Voluntary disclosures of payments from pharmaceutical companies to healthcare professionals in Germany: a descriptive study of disclosures in 2015 and 2016.

OBJECTIVE: To analyse voluntary payment reports of pharmaceutical companies to German healthcare professionals (HCPs) in 2015 and 2016 based on an industry-self-regulating transparency codex. DESIGN AND PARTICIPANTS: Cohort study of all German HCPs who voluntarily agreed that at least one payment they received in 2015 and 2016 from pharmaceutical companies is disclosed. MAIN OUTCOME MEASURES: Number of HCPs who disclosed at least one payment in the database; separated by year of disclosure and whether they disclosed once or repeatedly. Amount of disclosed payments and distribution parameters of disclosed annual payment sums per person; separated by year of disclosure und whether they disclosed once or repeatedly.
RESULTS: 28 230 HCPs agreed to the disclosure of at least one payment in the database. In 2015, 19 905 HCPs agreed to the disclosure, decreasing to 15 782 HCPs in 2016. Whereas 7457 disclosed payments in both years, 12 448 disclosed only in 2015 and 8325 only in 2016. Payments of €32 426 721 in 2015 and €23 289 343 in 2016 were disclosed, that is, 27% and 23% of the total amounts spent on HCPs, respectively. Distribution of annual payments was skewed: the top 1% of HCPs disclosed annual payment sums between €17 049 and €200 194, while the median disclosed annual payment sum per person was €536 (IQR €1092). Disclosed payments were higher in male physicians and in physicians with higher academic degree.
CONCLUSIONS: If voluntary, disclosure rates of payments are low and therefore only provide a fragmented picture of interactions between HCPs and pharmaceutical companies. Efforts must be intensified to ensure obligatory disclosure of all payments worldwide. © Author(s) (or their employer(s)) 2020. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.

This is the first study analysing the publicly available payments of pharmaceutical industry to healthcare professionals (HCPs) in Germany over 2 years.

Payment data of a large number of HCPs (n=28 230) are analysed.

Data are given at the individual level and only with the individual consent for disclosure, thereby allowing to analyse disclosure behaviour, that is, how many HCPs disclosed both years or only once.

Disclosed annual payments only give total payments with no specification for which the payments were given.

In cases of non-disclosure it remains unclear whether the HCP had not agreed to disclose or did not receive any payments.

Introduction

Pharmaceutical companies pay large sums to physicians to support their continuing medical education, to pay for lectures, meals, travel, speaking and consulting, and to support drug research.1 2 In 2017, US$8.4 billion were paid by US pharmaceutical companies to about 628 000 US physicians and 1100 teaching hospitals.3 Studies show that in 2017, nearly half of all US physicians received payments by a pharmaceutical company4 and between 2015 and 2017, pharmaceutical companies paid US$5.8 billion to US physicians with estimated mean annual payment values of about US$1100–US$4700 per physician.5 Such financial ties create conflict of interests (COIs), that is, the risk that the professional judgement or action of a physician or researcher is unduly influenced.6–8 Indeed, a large body of evidence exists that industry sponsorship is related to research outcomes2 9–11 and may influence physicians’ prescribing patterns.12–16

Some form of regulation regarding pharmaceutical companies’ disclosure of such financial ties to the public exists in many countries worldwide but these rules vary considerably.17 18 One important difference is whether pharmaceutical companies’ disclosure is based on legally binding laws and therefore mandatory or whether disclosure is introduced through self-regulation codices by the pharmaceutical companies themselves. Legally binding disclosure laws have been implemented in the US through the Physician Payment Sunshine Act,19 but also in Australia, France, Belgium, Portugal, Denmark and Romania.17 18 20 In contrast, some European countries such as Germany, Italy and Spain have no transparency laws but leave transparency entirely to a self-regulatory approach of the pharmaceutical industry, implemented in form of a Disclosure Code by the European Federation of Pharmaceutical Industries and Associations (EFPIA).21 Combined with data protection laws this means in Germany that payments are only made transparent if the individual healthcare professional (HCP) agrees. Therefore, disclosure is voluntary, because companies are not allowed to publish individual information without consent. Another major difference between disclosure policies is whether the disclosed information is aggregated centrally or exists only on the different pharmaceutical companies’ websites. The US Open Payments website, for example, provides a publicly visible search interface that supports the search of the individual HCP and provides analysable files for free download. In most European states, however, the information is only provided on the websites of the pharmaceutical companies in non-searchable and non-analysable documents.

