Spread payments have not yet been widely implemented in practice but several authors propose modalities to correct payments for real-world performance. However, ambiguity exists on how payments should be linked to outcomes and several forms of outcome correction have been described either based on individual or population-level data. For spread payments, outcome correction could be based on individual patient data by completely terminating all payments after treatment failure within the individual patient (Edlin et al., 2014; Kleinke and McGee, 2015; Carlson et al., 2017; Jorgensen and Kefalas, 2017; Hettle et al., 2017; Faulkner et al., 2018; Hampson et al., 2018; Oren and Oren, 2018; Schaffer et al., 2018; FoCUS, 2019a; FoCUS, 2019b; Towse and Fenwick, 2019). Outcome correction via individual patient data enables real-time adjustment of payments (Annemans and Pani, 2017; Fox and Watrous, 2017) but could stimulate payers or manufacturers to adversely select high or low risk patients to influence payment amounts (Garrison Jr et al., 2015; Maes et al., 2019). Furthermore, individual patient outcome is dependent on the correct administration of the treatment (Edlin et al., 2014) and individual performance is not useful to verify real-world effectiveness of the treatment for the complete patient population (Launois et al., 2014; Wenzl a Chapman, 2019). Contrarily, payments could be adjusted based on population-level data where the payment for all patients is adjusted post-hoc if aggregated outcomes do not meet a predefined target (Garrison Jr et al., 2013; Van De Vijver et al., 2016; Yeung et al., 2017; Cole et al., 2019; Maes et al., 2019). However, ambiguity remains whether the whole patient population or a sample of patients should be tracked, how missing data should be handled and whether this is feasible for small patient populations (AMCP, 2017; FoCUS, 2019b). Furthermore, several possibilities exist for the adjustment of payment amounts based on population-based outcomes. Payments could be adapted binary with a fixed decrease in payment if patient responses drop below a certain threshold, in a stepped manner where the payment amount can have different levels based on the outcomes achieved or in a continuous manner where the payment is a function of the outcomes measured (Cole et al., 2019; FoCUS, 2019b). Furthermore, an increase in payment amount when the therapy performs better than expected could reward manufacturers and incentivize the development of highly effective therapies (Annemans and Pani, 2017; Hettle et al., 2017; Cole et al., 2019). However, an increase of payment amounts due to better performance and thus a total price increase is currently not possible due to international reference pricing and national legislations in a European context (Annemans and Pani, 2017; Tuffaha and Scuffham, 2018; Cole et al., 2019).

Organising and performing data collection are two of the most frequently discussed barriers for implementation of OBAs for payers, developers and healthcare providers (Sudlow and Counsell, 2003; de Pouvourville, 2006; Carlson et al., 2010; McCabe et al., 2010; Raftery, 2010; Stafinski et al., 2010; Williamson, 2010; Jaroslawski and Toumi, 2011b; Klemp et al., 2011; Neumann et al., 2011; Cascade et al., 2012; Goldenberg and Bachman, 2012; Xoxi et al., 2012; Bibeau et al., 2014; Gibson and Lemmens, 2014; Li et al., 2014; Garattini et al., 2015; Lu et al., 2015; Lucas and Wong, 2015; Mohseninejad et al., 2015; Barlas, 2016b; Carr and Bradshaw, 2016; Malik, 2016; Pouwels et al., 2016; van de Wetering et al., 2017; Duhig et al., 2018; Ernst and Young, 2018a; Ernst and Young, 2018b; EXPH, 2018; Goldenberg et al., 2018; Jorgensen et al., 2018; Stirnadel-Farrant et al., 2018; Urbinati et al., 2018; Federici et al., 2019; Macaulay and Turkstra, 2019; Mundy et al., 2019; Pace et al., 2019; Kannarkat et al., 2020). First, experiences with OBAs in the Netherlands and the United Kingdom highlight that payers should perform a value of information analysis to guide the decision to engage in an OBA to confirm that the benefits from additional evidence collection are higher than