Abstract
In this paper we ascertain that the Italian market for medical devices is characterized by significant price dispersion. We have, therefore, carried out an econometric analysis, as well as a Bayesian network analysis to verify if price dispersion is due to price discrimination. We have found that ASLs (Aziende Sanitarie Locali) incur higher procurement costs than AOs (Aziende Ospedaliere), which purchase larger quantities as Centralized purchasing agencies do. Consequently, second-degree price discrimination may be one of the causes of price differences. Price levels are also inversely related to product age because of intense innovative activity, making product differentiation more likely than price discrimination. Public procurement agents located in Southern Italy pay higher prices than those located in Northern or Central Italy. This is due to the higher probability for Southern procurement agents to purchase from independent wholesalers, rather than from producers, implying a double marginalization effect which raises final prices. It is also more likely that obsolete medical devices are sold to Southern health care providers.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
For example, as pointed out in Ministero della Salute (2014), a direct information flow from the public medical device procurers to the Italian Ministry of Health was introduced in 2010, in order for the latter not to have to approximate the value of medical device consumption (decay) anymore, but to be able to assess it precisely at the device and hospital levels.
The HHI is calculated as the sum of squared market shares \(s_{i}\) on the \(N\) firms in the market, i.e. \(\sum\nolimits_{i = 1}^{N} {s_{i}^{2} }\). Especially for very asymmetric distributions of market shares, the resulting HHI will be rather high, with a maximum value of 1 for a perfect monopoly. It should be noted, however, that the HHI measures market concentration, which is not necessarily accompanied by higher market power, i.e. the ability of firms to charge higher relative markups.
More than 12,200 patent applications were filed with the European Patent Office (EPO), representing 7.7% of the total; 41% of these applications came from European countries (EU28, including Norway and Switzerland) and 59% from other countries, the majority of which arrived from the US (38%).
According to Stigler (1987), whose definition is the one used in the relevant price discrimination literature as in Varian (1989), Tirole (1988), or Stole (2007). As far as this definition is concerned, a firm engages in price discrimination, when it sells two similar products at price ratios that are not equal to the marginal cost ones, i.e. when it charges different relative markups.
In this paper, we are focusing on publicly procured medical devices mostly due to space and data constraints, but also because many privately paid-for medical devices are traded in rather standard commodity market environments, as for example, dental implants.
These models, however, do not take into account other, more dynamic factors that could potentially be relevant such as firms’ general reluctance to lose market shares in an uncertain future. For differentiated goods and physician preferences, winning a large procurement contract might also imply larger future profits from lock-in effects.
This analysis is based on supply contracts exceeding the EU publication threshold. Smaller contracts that are published only in Italy are excluded. The overall incidence of centralized procurement is therefore likely to be lower.
It should be noted that the underlying reasons for such strong preferences can be legitimate, but also illegitimate, e.g. connected to corruption. In addition, the practice of sponsoring physicians’ congresses in attractive locations and paying for doctors’ expenses might be able to influence their preferences. In a number of cases, it is not entirely clear where marketing methods end, and more illicit practices begin. This topic, however, goes beyond the scope of this paper.
The CND is a code starting with a letter, say C for cardio-circulatory devices, to which 2-digit sequences are added for an increasingly refined product grouping, so that we have C01 for arterial-venous devices, C0104 for angiography and hemodynamic devices, etc., until we reach the most disaggregated stage of, C010402020101 for example, denoting vascular dilatation balloon catheters for percutaneous transluminal angioplasty (PTA), a product group that constitutes by far the most frequently occurring type of medical device in our dataset with 1585 observations out of the total 7612.
Benchmarking Prezzi e Acquisti (BPA). Thanks to this service, healthcare providers can contact CBIM with a detailed list of medical devices that they either plan to purchase or for which a preliminary procurement procedure has already been carried out, with the aim of acquiring information about prices other healthcare providers have paid for them in the last 36 months.
The fact that there are 5 or more price observations at the ID level, for only 189 out of the 2886 comparable devices, seems to be a relevant problem at first sight. However, we can still make these inferences at an aggregated level by adequately constructing the dependent variable, although it could prevent inferences about the extent of price dispersion and its determinants for a single medical device type.