Voluntary disclosure databases have been described for the UK22 and Austria,23 by analysing disclosed payment sums by the pharmaceutical companies. However, to the best of our knowledge, the disclosure behaviour of individual HCPs in voluntary databases has not been examined in detail before. In the current study, we analysed disclosed data from 54 pharmaceutical companies representing about 75% of the German pharmaceutical industry market for prescription drugs.24 25 The database is unique as it not only allows for analysis of the number of HCPs that agreed to the disclosure of their financial ties and the amount of annual disclosed payments, but also allows to examine individual HCPs’ disclosing behaviour, that is, whether HCPs agreed to disclosure only in 1 year or repeatedly in both years. The objective of this study was, therefore, to investigate (1) how many German HCPs agreed to disclose their payments in the database in total and per year, and whether they agreed to disclose repeatedly or only once, (2) how much money was disclosed (A) in total and (B) on average per HCP and (3) how payment amounts differ with respect to disclosure behaviour, year of disclosure, gender and academic degree of the disclosing HCP.

Methods

Database

Since 2016, German pharmaceutical companies organised in the association of voluntary self-regulation in the pharmaceutical industry (Freiwillige Selbstkontrolle für die Arzneimittelindustrie e.V., FSA) have agreed to annually publish documents on their homepages containing information about transfers of value (ToVs) to healthcare organisations and individual HCPs during the previous year, in the context of a transparency codex. Due to data protection laws in Germany, pharmaceutical companies are not allowed to make the ToV to individual HCPs public without their consent. Thus all individual disclosures are voluntary for the HCP. HCPs are free to disclose none, some or all payments made to them. Where consent cannot be acquired, companies publish aggregated data for this year. Irrespective of this, all ToVs related to research and development are disclosed in aggregated form without identification of individuals.26 The codex further determines the location of the disclosure—the companies’ public websites—and provides a non-binding example for the structure of the disclosure document, but otherwise it does not state how disclosure should occur. In the past, this resulted in different document formats from different companies, of which many were non-searchable. The FSA explicitly does not aim to aggregate these data.27 In Germany, first data of the transparency codex were published in summer 2016 for the payments of the year 2015.

The German non-profit investigative journalism newsroom CORRECTIV combined the separate disclosure documents for the years 2015 and 2016 into the database ‘Euros for Doctors’, aiming to integrate the companies’ individual data per HCP. Based on extrapolations using the total amount of payments spent by pharmaceutical companies to HCPs and the disclosed payment sums per HCP, CORRECTIV estimated the number of HCPs who received payments by these companies to be 71 000 in 2015 and 66 000 in 2016.28

The publicly available online version of the CORRECTIV database offers a search tool by which individual HCPs can be searched by name, location or postcode. All individual entries are headed with ‘Mr/Ms (name of HCP) agreed to disclose the following payments:’ and include a table with all ToV the HCP agreed to disclose. Each entry concludes with the overall payment sum the HCP received in that year. In the case that data for 1 year are missing, it is not apparent whether HCPs did not agree to disclose the payment or whether they did not receive any payments that year. CORRECTIV provided us with the data behind this database in form of an analysable excel sheet which contained the following information for each HCP who is listed in the ‘Euros for Doctors’ database: name, academic degree, gender, address, number of companies from which payments were received for 2015 and 2016, and disclosed annual payment sums in 2015 and 2016. We, therefore, could not distinguish between different kind of payments or companies. Data of the pharmaceutical companies’ total payment amounts were taken from press releases by the FSA.24 25

Procedure

Preparation of the database

CORRECTIV’s data collection for the ‘Euros for Doctors’ database was machine based. Parts of the pharmaceutical companies’ documents were in a non-editable format, leading to spelling errors that complicated the correct mapping of payments to HCP. We, therefore, screened the data for duplicates (see figure 1) and matched entries that obviously belonged to the same person but were separated because of, for example, two working addresses or typing errors—2076 double or triple entries were found and could be merged into 1030 entries. A total of 1202 duplicates could neither be traced to one HCP, nor could we identify an obvious error, so we excluded these entries from further analyses.