the cost of collecting the data (Ferrario and Kanavos, 2013; Pauwels et al., 2017; Makady et al., 2019; Towse and Fenwick, 2019). When the value of information is high, the selection of a fitting study design should ensure the gathering of high-quality data that alleviates remaining uncertainties relevant to the payer (Theunissen et al., 2010; Towse and Garrison Jr, 2010; Vitry and Roughead, 2014; Gerkens et al., 2017). However, uncertainty exists on which evidence is needed, which outcomes should be selected and from which sources the data should be gathered (Ferrario and Kanavos, 2013; AMCP, 2017). To enable the collection of population- and individual-based data, the use of registries as prospective observational study design has been recommended (Theunissen et al., 2010; Towse and Garrison Jr, 2010; Menon et al., 2011; Vitry and Roughead, 2014; Pritchett et al., 2015; Schmetz et al., 2018; Jönsson et al., 2019). An expert panel reviewing the applicability of current HTA practices to the needs of ATMPs and the evaluation of MEAs by the Belgian healthcare knowledge center recommend registries to consist of standardized data elements and cover the continuity of care, including the primary care setting for possibly cured patients, by being linkable to other databases resulting in the provision of accurate, reliable and complete information (Towse and Garrison Jr, 2010; Ferrario and Kanavos, 2013; Cole et al., 2019).

Next to the set-up of registries, several publications report that payers and developers face difficulties to select appropriate outcomes as decision-rule for payment adjustment based on treatment performance (Breckenridge and Walley, 2008; Goldenberg and Bachman, 2012; Kocsis et al., 2015; Abou-El-Enein et al., 2016; Hettle et al., 2017; Holleman et al., 2017; Hanna et al., 2018; Pham and Carlson, 2018; Sandhu and Heidenreich, 2018; Jönsson et al., 2019; Mahendraratnam et al., 2019). Accessible and easily measurable outcomes in the short-term to medium-term which are clinically relevant, useful and important to all stakeholders are recommended (Dankó et al., 2009; Kiernan, 2016; Pouwels et al., 2016; Goble et al., 2017; Kazi et al., 2017; Yeung et al., 2017; Danzon, 2018; EXPH, 2018; Cole et al., 2019). A minimal core outcome set per disease could be built such as survival, disease progression, relapse or recurrence, long-term side effects and return to normal activities for oncological diseases as proposed by Cole et al. (Cole et al., 2019) or disease progression, unacceptable toxicity not allowing continuation of treatment and toxicity-related death as used by the national health service in Italy (Garattini and Casadei, 2011). However, payers are faced with the fact that effects of possibly curative therapies might only appear in the long-term while actionable outcomes are restricted to those measured in the short-term which is similar to the experienced difficulties of the United Kingdom MS scheme to provide conclusive answers on long-term functional outcomes (Lage et al., 2013; Carr and Bradshaw, 2016; Kiernan, 2016; Seeley and Kesselheim, 2017). This would require the use of surrogate endpoints which may give false reassurance of performance (Hettle et al., 2017; Pauwels et al., 2017; Seeley and Kesselheim, 2017; Toumi et al., 2017). Therefore, validated surrogate endpoints that are proven predictors of hard clinical endpoints and uninfluenced by other treatments should be selected (Garber and McClellan, 2007; McCabe et al., 2010; Brennan and Wilson, 2014; Franken et al., 2014; Launois et al., 2014; Carr and Bradshaw, 2016; Gerkens et al., 2017; Toumi et al., 2017; Yeung et al., 2017). However, proven validated surrogate endpoints may not exist for all disease areas which complicates agreement between payer and developer on the chosen outcome. Furthermore, multi-stakeholder agreement on the definition of success, by determining the baseline and target performance, has shown to be difficult but is crucial to define the link between the achieved outcomes and payment (McCabe et al., 2010; Neumann et al., 2011; Towse et al., 2012; Marsden et al., 2017; Jönsson et al., 2019).

