Analysis of Pharmaceutical Industry Payments to UK Health Care Organizations in 2015

Key Points Question How are drug company payments to health care organizations distributed in the UK health care system? Findings This cross-sectional study of the Disclosure UK database found that in 2015, 4028 health care organizations received US $72 110 156.6 from 100 companies. Although financial relationships were spread across the health care system, a few key donors and beneficiaries of industry funding were found. Meaning More policy attention is needed to disclose organizational conflicts of interests, particularly in areas of the health care system with a high concentration of industry payments.

This supplementary material has been provided by the authors to give readers additional information about their work.

eBox 2. Definition of Health Care Organization According to the European Federation of the Pharmaceutical Industry and the Association of the British Pharmaceutical Industry
"Any legal person (i) that is a healthcare, medical or scientific association or organisation (irrespective of the legal or organisational form) such as a hospital, clinic, foundation, university or other teaching institution or learned society (except for patient organisations within the scope of the EFPIA PO Code[EFPIA Patient Organisation Code of Practice]) whose business address, place of incorporation or primary place of operation is in Europe or (ii) through which one or more HCPs provide services." (EFPIA Code on Disclosure of Transfers of Value to Healthcare Professionals and Organisations and ABPI Code of Practice for the Pharmaceutical Industry) Organizations focused on providing material support for patients or healthcare organizations third-sector organization (non-charity) focused on providing funding or material support to patients or NHS organizations charity focused on providing funding or material support to patients or NHS organizations Organizations focused on providing patient support third-sector organization (non-charity) focusing on providing patient support charity focused on providing patient support (membership organization)

eMethods. Details of Data Collection and Management
This eAppendix provides the full list of steps taken in preparing Disclosure UK data at the HCO level for analysis for the purposes of this study.
The eAppendix follows the structure of the Methods section of the paper.

Version of Disclosure UK
We analyzed the Disclosure UK dataset version 20160630, published on the 1 st July 2016, and downloaded from the ABPI website on the same day. 1 We note the version and the dates because the have noticed that ABPI sometimes introduces minor changes to the dataset following its initial publication. The extent and nature of these changes are difficult to establish. The versions of Disclosure UK released in subsequent years have a different format so all statements made in this eAppendix refer to this particular version of Disclosure UK.

Identification of Healthcare Organizations in Disclosure UK
As the 2015 version of Disclosure UK does not have a separate column allowing for distinguishing between payments made to Healthcare Professionals (HCPs) and Healthcare Organizations HCOs, we extracted payments (otherwise termed Transfers of Value) to HCOs by filtering out rows in the dataset with no values entered in the "First Name" and "Last name" columns, which are filled for payments made to HCPs but not HCOs. We assumed that the lack of entry in those columns indicated that the payment recipient was a HCO.

Variables in Disclosure UK associated with HCOs
There are 13 variables applicable to HCOs in the version of Disclosure UK we analyzed: Company Fee for service and consultancy include two more detailed payment types: Fees and Related expenses agreed in the fee for services or consultancy contract.

Donations and Grants to HCOs and Joint working appear in both paymentcategories and payment
types.
For the purposes of our study, we rely on payment categories rather than on payment types as they have a smaller number of more comprehensive values.

Cleaning payment values
Before proceeding to analysis, we turned 20 payments with negative values, worth $160,210.2, into positive ones. We assumed that the negative values resulted from the lack of standardisation of accounting systems between and within companies.

Three levels of categorization
Our categorization framework has three primary levels.
 The detailed level captures unique characteristics of payment recipients, including the regional differences between HCOs based in England, Scotland, Wales and Northern Ireland.

Supplementary principles of categorization
We introduced supplementary principles of categorization so as to  distinguish organizations from the third sector from organizations from the private, public and mixed (public and third) sectors

Development of the categorization framework for healthcare organizations and its validity
We developed the categorization framework for HCOs iteratively. This process involved the following steps.
 We started by coding all payment recipients using detailed categories. Subsequently, we aggregated the detailed categories to create general and top-level categories. We were refining the category names at all levels throughout the categorization process. This involved merging or splitting existing categories as well as developing new ones. These revisions were intended to allow for organizations to be grouped together more adequately. We also sought to reduce the number of codes, especially at the detailed level, to maximize the clarity of the presentation of findings.
 Each level of categorisation was applied in relation to the recipient of each payment. The presence of three levels, including the detailed one, was crucial for ensuring the validity of decisions about placing payment recipients in specific categories. The categories associated with each payment recipient can be scrutinised in the Online Supplement linked with this publication.

Sources of data
We extracted data from the healthcare organizations' main webpage, and when necessary, the "About us", "History" or "Membership" sections. If an organizational website was not available, we used other available websites mentioning the organization to minimize the number of payments with unclear recipients. In particular, we checked the following websites.
 The Companies House website 5 -to establish whether a healthcare organization was a private company  The websites of the Charity Commission for England and Wales, Scottish Charity Regulator and the Charity Commission for Northern Ireland to establish whether a healthcare organization was a registered charity.
Dealing with discrepancies between Disclosure UK and website data In rare instances of discrepancies between recipient names reported in Disclosure UK and on the organizational websites we relied on the information provided on the websites when assigning the organization categories to payment recipients. For example, if a health center (Disclosure UK) appeared as a medical practice during the web search it was coded as medical practice. All instances of differences between Disclosure UK and the websites were minor and affected only the detailed level of coding.