Figure 1

Decision tree for handling of duplicate entries in the database.

Endpoints of interest

The main outcomes of the analyses were the number of HCPs who disclosed at least one payment in the database, further separated by year of disclosure and disclosure behaviour; and the amount of disclosed payments as well as distribution parameters of disclosed annual payment sums per person, separated by year of disclosure and disclosure behaviour.

Analysis

First, we analysed the number of HCPs who disclosed their received payments in total, dependent on year (2015 vs 2016) and disclosure behaviour (repeatedly vs once). We then examined the total amount of disclosed payments as well as the annual disclosed payments and number of companies from which HCPs received payments per person. To gain insight in the distribution of payment sums, the median, IQR and the 99th percentile was calculated. These analyses were also examined separately for year, disclosure behaviour, gender and academic degree. For quantification of the differences between various groups of HCPs regarding payment amount, we calculated two-sided non-parametric tests and separated the analyses into examination of independent samples on the one hand and dependent samples on the other. For the comparison of two independent samples, Wilcoxon rank-sum tests, and for the comparison of three independent samples, Kruskal-Wallis tests were conducted. For dependent samples, Wilcoxon signed-rank tests were performed. Effect sizes are given as rank-biserial correlation rb (Wilcoxon tests) and epsilon squared ε2 (Kruskal-Wallis test) with the corresponding 95% CI.

Analyses were done in JASP V.0.10.2,29 RStudio, R V.3.6.130 and Microsoft Excel (2011).

Patient and public involvement

Patients or the public were not involved.

Results

Number of HCPs who disclosed payments

In total, 28 230 HCPs disclosed payments in at least 1 year. Figure 2 shows the total number of HCPs separated by year and disclosure behaviour. The total number of HCPs decreased from 2015 to 2016 by 21%. Of all HCPs, 26% agreed to disclose repeatedly in both years, and 74% only in 1 year (44% in 2015, 29% in 2016).

Figure 2

Number of HCPs who disclosed their annual payment sums per year, separated by disclosure behaviour (onetime vs repeated disclosure). HCPs, healthcare professionals.

The proportion of HCPs who disclosed payments in relation to the number of HCPs who received payments (as estimated by CORRECTIV) was about 28% in 2015 (19 905/71 000) and 24% in 2016 (15 782/66 000).

Amount of disclosed payments

Total payment amounts

The investigated 54 companies paid €119 000 000 to HCPs in 2015 and €105 000 000 to HCPs in 2016, resulting in a total of €224 000 000.24 25 Summarised over both years, €55 716 063 (25%) of payments were disclosed in the form of 35 687 annual payment sums. The minimum disclosed annual payment sum per person was €1 and the maximum sum €200 194. The mean annual payment sum per person was €1561 (SD €4221). Half of all HCPs disclosed annual payment sums of €536 or less (IQR €1092). The top percentile included payments equal or larger than €17 049. This indicates a distribution of annual payment sums per person that is strongly skewed to lower payments, see figure 3.

Figure 3

Absolute frequencies of disclosed annual payment sums in respective categories 2015–2016 up to the maximum amount of payment. HCP, healthcare professional.

Disclosed payment amounts in 2015 vs 2016

A higher percentage of the total sum of payments was disclosed to HCPs in 2015 (€32 426 721/€119 000 000; 27%) than 2016 (€23 289 343/€105 000 000; 22%). The total amount of disclosed payments per year decreased by 28% from €32 426 721 in 2015 to €23 289 343 in 2016. The mean annual payment sum per person decreased by 9% from 2015 to 2016 (see table 1 for further descriptive data). In 2015, the top 10 highest annual payment sums per person summarised to €1 048 929, meaning that the ten HCPs with the highest payments disclosed 3% of the total disclosed payments; in 2016, the top 10 highest annual payment sums summarised to €845 597, which is 4% of the disclosed payments.