Even when the ideal study set-up could be reached and the optimal outcome is selected, data collection is dependent on intensive human and financial resources from the healthcare provider and payer. Administrative and claims databases, as already used in six of twelve countries using OBAs interviewed by OECD (Wenzl and Chapman, 2019), might offer an opportunity for low-burden data collection by measuring hospitalizations and alternative drug interventions as surrogate endpoints for disease progression (Garrison Jr et al., 2015; Yu et al., 2017; Robinson et al., 2018; Schaffer et al., 2018; Makady et al., 2019; Wenzl and Chapman, 2019). However, these data do not provide detailed information on clinical outcomes which might limit their usefulness (Garrison Jr et al., 2015; Seeley and Kesselheim, 2017). Another option to increase data collection capabilities in Europe is enabling cross-country collaboration by coordinating multi-country clinical data collection with interoperable registries to reduce the burden of data collection and avoid duplication of collection efforts (Towse and Garrison Jr, 2010; Ferrario and Kanavos, 2013; Morel et al., 2013; Marsden et al., 2017; Jorgensen and Kefalas, 2017; Kanavos et al., 2017; Bouvy et al., 2018; Alliance for Regenerative Medicine, 2019; Jorgensen et al., 2019; Jorgensen and Kefalas, 2019; Maes et al., 2019). This collaboration could be organised by aligning the evidence needed for follow-up requirements of the European Medicines Agency (EMA) with identical requests for post-reimbursement evidence by several national payers (Ferrario and Kanavos, 2013; Marsden et al., 2017; Gerkens et al., 2017; Jorgensen and Kefalas, 2017; Geldof et al., 2019; Jorgensen et al., 2019; Maes et al., 2019). A first step toward international collaboration has been made by the EMA that will develop the Data Analysis Real World Interrogation Network (DARWIN) to access and analyze healthcare data from across the EU (Hines et al., 2020). Currently, only the Italian national health service possesses a national registry for both regulatory and reimbursement purposes and joint scientific advice will be needed from EMA and the different European health technology assessment bodies to enable collaboration between the regulatory agency and multiple national payers (Bouvy et al., 2018; Jorgensen and Kefalas, 2019).

The collection of detailed clinical data will require the set-up of specific data collection systems or the advanced extraction of data from electronic medical records (Lewis et al., 2015; Goble et al., 2017; Yeung et al., 2017; Wenzl and Chapman, 2019). The presence of a robust data infrastructure requires the establishment of data systems that operate in an automated manner and are virtually connected to integrate all existing information systems within common data formats (Garrison Jr et al., 2015; Gerkens et al., 2017; NEHI, 2017; Goncalves et al., 2018; Maes et al., 2019; Makady et al., 2019). Automation of data collection efforts will require the extensive use of electronic medical records which may be converted into an analysable form (Seeley and Kesselheim, 2017; Robinson et al., 2018; Maes et al., 2019) and the integration and interoperability of data systems that link clinical and financial data (Value in Health, 2017; Antonanzas et al., 2019; Jorgensen et al., 2019; Mahendraratnam et al., 2019). The Italian web-based monitoring registries are currently the most frequently cited as good practice serving as a complete post-marketing surveillance system to share information between health authorities, clinicians, pharmacists and payers with automated procedures and analysis (Carlson et al., 2011; Garrison Jr et al., 2013; De Rosa et al., 2015; Garattini et al., 2015; Montilla et al., 2015; Pauwels et al., 2017; Maskineh and Nasser, 2018; Jorgensen et al., 2019). Other European countries, such as Belgium, Spain and the United Kingdom, highlight the importance of leveraging existing databases to collect both clinical and budgetary data with possible infrastructure upgrades or including new organisational circuits to avoid delays in obtaining data (Clopes et al., 2017; Gerkens et al., 2017; Bouvy et al., 2018; Kefalas et al., 2018; Jorgensen et al., 2019). Jörgensen et al. have calculated the cost estimate for upgrading the current oncological data collection infrastructure via manual workaround or via automation to enable outcome-based reimbursement in the United Kingdom. The analysis showed that in both cases the upfront cost would be mainly caused by the technological upgrade and would be five times higher to allow automation. However, the high annual costs for the manual workaround are substantially lower when the infrastructure is automated and will not be sensitive to the number of contracts (Jorgensen and Kefalas, 2019). Another option is mimicking the Italian system by building a national registry independent of disease area which would lower the cost per MEA set-up but would require a substantial initial investment (Kefalas et al., 2018; Jorgensen and Kefalas, 2019).

