offering its customers reimbursement from its minimum standards of quality and service. If
insurers. deemed necessary, it is also possible to incorporate
online pharmacies into cross-subsidy schemes,
thus reducing the impact on existing pharmacies.
Policy discussion Finally, our empirical findings also suggest that
normal market forces are working in this online
This article shows that the online pharmacy market towards the development of conventional
industry is split into two segments of poor and market-driven assurance mechanisms, based on
high quality, as a consequence of the simultaneous the reputation of market participants and specia-
presence of rigorous licensing in some countries lized third parties. It has been shown, first, that
and ineffective enforcement everywhere. It also insurers demand quality of service. Secondly, firms
identifies the determinants of quality provision of diversifying into online operations start out by
services. In particular, it finds that quality is higher providing higher quality, probably to capitalize on
for pharmacies with insurance coverage, those their reputational assets. Thirdly, pure online
which also operate conventional pharmacies and operators and those without insurance coverage
those which have been open for a longer time. are more likely to use private third-party certifica-
These results suggest that provision of high- tion than those also running conventional phar-
quality service in this industry is possible, can be macies. Lastly, the quality of purely online
regulated by public intervention and can be operators tends to improve with age. These
powered by market forces. Possibility has been findings advise a prudently patient regulatory
demonstrated by the functioning of licensed online attitude, as markets need time to develop safe-
pharmacies. The efficacy of regulation has been guarding institutions adapted to the possibilities of
indirectly shown by the significant impact that new technologies.n Overall, and given also the
insurance coverage, a close proxy of public difficulties to enforce prohibitions effectively,
licensing, has on all measures of quality. Finally, regulators should probably focus their activity on
the importance and compatibility of market forces the setting of standards and the education of
is also clear, as quality is also driven by reputa- consumers. To some extent, this is what the US
tional investments linked to reputational spillovers Federal Drug Administration has been doing with
(proxied through the simultaneous running of respect to e-pharmacies – in addition to investigat-
conventional operations) or accumulated over ing and prosecuting criminal activity, it has been
time through compliance with the promised levels busy in requiring quality certifications, intimidat-
of quality (proxied by pharmacies’ age). ing rogue sites with ‘cyber’ letters and educating
This interpretation supports the lifting of the users in sound Internet purchasing practices.o
current prohibitions of this activity in most Many issues remain open for further research,
European countries. Such liberalization would however. In particular, future studies would
facilitate the benefits of electronic commerce in benefit from additional information on the strate-
this area without causing any drop in service gies followed by new entrants and pharmacies’
quality. On the contrary, the current activity of actual performance. For instance, information on
fraudulent sites would probably be hindered by the advertising expenditures – an alternative source of
functioning of licensed online pharmacies, as the reputational capital – would make it possible to
later would take part of the current demand of examine in greater depth the comparative advan-
fraudulent sites. tage that firms also running conventional phar-
Liberalization and licensing of online pharma- macies seem to enjoy over those that only operate
cies are also compatible with the current systems online. Secondly, on the issue of performance, the
of protective regulation, whatever their real merits article has relied on evidence from the services
are. The regulatory framework for online phar- offered by online pharmacies to estimate the effects
macies would therefore be stricter than that for of quality assurance mechanisms. This evidence
other online operations which are regulated by nicely complements studies that examine the
rules applying the electronic commerce Directive quality of online pharmacies via a sample of real

Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
Quality Safeguards and Regulation of Online Pharmacies 343