Table 1

Disclosed annual payments per person and in total, separated per disclosure behaviour, gender and academic degree for each year

	2015						2016						2015 and 2016
	n (paym.)	amount of disclosed annual paym. per person in €					n (paym.)	amount of disclosed annual paym per person in €					total n	total amount in €
		Mean (SD)	Median (IQR)	Min*	99th Perc	Max		Mean (SD)	Median (IQR)	Min*	99th Perc	Max
All	19 905	1629 (4363)	551 (1162)	1	17 429	200 194	15 782	1476 (4033)	516 (1033)	1	16 667	199 775	35 687	55 716 063
Disclosure behaviour
Once	12 448	1021 (2367)	452 (773)	1	10 082	61 380	8325	760 (1490)	387 (660)	1	6352	59 250	20 773	19 035 707
Rep.	7457	2644 (6310)	899 (2102)	4	26 735	200 194	7457	2275 (5544)	790 (1732)	1	23 198	199 775	14 914	36 680 357
Gender
Female	7516	935 (2298)	380 (669)	3	9651	85 062	6122	830 (2103)	347 (619)	1	8547	69 906	11 375	12 108 475
Male	12 385	2051 (5188)	741 (1492)	1	21 866	200 194	9655	1886 (4832)	712 (1353)	1	19 859	199 775	16 848	43 601 289
Academic degree
Professor	1764	5675 (10 740)	2404 (5309)	4	46 493	200 194	1463	4904 (9756)	2062 (3928)	3	50 609	199 775	2211	16 885 282
Dr	8370	1650 (3608)	669 (1301)	3	15 657	86 654	6398	1512 (3371)	629 (1167)	1	15 938	70 923	11 005	23 623 640
None	9771	881 (1817)	393 (686)	1	9040	57 787	7921	813 (1745)	375 (647)	2	7879	54 483	15 014	15 207 142

*In contrast to other databases, there is no threshold for disclosure. N=87 HCPs received annual payments sums ≤ €10; the reasons for these low payments are unclear.

HCP, healthcare professional; Paym, payments; Perc, Percentile.

On average, an HCP received payments from two pharmaceutical companies per year (Mdn=1) both in 2015 and 2016. The top percentile for company count per HCP was 10 in both years, which means that 99% of HCPs received payments by less than 10 pharmaceutical companies per year. The HCP with the highest company count per year received payments by 31 pharmaceutical companies in 2015 and by 28 companies in 2016.

Comparison of disclosed payments: year and disclosure behaviour

Table 1 shows descriptive statistics split up for disclosing behaviour and year of disclosure and table 2 shows the quantification of differences between the analysed groups.

Table 2

Effect sizes and 95% CI for the comparison of various groups regarding annual payment amounts per person

Independent variable	Factor levels	Effect size	(95% CI)
Disclosure behaviour: once
Year	2015 vs 2016	rb=0.07	(0.06 to 0.09)
Disclosure behaviour: repeatedly
Year	2015 vs 2016	rb=0.14	(0.12 to 0.17)
2015
Disclosure behaviour	Once vs repeatedly	rb=−0.31	(−0.32 to −0.29)
Gender	Male vs female	rb=0.28	(0.26 to 0.29)
Academic degree	None vs Dr vs Professor	ε2=0.12	(0.11 to 0.13)
	None vs Dr	rb=−0.24	(−0.25 to −0.22)
	None vs Professor	rb=−0.68	(−0.69 to −0.66)
	Dr vs Professor	rb=−0.51	(−0.53 to −0.49)
2016
Disclosure behaviour	Once vs repeatedly	rb=−0.32	(−0.34 to −0.31)
Gender	Male vs female	rb=0.31	(0.30 to 0.33)
Academic degree	none vs Dr vs Professor	ε2=0.12	(0.11 to 0.13)
	None vs Dr	rb=−0.21	(−0.23 to −0.20)
	None vs Professor	rb=−0.68	(−0.70 to −0.66)
	Dr vs Professor	rb=−0.50	(−0.53 to −0.48)

rb=rank-biserial correlation: positive values indicate that the after scores are smaller than the before scores.

Regarding year of disclosure, those who agreed to disclose only once in 2015 disclosed slightly higher payments than those who disclosed only once in 2016, as indicated by a Wilcoxon rank-sum test with a trivial effect size of rb=0.07 (0.06; 0.09). Those who agreed to disclose both in 2015 and 2016 also showed slightly higher annual payment sums in 2015 than in 2016, indicated by a Wilcoxon signed-ranked test with a small effect size of rb=0.14 (0.12; 0.17).