Based on experiences with OBAs in Europe, the incomparability of patients, missing data and the presence of confounders pose threats to data quality which may complicate the analysis and interpretation of evidence to support the adjustment of spread payments (Boggild et al., 2009; Adamski et al., 2010; Lage et al., 2013; Van Der Meijden et al., 2013; Vitry and Roughead, 2014; Franken et al., 2014; Garrison Jr et al., 2015; Touchot and Flume, 2015; Carroll and Truglio, 2016; Value in Health, 2017; Kim et al., 2018; Sachs et al., 2018; Tuffaha and Scuffham, 2018; Cole et al., 2019). Furthermore, studies investigating payer opinions showed that low data quality may cause skepticism of payers toward real-world evidence possibly influencing their willingness to use OBAs (Carr et al., 2018; Makady et al., 2019). Additionally, gathering high-quality data is dependent on the input of data from healthcare professionals which was cited as insufficient during CED programs in the Netherlands and during OBAs in Italy due to low compliance with registry procedures (Garattini and Casadei, 2011; Jaroslawski and Toumi, 2011b; Pauwels et al., 2017; Bouvy et al., 2018). Therefore, Makady et al. recommend to invest in the extensive training of healthcare professionals, manufacturers and payers on analysis and interpretation of the results for OBAs (Makady et al., 2019). Furthermore, progress of data collection could be monitored by frequent payer audits to correct for errors during the course of the agreement (Makady et al., 2019; Makady et al., 2019) and to verify if the registry is providing useful information (Mohseninejad et al., 2015; Annemans and Pani, 2017; Gerkens et al., 2017; FoCUS, 2019b).

Lastly, use of data collection systems to collect personal data will require compliance with data privacy regulations such as the General Data Protection Regulation (GDPR) to protect personal data during OBAs (Klemp et al., 2011; Coulton et al., 2012; Franken et al., 2014; Drummond, 2015; Barlas, 2016a; Proach et al., 2016; Van De Vijver et al., 2016; PWC Health Research Institute, 2017; van de Wetering et al., 2017; Duhig et al., 2018; Ernst and Young, 2018a; Ernst and Young, 2018b; EXPH, 2018; Mundy et al., 2019). According to the GDPR, patients may withdraw consent and withdraw their data from the holder files which may increase missing data in registries (Alliance for Regenerative Medicine, 2019; FoCUS, 2019b). Additionally, contractual terms of OBAs may require sharing of identifiable personal data to allow for adjustment of payments. Therefore, additional risk management may be needed to ensure that personal data protection remains guaranteed (Menner and Lewandowska, 2018; Maes et al., 2019) especially for agreements based on individual patient data where the need for identifiable patient data is higher than agreements dependent on aggregated population-level data (Gerkens et al., 2017). In both cases, correct provisions for data protection and sharing are needed with adequate safeguards determining which parties will have access to the data and in which form (Faulkner et al., 2016; Gerkens et al., 2017; Sachs et al., 2018; Jorgensen and Kefalas, 2019; FoCUS, 2019b; Maes et al., 2019).