Our proxy of innovation can be considered also at the ID level, to the extent that we can link every ID to a given RDM.
Apart from these robustness tests, we shall generally disregard the time dimension in our analysis, because we have treated data as a pooled cross section. There is no need to treat it as a panel data set since data were sampled over a time span and do not follow specific procurers over the relatively short period from 2010 to 2014.
See “Appendix I” for further details.
We will omit summary statistics of the relative age difference variable since it is, centered around an average of about zero with 99% of the values ranging from − 80% to + 80%, similarly to the price PAM variables.
As a side note, we did not create distinct variables for these two hospital types because of the rather low number of observations and due to the fact that IRCCSs and university AOs/polyclinics sometimes seem to coincide.
Estimations are available on request.
In “Appendix II” we will present a model with regional organization type dummies, i.e. the integrated, semi-integrated, separated and semi-separated one. We cannot use the regional organization type and the south/island dummy in the same model given the high correlation among dummies (the correlation between the semi-separated dummy and the southern one is 0.8).
For this analysis, we have dropped all observations with procurers being of the other type (not centralized) as specified in Table 4 in the previous subsection.
Penalizing complexity is crucial, because a complete network structure, i.e. where every node is connected to every other node, will naturally always be the most accurate, given the data, when assigning the right parameters. As already mentioned above, a complete graph does not have any additional explanatory power if compared to the intractably large joint probability distribution among all our variables. When the parameters of a complete model are learnt from the data, we will probably encounter the problem of overfitting, since the resulting model will exactly represent the sample but possibly not the underlying data generating the network.
For a complete overview of Bayesian network analysis, see Nielsen and Jensen (2009).
The relevant positive correlation between the South/Islands dummy and the product age is probably not representative of the underlying real-world purchasing processes, but it is grounded in the way data have been sampled. Since the BPA requesting parties, which are both from the South, represent a large part of the more recent data derived from the years 2013 and 2014, the products they buy are naturally older than the ones of the public procurement agents they are compared with using market research on procurement procedures carried out in the past.
To test this result, we have set up a dummy variable to indicate when a device has been sold by an intermediary in the South/Islands area; this variable is always positive and highly consistent (p < 0.05) for all model specification with a coefficient value ranging from 16.72 to 34.28 per cent on PAM measurement.
By collecting data both from the National Firms Register (“Registro delle Imprese”) and the Orbis data base, we have observed that in the Southern and Islands regions the share of producers amounts to 2%, while the wholesalers’ is 91%; the residual share of 7% accounts for integrated firms.
Of course, the distinction between AOs and ASLs as purchasers is relevant only for those regions where this distinction is valid (like, e.g. Lombardy, Sicily, and Lazio). In others purchase is centralized by ASL, and of course, no difference arises.
References
Assobiomedica (2010). La posizione associativa in tema di prezzi dei dispositivi medici: Attenzione ai confronti impropri. http://www.assobiomedica.it/static/upload/pp_/pp__prezzidispositivi-03.pdf. Accessed Dec 2017.
Bonaccorsi, A., Lyon, T. P., Pammolli, F., & Turchetti, G. (2000). Auctions vs. bargaining: An empirical analysis of medical device procurement. LEM Working Paper Series. 1999/20.
Boscolo, P. R., & Tarricone, R. (2013). La rilevanza economica e sociale del mercato dei dispositivi medici. caratteristiche distintive del prodotto, delle imprese produttrici e delle relazioni con il settore pubblico. Aidea 2013 Conference Papers.
Burns, L. R., & Lee, J. A. (2008). Hospital purchasing alliances: Utilization, services, and performance. Health Care Management Review, 33(3), 203–215.
Centro Studi Assobiomedica (2012a). I tempi medi di pagamento delle strutture sanitarie pubbliche. Studi n.24. https://www.assobiomedica.it/it/pubblicazioni/index.html. Accessed Dec 2017.