Timing of web searches
The timing of the web searches (February-May 2017) was close to the time when the payments were made (until the end of 2015, with the database released in June 2016). Therefore, it is reasonable to assume that the results of the web searches provided an accurate capture of the nature of healthcare organizations.
With very few exceptions, the healthcare providers listed in Disclosure UK still existed at the time of the searches. However, some private sector companies providing health services and private companies other than providers of health services had ceased to exist. In a vast majority of these cases we were able to establish their status as closely as possible to 2015 based on the Companies House Website.
Inter-coder reliability Two researchers, ER and PO, coded half of the dataset each at the detailed level, with any revisions being discussed and adopted in real time. PO then coded the payments using the general and top-level categories and checked the consistency of coding between and within categories at detailed, general and top-level. The consistency checks focused on ensuring that  organizations covered by the same code were of the same nature;  organizations covered by different codes were sufficiently different;  each organization was assigned one category at the detailed, general and top-level.

Distinguishing between different sectors of healthcare provider organizations
When deciding about the sector of healthcare providers (public, private, third) we applied the following principles:  Public sector secondary and tertiary care providers -we assumed that all hospitals were NHS hospitals (that is, public sector hospitals) unless it was specifically stated on their main website that they were private healthcare organizations or alternative healthcare providers (e.g. social enterprises, community interest companies or charities). However, if a ward or unit within an NHS trust provided services exclusively for private patients we categorized it as a private sector healthcare provider.
 Public sector primary care providers -we assumed that all primary care organizations (e.g. surgeries, health centers, clinics) were public sector if they were listed on the NHS website or had an NHS logo on their website. We did not investigate separately whether they provided additional health services for private patients.
 Alternative healthcare providers -we assumed that a healthcare organization was an alternative healthcare provider if its main website specifically mentioned that it was a community interest company, social enterprise, or a charity.

Notation of "Names" and "Locations"
We use the terms "Names" and "Locations" in inverted commas when referring to the way in which they were used in Disclosure UK. We do not use inverted commas when referring to the approach to naming HCOs established for the purposes of this study.

Approach to reporting HCO names taken in this study
As noted in the paper, we report our results by HCO names and categories based on "Locations" unless they were categorized as unclear; in that case they were substituted by "Names". There were three reasons. First, "Locations" had fewer unclear recipients (1,987, 9.9%) than "Names" (2,940, 14.7%).
Second, using "Locations" minimized the loss of information associated with arriving at a single level of aggregation for the HCO category receiving the greatest number and value of payments, that is secondary and tertiary public healthcare providers (see Results). Specifically, both "Names" and "Locations" comprised a combination of NHS trust and hospital names, but the share of the former was considerably greater in "Locations". It was therefore straightforward to replace NHS hospitals in "Locations" by their respective trusts, and not the other way round (NHS trusts typically comprise more than one hospital). Third, focusing on the trust level allowed for reducing the complexity of the dataset and comparing the number of HCOs with other main UK healthcare statistics (e.g. NHS Confederation data on NHS trusts).
The HCO categories in "Names" and "Locations" disagreed in relation to 2,093 (10.4%) payments.
These payments were spread across different donors, HCO and payment categories. We kept them in the analysis as the general rule of prioritizing information from "Locations" allowed for resolving these discrepancies; for example, we interpret 130 payments to primary and 980 to secondary care organizations ("Names") as having been made to healthcare commissioning organizations ("Locations") which fund their activity.

Addressing inconsistencies in HCO names reported in Disclosure UK
In creating the list of HCO names reported here, we addressed inconsistencies in the naming of HCOs in Disclosure UK, such as the same HCO being referred to with different names (e.g. "NHS Bristol CCG [Clinical Commissioning Group]" and "Bristol CCG") and different HCOs appearing under the same name (e.g. Grove House Surgery in West Yorkshire and Somerset).
First, we standardized the original HCO "Names" (4,224, 21.1% entries changed) and "Locations" standardized HCO "Names" (1,167, 5.8% changes) and "Locations" (1,237, 6.2%) to allow for reporting the same HCOs at a single level of aggregation (for example, at the NHS trust level and not the trust, hospital, or department levels for different payments). Third, we placed the simplified HCOs from "Names" (208, 1.0% changes) and "Locations" (481, 2.4%) at the same level of aggregation particularly by replacing hospitals from "Names" with their respective trusts from "Locations". Fourth, we generated the final list of HCO names by replacing unclear "Locations" by the "Names" associated with the same payments. Finally, we used postcode information to eliminate any remaining instances of different organizations not being counted separately or the same HCOs appearing under different names.

Website checks
Consistent with the evolving nature of the healthcare system, the names of some of healthcare organizations changed between the time when the payments were made and the time of the web checks. For example, some commissioning or regulatory organizations had merged or been renamed (e.g. two clinical commissioning groups became one clinical commissioning group). Similarly, some NHS trusts merged or ceased to exist.
In these instances, we did not change the organization names reported in Disclosure UK as in some instances this would affect the number of healthcare organizations in the dataset.