High-quality governance of OBAs is crucial to support financial flows of the agreement, data collection and reinforcement of the relevant legislation. Therefore, several publications, discussing barriers from payer, developer and provider perspective, highlight the current lack of clear governance structures and recommend to build a framework that details every step of the process with specification of every stakeholders’ roles, responsibilities, interests and incentives (Carlson et al., 2009; Trueman et al., 2010; Ferrario et al., 2011; Jaroslawski and Toumi, 2011a; Jaroslawski and Toumi, 2011b; Neumann et al., 2011; Goldenberg and Bachman, 2012; Xoxi et al., 2012; Towse et al., 2012; Gottlieb and Carino, 2014; Li et al., 2014; Vitry and Roughead, 2014; Drummond, 2015; Lu et al., 2015; Lu et al., 2015; Lucas and Wong, 2015; Lucas and Wong, 2015; Polimeni et al., 2016; Calabrese et al., 2017; Value in Health, 2017; Slocomb et al., 2017; Toumi et al., 2017; Nazareth et al., 2017; NEHI, 2017; Seeley and Kesselheim, 2017; Duhig et al., 2018; Ernst and Young, 2018b; Faulkner et al., 2018; Salzman et al., 2018; Goncalves et al., 2018; Drummond et al., 2019; Mahendraratnam et al., 2019; Moradi et al., 2019; Patel et al., 2019). Furthermore, this framework should entail a clear structure to initiate payments, specify the data collection process with attention to ownership of the data and foreseeing regular data audits, establish a defined management framework, define the funding arrangements of the agreement and clearly state the opportunities for appeal when requirements are not met (Adamski et al., 2010; McCabe et al., 2010; Williamson and Thomson, 2010; Klemp et al., 2011; Ferrario and Kanavos, 2013; Garrison Jr et al., 2013; Annemans and Pani, 2017; Clopes et al., 2017; EXPH, 2018; Wenzl and Chapman, 2019). Most importantly, the governance procedures should indicate the standardised decision-making criteria that are used to support the adjustment of payments based on the outcomes achieved (Coulton et al., 2012; Kornfeld et al., 2013; Gerkens et al., 2017; Kanavos et al., 2017; Makady et al., 2019).

This structure could be reinforced by a core steering committee to establish and oversee the agreement as recommended by the ISPOR good practices for performance-based risk-sharing arrangements task force (Chapman et al., 2003; Garrison Jr et al., 2013). The members of this committee could be the payer, the health technology assessment agency, the manufacturer and the healthcare provider as they are deemed crucial due to their budgetary and/or administrative responsibilities within these agreements (Menon et al., 2011; Ferrario and Kanavos, 2013; Garrison Jr et al., 2013; Drummond, 2015). Additionally, this committee could provide minimum transparency, without disclosing sensitive financial information, on the process of OBA establishment, the governance framework, the progress of data collection and the gathered real-world evidence by annual reports to ensure that all stakeholders groups are held accountable (Hutton et al., 2007; Boggild et al., 2009; McCabe et al., 2010; Raftery, 2010; Cole et al., 2019; Wenzl am]nd Chapman, 2019). Furthermore, studies investigating both experiences with OBAs and stakeholder perspectives highlight the possibility of including an external advisory board with independent experts, having no affiliation toward the payer or manufacturer, such as researchers, statisticians, health economists, IT experts, healthcare professionals and patient representatives to evaluate proposed schemes (while respecting the necessary confidentiality) on their implementation, the data collection process and the analysis of the gathered evidence (Hutton et al., 2007; Adamski et al., 2010; Menon et al., 2011; Thompson et al., 2016; Gerkens et al., 2017; Alliance for Regenerative Medicine, 2019; Makady et al., 2019; Wenzl and Chapman, 2019). The enablement of such a governance structure will require a multi-stakeholder approach guaranteeing good communication and clear entry points into the relevant OBA processes for every stakeholder (Adamski et al., 2010; Klemp et al., 2011; Menon et al., 2011; Thompson et al., 2016; AMCP, 2019; Macaulay and Turkstra, 2019).