Centro Studi Assobiomedica (2012b). Produzione, ricerca e innovazione nel settore dei dispositivi medici in Italia. Rapporto. https://www.assobiomedica.it/it/pubblicazioni/index.html. Accessed Dec 2017.
Centro Studi Assobiomedica (2013). I tempi medi di pagamento delle strutture sanitarie pubbliche. Studi n.25. https://www.assobiomedica.it/it/pubblicazioni/index.html. Accessed Dec 2017.
Centro Studi Assobiomedica (2014). Average payment times of public and private health-care organisations. Studies n.28. https://www.assobiomedica.it/it/pubblicazioni/index.html. Accessed Dec 2017.
Centro Studi Assobiomedica (2015a). I tempi medi di pagamento delle strutture sanitarie pubbliche e private. Studi n.31. https://www.assobiomedica.it/it/pubblicazioni/index.html. Accessed Dec 2017.
Centro Studi Assobiomedica (2015b). Le politiche pubbliche d’acquisto di dispositivi medici. Studi n.30. https://www.assobiomedica.it/it/pubblicazioni/index.html. Accessed Dec 2017.
Centro Studi Assobiomedica (2016). Le politiche pubbliche d’acquisto di dispositivi medici, secondo aggiornameto. Studi n.33. https://www.assobiomedica.it/it/pubblicazioni/index.html. Accessed Dec 2017.
Dimitri, N., Dini, F., & Piga, G. (2006). When should procurement be centralized? Handbook of procurement, 47–81. Cambridge: Cambridge University Press.
Eucomed (2014a). The European medical technology industry in figures. http://www.eucomed.be/uploads/Modules/Publications/20141003-medtech-brochure-digital-1.pdf. Accessed Dec 2017.
Eucomed (2014b). Key principles of smart procurement for medical devices. http://www.medtecheurope.org/uploads/Modules/Publications/20141020_eucomed-key-principles-of-smart-procurement-for-medical-devices-final.pdf. Accessed Dec 2017.
European Council Directive 93/42/EEC of June 14 (1993).
Ferré, F., de Belvis, A. G., Valerio, L., Longhi, S., Lazzari, A., Fattore, G., Ricciardi, W., & Maresso, A. (2014). Italy: Health system review. European Observatory on Health Systems and Policies. http://www.who.int/iris/handle/10665/141626.
Grennan, M. (2013). Price discrimination and bargaining: Empirical evidence from medical devices. American Economic Review, 103(1), 145–177.
Hahn, R. W., Klovers, K. B., & Singer, H. J. (2008). The need for greater price transparency in the medical device industry: An economic analysis. Health Affairs, 27(6), 1554–1559.
Italian Legislative Decree n. 46 of February 24 (1997). Attuazione della direttiva 93/42/CEE concernente i dispositivi medici. Gazzetta Ufficiale della Repubblica Italiana n. 54 del 6 marzo 1997.
Lerner, J. C., Fox, D. M., Nelson, T., & Reiss, J. B. (2008). The consequence of secret prices: The politics of physician preference items. Health Affairs, 27(6), 1560–1565.
Medtech Europe (2018). The European Medical Technology industry in figures. http://www.medtecheurope.org/sites/default/files/resource_items/files/MedTech%20Europe_FactsFigures2018_FINAL_1.pdf. Accessed Apr 2018.
Ministero della Salute (2014). Rapporto sulla spesa rilevata dalle strutture sanitarie pubbliche del SSN per l’acquisto di dispositivi medici - anno 2013. http://www.salute.gov.it/imgs/C_17_pubblicazioni_2259_allegato.pdf. Accessed Dec 2017.
Neapolitan, R. E. (2004). Learning Bayesian Networks (Vol. 38). Upper Saddle River: Prentice Hall.
Nielsen, T. D., & Jensen, F. V. (2009). Bayesian Networks and Decision Graphs. New York: Springer.
Novi, D., & Rizzi, Z. (2017). Scale effects and expected savings from consolidation policies of Italian local healthcare authorities. Applied Health Economics and Health Policy., 16(1), 107–122.