Regarding disclosure behaviour, those HCPs who disclosed only in 1 year disclosed lower payments than HCPs who agreed to disclose in both years, with a medium effect sizes as indicated by a Wilcoxon rank-sum test, rb=−0.31 (−0.32; −0.29) (see table 2 for similar results for 2016).

Comparison of disclosed payments: gender and academic degree

The descriptive annual and total payment sums and absolute frequencies separated by gender and academic degree can be seen in table 1. Of all 28 230 HCPs, 40% were female and 60% were male. Regarding academic degree, 8% had a German postdoctoral degree (‘habilitation’ / ‘Prof.’), 39% had a doctoral degree but no habilitation, and 53% had no academic degree stated.

The proportion of female HCPs in our sample was higher for those who agreed to disclose only once: 44% female vs 56% male, compared with those who disclosed repeatedly: 30% female vs 70% male. In 2015 (see table 2 for similar results for 2016), male HCPs disclosed higher annual payment sums than female HCP, indicated by a Wilcoxon rank-sum test delivering a small effect size: rb=0.28 (0.26; 0.29).

The proportion of habilitations and doctoral degrees in our sample was higher for those who disclosed only once: 6% habilitation vs 35% doctoral degree vs 59% no academic degree, compared with those who disclosed repeatedly: 14% habilitation, 50% doctoral degree and 36% no academic degree. Academic degree had a medium effect on disclosed payment sums, ε2=0.12 (0.11; 0.13), as indicated by a Kruskal-Wallis test. Pairwise comparisons indicate that habilitated HCPs disclosed higher annual payment sums than doctoral HCPs and those without academic degree; and doctoral HCPs disclosed higher annual payment sums than those without academic degree (see table 2 for further details). Median annual payment sums including IQR, separated by year of disclosure, disclosure behaviour, gender and academic degree are depicted in figure 4.

Figure 4

Median annual payment sums with IQR separated by year, disclosing behaviour, gender and academic degree. HCP, healthcare professional.

Discussion

Principal findings

To our knowledge, this is the first analysis of a 2-year database of voluntary disclosures of payments from pharmaceutical companies to HCPs allowing for a detailed description of disclosing behaviour of individual HCPs. Our analyses showed that the amount of disclosed payments, as well as the number of HCPs who agreed to disclose payments, decreased over the period of 2 years. Furthermore, of all HCPs disclosing at least one payment in 2015 and 2016, only 26% disclosed payments in both years. The disclosures by German pharmaceutical companies can, therefore, be described as a very fragmented and limited insight into the actual interactions with German HCPs. The distribution of disclosed annual payment amounts was skewed to lower amounts of payments with half of all disclosing HCPs disclosing annual payment sums up to €536, while one percent of HCPs disclosed payment sums of €17 049 to €200 194. This is comparable to the distribution of non-voluntary disclosed payments in the US5: a large proportion of the total payments is concentrated among few physicians, whereas many physicians receive comparable smaller amount of payments.

We observed that on average, female HCPs disclosed lower amounts of payments than male HCPs and HCPs with lower academic degrees disclosed lower amounts of payments than HCPs with higher academic degrees. This is comparable to studies from the US that have shown that male physicians,4 5 31 physicians who have been practising longer, and those who graduated from a top-ranked US medical school receive higher industry payments.5 These factors, however, might interact at various levels: for example, women are less likely to attain senior-level positions, and therefore, are more likely to have lower academic degrees.32 33