However, collaboration between different stakeholders may be complicated by the underlying interests of every stakeholder group and their respective incentives ( de Pouvourville, 2006; Carlson et al., 2009; Puig-Peiró et al., 2011; Neumann et al., 2011; Brennan and Wilson, 2014; Barlas, 2016a; Carr and Bradshaw, 2016; Kiernan, 2016; Proach et al., 2016; Van De Vijver et al., 2016; NEHI, 2017; Value in Health, 2017; Cole et al., 2019; FoCUS, 2019a; Federici et al., 2019; Pace et al., 2019; Kannarkat et al., 2020). Therefore, several authors recommend performing an analysis of the interests of all stakeholders at all stages of the agreement allowing the declaration of possible conflicts of interests and affiliations before the start of the agreement (Chapman et al., 2003; Raftery, 2010; Menon et al., 2011; Clopes et al., 2017; Gerkens et al., 2017; Makady et al., 2019). Although payers, developers and providers have several positive incentives to engage in OBAs, different negative incentives and conflicting interests are reported in literature and may be an additional barrier to OBA implementation. However, several facilitators are proposed to mitigate these negative incentives (Table 2). Next to diverging interests and incentives, a survey with payers and manufacturers showed that a barrier for the collaboration between stakeholders during OBAs is the lack of trust between healthcare professionals, payers and manufacturers (Mahendraratnam et al., 2019). This trust is imperative to ensure that all stakeholders trust that the data was gathered and analyzed in an unbiased manner and will be shared correctly (Garrison Jr et al., 2015; AMCP, 2017). Therefore, stakeholders are encouraged to act in a trustworthy manner by being transparent about their goals and respect the terms dictated in the agreement (Thompson et al., 2016; Mahendraratnam et al., 2019). To enhance trust, several authors propose working with an independent third party such as academic institutions or non-profit, publicly funded organisations to perform all stages of the data collection process (Adamski et al., 2010; McCabe et al., 2010; Stafinski et al., 2010; Menon et al., 2011; Goldenberg and Bachman, 2012; Ferrario and Kanavos, 2013; Drummond, 2015; Thompson et al., 2016; Kanavos et al., 2017; Bouvy et al., 2018; EXPH, 2018; Mahendraratnam et al., 2019; Wenzl and Chapman, 2019). The importance of such independence was shown by the decision made by the scientific advisory group, consisting of representatives with high interest in continued access to the medicines, of the United Kingdom multiple sclerosis (MS) agreement to maintain reimbursement even though many patients experienced worse outcomes (Raftery, 2010; Towse et al., 2012; Gibson and Lemmens, 2014).