purchases, gathering data on the actual prices and prescription history with or without drug name
quality provided. The integration of both ap- display – you choose!; manage your household’s
proaches, however, would be a promising avenue account; personalized health topics and tools – based
for future research. By combining information on on your expressed interests; time-saving non-pre-
scription shopping lists’’ (http://www.merck-medco.
quality assurance and performance, such integra- com/medco/index.jsp).
tion would make it possible to test whether quality d. CVS, an US leader in both conventional and online
assurance explains not only the quality offered but pharmaceutical retailing, estimates that half its
also actual quality. It would also allow the net prescription sales contain some element of urgency
benefits for consumers to be gauged with greater and will never be sold online (‘‘Batalla de precios en
precision. Internet,’’ Diario M!edico, January 13, 2000).
e. Of up to 799 per cent as with ranitidin: 60 tablets
of Zantac cost 98.70 dollars in October 2000 in
Drugstore.com, a certified pharmacy, whereas it sold
Acknowledgements the same amount of generic ranitidin for 10.98 dollars
(accessed October 15, 2000).
The author thanks Merck Company Foundation, the f. ‘‘Prescription for Trouble’’ (Consumer Reports,
philanthropic branch of Merck & Co. Inc., Whitehouse February 2001, 66(2), 19–22).
Station, New Jersey, United States, for the uncondi- g. See [22–24] for a similar interpretation of age as a
tional support received during preparation of this study. proxy of reputation in the service sector
He also thanks Veneta Andonova, Joan Dur!an, Jos!e h. In http://dir.yahoo.com/Business and Economy/Sho-
Manuel Freire, Maia Gu. ell, Albert Oriol, Vicente Ortu! n, pping and Services/Health/Pharmacies/. Six further
Amadeo Petitbo! , Albert Satorra, St!ephane Saussier, pharmacies found using Altavista were added. No
Andreu Segura, Xos!e H. V!azquez-Vicente and Ernesto others were found using other search engines. All the
Villanueva for comments and Petar Balachev and pharmacies that sold more than one product were
Andr!es Jariod for research assistance. Usual disclaimers taken into account. For those offering a single
apply. product, a sample of 39 observations was taken,
giving a sample error of 1 per cent. Internet sites that
only offered information or sold over-the-counter
Notes products were excluded
i. Giving three values to this variable for distinguishing
a. See [2]. Three EU directives are involved in defining these two possibilities did not make any significant
the regulatory environment of European online difference.
pharmacies. In principle, Directive 2000/31/EC on j. See the following sites for information: http://
electronic commerce enables providers to serve www.nabp.net/vipps/intro.asp, http://www.hon.ch/
consumers in all EU member States, and Directive HONcode/Conduct.html and http://www.verisign.
97/7/EC on distance selling makes it possible to com (visited September 4, 2001). To obtain a VIPPS
deliver prescription medicines by mail. However, they certification, which is the most important, an online
may be superseded by national laws which, with the pharmacy must meet 17 criteria about patient
excuse of protecting public health and consumers, confidentiality, prescription security, quality assur-
may in fact be protecting local and conventional ance, and patient–pharmacist consultation [25].
providers. The stringent regime set by Directive 95/ k. Lack of data on alternative reputational investments,
46/EC on data protection might also unnecessarily such as advertising in launching the online operation,
constrain the efficient use of consumers’ databases, an is not likely to be a major problem in this case, given
important comparative advantage of online pharma- the relative values of the assets involved.
cies [3]. A more general analysis of the comparative l. For these, a value of 100 was taken. The signs and
prospects faced by online pharmacies in the US and statistical significance of the parameters and the
the EU is given in [4]. model did not change when other assumptions (35,
b. See [5, 6] on this case, and [3] for a description of the 1000 and 10 000) were introduced.
business strategies of three other pioneer European m. In this sample, it seems sensible to interpret the lack
online pharmacies: the British Pharmacy2U, Allcur- of observations with certain characteristics (zeroes in
es.com and the Swiss Pharmaworld.com. Table 5) as being ‘fixed zeroes’ (i.e. ‘structural’),
c. For example, the services offered by the CVS-Merck- which are impossible to observe, instead of as
Medco on-line pharmacy included the following in ‘random’ (‘sampling’) zeroes, which might have
September 2000: ‘‘Coverage and pricing comparisons simply remained unobserved in the sample [28, pp.
for brand-name and generic medications for both 135–140]. The reason lies in the additional adverse
mail service and retail; point-and-click refills and selection and moral hazard that would arise when
renewals; 12 months of mail service and retail insuring medicine expenditures without requiring

Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)
344 B. Arrunada

that prescriptions be provided by independent third 11. Zbar JD. Pharmacies surge online. Advertising Age
parties. 2000; 71(14): S8–S9.
n. Time is needed also for designing the services with 12. Scanlon B. Vital signs. Inter@ctive Week 2001;
the highest value, as shown by some surveys of 8(11): 26–29.
online shoppers [29], probably because of the 13. Gambardella A, Orsenigo L, Pammolli F. Global
substantial change implicit in e-commerce. It is competitiveness in pharmaceuticals: a European
hardly surprising, then, that developing effective perspective. Report Prepared for the Directorate
safeguards also takes time. General Enterprise of the European Commission,
o. See details of these activities of the FDA in http:// November 2000 (http://pharmacos.eudra.org/F2/
www.fda.gov/oc/buyonline/ (visited September 4, pharmacos/docs/Doc2000/nov/comprep nov2000.
2001). pdf, visited June 8, 2001).
14. Smith MD, Bailey J, Brynjolfsson E. Understanding
digital markets: review and assessment. In Under-
standing the Digital Economy: Data, Tools, and
References Research, Brynjolfsson E, Kahin B (eds). MIT
Press: Cambridge, 2000; 276–288.
1. Goetz J, Lund D. What the law allows. Pharma 15. Armstrong K, Schwartz JS, Asch DA. Direct sale of
Exec 2000; 20(8): 76–84. Sildenafil (Viagra) to consumers over the internet.
2. Bordoni L (rapporteur). Issue Paper. Working New Engl J Med 1999; 341(18): 1389–1392.
Group Manufacturing ‘‘Consumer Goods Industry. 16. Eysenbach G. Online prescribing of Sildanefil
Conference on The Economy in Europe, Brussels, ðViagra1 Þ, on the world wide web. J Med Internet
1–2 March, 2001. Res 1999, 1(2); e10 (http://www.jmir.org/1999/2/
3. Twibell D. Europe takes to the net. Pharma Exec e10/, visited September 1, 2001).
2000; 20(8): 116–122. 17. Tirole J. The Theory of Industrial Organization. MIT
4. Twibell D. E-pharmacies: strategic implications for Press: Cambridge, 1988.
the healthcare industry. Datamonitor (Reuters Busi- 18. Klein B, Leffler K. The role of market forces in
ness Insights, London), June 2000. assuring contractual performance. J Polit Econ
5. Weber W. German controversy over internet 1981; 89: 615–641.
pharmacy. Lancet 2000; 356(9245): 1912. 19. Shapiro C. Premiums for high quality products as
6. Zwick S. M!edecines sans Fronti"eres: A Dutch returns to reputations. Q J Econ 1983; 98: 659–679.
dotcom that fills prescriptions for its neighbors 20. Nelson P. Information and consumer behavior.
challenges Germany’s strict regulations. Time Eur- J Polit Econ 1970; 78: 311–329.
ope, 2001; June 8 (http://www.time.com/time/eur- 21. Darby MR, Karni E. Free competition and the
ope/biz/column/0,9868,101418,00.html, visited June optimal amount of fraud. J Law Econ 1973; 16:
8, 2001). 67–88.
7. European Commission. E-Europe: an information 22. Lafontaine F. Agency theory and franchising: some
society for all. Communication on a Commission empirical results. Rand J Econ 1992; 23(2): 263–283.
Initiative for the Special European Council of 23. Thompson RS. The franchise life cycle and the
Lisbon, 23 and 24 March 2000, COM(1999) 687 Penrose effect. J Econ Behav Organ 1994; 24(2):
final, December 1999 (http://europa.eu.int/informa- 207–218.
tion society/eeurope/news library/pdf files/initiative 24. Arrun* ada B, Garicano L, V!azquez L. Contractual
en.pdf visited June 8, 2001). allocation of decision rights and incentives: the case
8. European Commission. E-Europe: an informa- of automobile distribution. J Law Econ Organ 2001;
tion society for all. Objective 3}Stimulate the 17(1): 256–283.
use of Internet: accelerating e-commerce, 2001 25. Fisher SE. Online health care contends with fraud.
(http://europa.eu.int/information society/eeurope/ Info world 2000; 22(27): 33.
index en.htm, visited June 8, 2001). 26. Woldt J. Internet transforming the pharmacy
9. Henkel J. Buying Drugs Online: It’s convenient business. Chain Drug Rev 1999; 21(17): 1 and 86.
and private, but beware of ‘Rogue Sites’. FDA 27. Welch BL. The generalization of student’s problem
Consumer, January–February 2000, revised on when several different population variances are
June 2000 (http://www.fda.gov/fdac/features/2000/ involved. Biometrika 1947; 34: 28–35.
100 online.html, visited September 4, 2001). 28. Hosmer DW, Lemeshow S. Applied Logistic Regres-
10. Heinrich J. Internet pharmacies}adding disclosure sion (2nd edn.) Wiley: New York, 2000.
requirements would aid state and federal oversight. 29. Yang Z, Peterson RT, Huang L. Taking the pulse of
Report to Congressional Requesters, FDCH Gov- internet pharmacies. Marketing Health Services
ernment Account Reports, 2000; October 19. 2001; 5–10.

Copyright # 2003 John Wiley & Sons, Ltd. Health Econ. 13: 329–344 (2004)

Potrebbero piacerti anche