Pammolli, F., Riccaboni, M., Oglialoro, C., Magazzini, L., Baio, G., & Salerno, N. (2005). Medical devices competitiveness and impact on public health expenditure. Brussels: Enterprise Directorate-General, European Commission.
Pauly, M. V., & Burns, L. R. (2008). Price transparency for medical devices. Health Affairs, 27(6), 1544–1553.
Sorenson, C., & Kanavos, P. (2011). Medical technology procurement in Europe: A cross-country comparison of current practice and policy. Health Policy, 100(1), 43–50.
Stigler, G. (1987). Theory of price. New York: Macmillan.
Stole, L. A. (2007). Price discrimination and competition. Handbook of Industrial Organization, 3, 2221–2299.
Tirole, J. (1988). The Theory of Industrial Organization. Cambridge: MIT Press.
Varian, H. R. (1989). Price discrimination. Handbook of Industrial Organization, 1, 597–654.
Vellez, M. (2011). Auctions versus negotiations: Evidence from public procurement in the Italian healthcare sector. CEIS Working Papers. N 191.
Vellez, M. (2012). Determinants of price discrimination in the acquisition of medical devices. CEIS Working Papers. 235.
Wooldridge, J. (2009). Introductory econometrics: A modern approach (4th ed.). Boston: Cengage Learning.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We are grateful to two anonymous referees for their detailed comments and advices, that helped us to improve a previous version of this paper. We would like to thank CBIM (Consorzio di Bioingegneria e Informatica Medica) for allowing us to share their data on medical devices, and especially Paolo Cristiani and Stefania Pazzi for interesting discussions about the Italian market. We are also grateful to Claudia Tarantola for introducing us to the Bayesian Network Analysis, and to Valentina Falleri for her commitment in setting up a useful Data Base for our research. Finally, our deepest thanks to Federico Desperati and Marina Scarmato for their brilliant research assistance and to Roseanne Rogosin for her careful revision of the text. Any remaining errors are our responsibility.
Appendices
Appendix I
Table 7 shows the top 10 of most frequently occurring producers and suppliers in our data set. As can be seen, the three prevalent firms lead both categories in terms of frequency although the gap is much smaller for the final suppliers.
Appendix II: The regional organization model
Table 8 shows four different OLS regression specifications using three different dependent variables. The first specification uses the CBIM ID to define groups of comparable medical devices and includes all observations with at least two devices per group. The second equation also uses the CBIM ID as dependent variable but includes only devices with at least 3 comparable observations. The third comprehends yearly dummies. The fourth specification uses all comparable devices with at least two observations at the RDM level, this is a slightly more aggregated product specification. As can be checked, the three different Regional organization dummies in this specification have a negative impact on prices compared to the baseline case: the semi-separated model. In more detail, the separate organization model performs better, with an average reduction of PAM measure ranging from − 34.02 to − 21.93. The other two regional models also predict a significant and negative effect on prices ranging from an average of − 30.40 to − 18.88 for the integrated regional model, and from − 34.24 to − 15.04 for the semi-integrated one. When interpreting these results, we have to consider that the observations for the separated model are collected in Lombardy (the only region characterized by this model), accounting for 2586 out of 4017 observations collected in the North/Centre area. In the semi-separated model, Sicily and Campania also represent almost all of our observations (2672 out of 2806). These two Regions express 2672 out of 3299 collected in the South/Islands area. For these reasons we believe that the specification with regional models replicates, to some extent, the previous results obtained using the South/Islands dummy (Table 6), where we have found evidence about higher prices in the South/Islands macro-region compared to the Northern one. It is also interesting to point out that this same argument is valid for the comparison between the baseline model and the integrated and the semi-integrated ones. Because both models perform better compared with the baseline one and still the observations for these two models come mainly from regions located in Central and Northern Italy. The coefficients obtained from control variables in Table 8 confirm the result shown in Table 6.
Rights and permissions
About this article
Cite this article
Crea, G., Cavaliere, A. & Cozzi, A. Price discrimination in the Italian medical device industry: an empirical analysis. Econ Polit 36, 571–608 (2019). https://doi.org/10.1007/s40888-019-00149-5
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s40888-019-00149-5