Implementation of OBAs with spread payments will be further complicated by the current experienced cost of implementing and managing OBAs with upfront payment. The cost of managing registries for OBA-related data collection has been estimated to be around one million euros in Italy (Garattini et al., 2015). A similar cost estimate was made for the reimbursement of CAR-T cell therapy in the United Kingdom with a MEA, compared with CAR-T cell therapy without a MEA, which would result in an incremental administrative burden of £871.707 over 10 years (Kefalas et al., 2018). However, these costs can be highly variable since the MS risk-sharing scheme in the United Kingdom was estimated to cost nearly £50 million annually (Sudlow and Counsell, 2003; Raftery, 2010) while the costs of the OBA for gefitinib in Catalonia were estimated to be negligible due to the readiness of the organisational structures (Clopes et al., 2017). The majority of costs are attributed to the increase in personnel time and required personnel to perform data collection, the cost of the infrastructure required for data collection, guaranteeing compliance with the established agreement and initiating payments (Sudlow and Counsell, 2003; Raftery, 2010; Loveman et al., 2011; Garrison Jr et al., 2013; Garattini et al., 2015; Garrison Jr et al., 2015; Clopes et al., 2017; Jorgensen and Kefalas, 2017; Macaulay and Hettle, 2017; Kefalas et al., 2018; Mailankody and Bach, 2018; FoCUS, 2019a; Lorente et al., 2019; Mundy et al., 2019). This burden is predominantly experienced by the payer and the healthcare professionals who are responsible for overseeing the agreement and performing data collection. Payers struggle with their administrative capacity to handle (multiple) agreements under current staffing levels (Adamski et al., 2010; FoCUS, 2019b; Jorgensen et al., 2019; van de Wetering et al., 2017). Healthcare professionals, specifically clinicians, nurses, pharmacists and more broadly the hospital administration and finance department, are increasingly burdened with administrative tasks to collect data (Williamson and Thomson, 2010; Franken et al., 2014; Gibson and Lemmens, 2014; Garattini et al., 2015; Kiernan, 2016; Nazareth et al., 2017; Brown et al., 2018; Carr et al., 2018; Schaffer et al., 2018; Antonanzas et al., 2019). They require an increase in staff resources by hiring dedicated healthcare staff or increasing personnel time for management of agreements (Coulton et al., 2012; Goodman et al., 2019; Macaulay and Turkstra, 2019). This was evident from the United Kingdom MS agreement which required hiring approximately 150 additional nurses and an increase in neurology consultations (Sudlow and Counsell, 2003; Garattini and Casadei, 2011). Moreover, Jørgensen et al. emphasize that staff resource shortages are more apparent in primary care which complicates the data collection in the ideal situation where patients are cured after one-time treatment (Jorgensen et al., 2019).

Efficient governance can partly alleviate the administrative burden and cost of OBAs. Therefore, studies investigating the current experience with OBAs and their respective administrative burden emphasize the creation of a streamlined governance process similar for every agreement as being crucial to reduce the cumulative burden of all agreements (Espin et al., 2011; Ferrario et al., 2011; Garrison Jr et al., 2013; Faulkner et al., 2016; Macaulay and Turkstra, 2019; Towse and Fenwick, 2019). Furthermore, lessons learned from the administrative burden experienced in the Belgian healthcare system indicates that utilisation and optimisation of existing administrative systems is essential to ensure a streamlined implementation of OBAs with a reduced burden for healthcare staff (Gerkens et al., 2017). This will require an initial investment in an adequate IT infrastructure and continued investment in the education and support of experienced staff in data collection and OBA management (Menon et al., 2010; Franken et al., 2014; Garrison Jr et al., 2015; Hettle et al., 2017; Schaffer et al., 2018; Antonanzas et al., 2019; Federici et al., 2019; Makady et al., 2019). Adamski et al. stress that efficient governance also includes clear agreements on how the OBAs should be funded (Adamski et al., 2010). The costs for upgrading the data collection infrastructure, the cost of personnel training, the cost of additional personnel (time) and the general management of the scheme could be included in the underlying agreement as transaction costs to the healthcare system (Chapman et al., 2003; Coulton et al., 2012). The manufacturer has been proposed as funder due to the incapacity of government payers to fund the additional efforts to establish MEAs (Edlin et al., 2014; Thompson et al., 2016). Others suggest the payer as responsible for the financing of agreements since they are the ultimate buyers of the technologies that will benefit the entire healthcare system (McCabe et al., 2010; Stafinski et al., 2010; Menon et al., 2011; Edlin et al., 2014; Thompson et al., 2016). Several authors discuss shared responsibility for funding the agreement by both the payer and the manufacturer and possibly research organisations or granting agencies (McCabe et al., 2010; Stafinski et al., 2010; Thompson et al., 2016). Irrespective of the funding party, the governance structure must include who will be responsible for the funding of every task required to enable OBAs (Touchot and Flume, 2015; Carlson et al., 2017; Hettle et al., 2017; Schaffer et al., 2018; Value in Health, 2017).