Findings in context

It is often argued by pharmaceutical companies that adherence to self-regulation codices like the EFPIA provides a sufficient basis for transparency of payments to HCPs.26 Based on our data, we would argue against this view as only approximately a quarter of all payments were disclosed. Comparison of disclosure rates to other countries providing voluntary transparency of payments is difficult since transparency codices can be implemented in many different ways and because companies did not report the number of all HCPs who received payments. However, analyses of comparable European databases for the year 2015 and 2016 show that the proportion of disclosed payments to HCPs in 2015 was higher in Switzerland (estimated 56%)34 and Great Britain (48%),22 but even lower in Austria (17%).23 In contrast to our finding that disclosure rates decreased from 2015 to 2016, disclosure rates increased from 2015 to 2016 in Switzerland and Great Britain. Even more problematic to analyse is the proportion of HCPs agreeing to disclose payments because pharmaceutical companies in Germany did not give the exact number of HCPs who received payments. Indeed, we only had estimates of those numbers by CORRECTIV and thereof calculated consent rates of 28% (2015) and 24% (2016).28 In other countries, estimates show a development of consent rates from 2015 to 2016 of 22%–21% in Austria23 and of 56%–65% in UK.22 The reasons for these discrepancies between countries are unclear but might be found in the different implementations of the EFPIA. Even within one country, a transparency guideline might be implemented differently by different pharmaceutical companies. The German example shows that there is high variance of consent rates between the companies, ranging from 91% for GlaxoSmithKline (GSK) to 16% for Genzyme.35 In reply to an enquiry by CORRECTIV, the pharmaceutical companies with the highest consent rates stated that they actively endeavoured to reach HCPs’ consent, for example, by multiple inquiries if an HCP did not respond. GSK stated that they prefer working with physicians who agree to disclose ToVs and may quit collaborating with HCPs who are not supporting transparency. The companies with the lowest disclosure rates did either not provide a statement, or stated that they requested written consent by HCPs.35 Other factors that influence the consent rate might lie in the communication and framing of the transparency guideline towards the public. Some physicians in Germany reported that they did not disclose because they were afraid the public might get a ‘false impression’ by the disclosed information.36 Unfortunately, we have no data regarding these factors, nor on the further development of disclosure rates, as the CORRECTIV database was only available for 2015 and 2016.

Strengths and weaknesses

A strength of our dataset is that it allowed for an analysis of the individual disclosing behaviour of HCPs over 2 years. Our analyses showed that only 26% disclosed repeatedly. A major weakness of the given database was that no information was available whether HCPs disclosed all their payments or just some; and whether HCPs without an entry did not receive any payments or did receive payments but did not agree to disclose. This left us with the uncertainty that HCPs without an entry in one or both years could either have decided against disclosure or not have received any payments. An argument for the latter is that the group of HCPs who only disclosed in 2015 disclosed smaller annual payment sums than those who disclosed repeatedly. This group may, therefore, consist of HCPs who sometimes receive smaller payments and sometimes none at all, whereas repeatedly disclosing HCPs sometimes receive higher payments and sometimes lower payments. If this is the case and payments ‘naturally fluctuate’, however, the consent rate between the years should have been similar, where in reality it decreased from 27% to 22%. Also, in 2015, the ratio of those who disclosed only once was higher than 2016—63% vs 53% of all disclosing HCPs. This might be a clue that in 2016, more HCPs decided against disclosure. However, from the current body of evidence, we can only speculate.

A further weakness was that the database was not the original database as for example, in analyses of Open Payments/Disclosure UK but collected by a third party. This may have created transcription and transmission errors. We worked hard to omit obvious transcription errors by conscientiously examining the duplicates in the database, merging obvious duplicates and deleting cases in which it was unsure whether a transcription error was apparent from the analysis. A standardised procedure that not only builds and maintains a centralised database, but compares disclosed and actually transferred payments and provides the opportunity for HCPs to feedback errors in their entries would contribute to full transparency.

Policy implications

Our data are of relevance for policymakers. Voluntary disclosures of payments lead to low numbers of physicians agreeing to disclose and are dependent on how the guidelines are interpreted and implemented by those responsible, resulting in a fragmented picture of the interactions between HCPs and pharmaceutical companies. Full transparency of such payments, therefore, can only be provided if disclosures are required by legislation. In our view, full transparency is necessary for several reasons. Only full transparency provides an unbiased view on the true degree of financial ties between HCPs and pharmaceutical companies. This is the basis for all further measures of COI regulation.20 37 38 In addition, patients can use this information to make informed decisions about which physicians to place their trust in, and, under certain circumstances, how to interpret their physicians’ recommendations.20 37–40 All data should be made transparent through user-friendly, searchable and downloadable databases to allow for continuing analysis of interactions between HCPs and pharmaceutical companies.

Conclusion

This example of a self-regulating transparency codex of German pharmaceutical companies shows that only every third to fourth HCP decides to make their payments by pharmaceutical companies voluntarily transparent, and consent rates decrease over the years. The analysis shows that only a quarter of those HCPs who disclosed 2015 or 2016 disclosed payments in both years. To gain true transparency, disclosed information must be complete. Only on that condition the disclosed information adds value to current discussions about COIs and only then further regulation mechanisms can be discussed and implemented.