Enabling outcomes-based reimbursement through a universal platform for outcomes data

How do you prove the lifetime benefits of a potentially life-altering one-off therapy through one or two years’ of clinical trial data? The answer is you do not. The resulting uncertainty around how the therapy will perform in real life is one of the main challenges faced by health technology assessment (HTA) bodies and manufacturers when a new therapy is considered for reimbursement in the national health service (NHS).

Uncertainty is no one’s friend

HTA bodies and NHS payers want to be reasonably confident that the patient benefit seen in the trial will materialise in the real world and in the longer term, and thus that the NHS gets the value it pays for. Traditionally, payers have tended to manage this uncertainty through simple discounts, paired with the potential for treatment discontinuation to limit the budget impact and the potential for overpayment (e.g., if the therapy does not perform as expected). This approach makes sense for managing conventional drugs that are administered repeatedly, but it is less useful for one-off cell and gene therapies. If the conventional model of paying for the treatment in full upfront is applied to one-off therapies, it will not be possible to terminate the treatment to limit the financial exposure in cases where the desired health outcome does not materialise. This creates a challenge both for payers, who face a substantial upfront payment and uncertain outcomes, and for manufacturers who struggle to get their therapies adopted under terms that are commercially viable.

Performance- or outcomes-based reimbursement (OBR) schemes, whereby payments to manufacturers are made conditional on achieving and/or sustaining a certain patient benefit, is a potential way to address this uncertainty. There are several types of outcomes that can be collected and that are meaningful for the purposes of facilitating OBR, however, they can broadly be categorised into three:

Enabling outcomes-based reimbursement through a universal platform for outcomes data

OBR has been discussed extensively in the literature, however, in practice, payers and HTA bodies in Europe have conveyed a low appetite for schemes that require the long-term collection of outcomes data. This reluctance is typically due to concerns around their administrative complexity, and a central component of this administrative burden relates to the collection of outcomes data, which necessitates both having an appropriate data collection infrastructure (e.g. a database) and a (clinical) practice that make OBR possible. In previous work, we have looked at the databases and registries available in the therapy areas that are most likely to see launches of cell and gene therapies in the next few years, and we found that even the most advanced and promising of these can fall short of providing the data and frameworks needed for OBR.

We argue that this is a central issue that needs addressing in order for OBR to become a reality, so that decision uncertainty can be reduced and patient access to therapeutic innovation can be improved. Broadly speaking, there are three approaches that can be taken to implement a new system that can enable OBR:

  1. Updating existing databases and registries individually
  2. Building a new system, independent of existing databases, registries and other data sources
  3. Integrate existing systems and data collected into a new platform

In previous work, we have shown how updating individual databases and registries for the purposes of OBR is possible, and can be done through a manual workaround or through part automation. However, this approach is highly fragmented by nature, and does not lend itself to scaling up OBR into other therapy areas than the ones pertaining to the registry in question. This is an inefficient and costly approach, so in the below, we expand on the pros and cons of building a new system, or integrating existing sources into a new platform.

Building a new and independent system

This approach can be exemplified by the efforts undertaken by the Italian Medicines Agency (AIFA) in 2005 and after, when it introduced a registry for oncology treatments. The purpose of the AIFA registry was to address clinical uncertainty to increase evidence-based decision-making (as well as avoiding off-label use, which was an issue in Italy at the time). When the cancer registries were introduced, there were few patient registries available in Italy, and the AIFA registries were considered somewhat of a revolution. The AIFA registry was used to facilitate OBR for high-cost cancer drugs with high uncertainty, and over time, this was expanded to include also other therapy areas; it currently captures data also for drugs used in Hepatitis C, haemophilia, cystic fibrosis, ADA-SCID, spinal muscular dystrophy, cytomegalovirus.

The registry collects (among others) demographics, clinical data, and information on prescriptions and follow-up appointments. It is designed and developed to be adaptable to accommodate the collection of different patient outcomes as new indications are added, and allows for longitudinal data collection, although the majority of reimbursement schemes range over a one-year period (some ranging up to two years). Another key strength is that the system runs through a web-based portal, which only requires a computer or device with internet access to function. This has enabled its implementation across all the 21 autonomous regions, which have substantial discrepancies among them in terms of technological infrastructure. Also, the operation and maintenance of the registry is largely covered by fees levied on the industry (per product per indication): €32k for the first three years, and €15k for each additional year, which can be considered a reasonable amount in the context of costly specialist therapies.

The introduction of the Italian registry system was no small feat, and is widely referenced in the literature as an example of how to make OBR a reality. However, as the starting point in England in 2019 is quite different from that of Italy in 2005, creating a new standalone system is not necessarily the best way forward. A key concern voiced in our research with key opinion leader physicians and experts in outcomes data collection is that clinical staff are already extremely thinly stretched, and that fully complying even with the existing data collection requirements is not always possible. Clinical tasks will take priority over administrative tasks if staff is forced to choose between the two, and the stakeholders we spoke to found it hard to imagine how the entry of data into an additional system can be facilitated successfully given current staffing and resource limitations.

Given the plethora of existing databases and registries currently in use in England, we believe that it would be counter-productive to introduce yet another framework for frontline staff to comply with, as they are already experiencing capacity issues. Rather, we believe that leveraging the data that is already being collected through existing frameworks and practices is a more fruitful approach.

Integrate existing systems and data into a new platform

All the data that is required to operationalise OBR through clinical and/or economic outcomes are already being collected, which we believe represents an opportunity to be seized upon. While it is not possible to list all the data sources in the NHS here, we detail some examples in the table below.

Enabling outcomes-based reimbursement through a universal platform for outcomes data

As stated above, outcomes data are stored in silos across different databases, registries and repositories, and none of these in isolation provide data sufficient to enable OBR. However, if it were possible to utilise the data that is currently being collected in these various sources in a central platform, OBR could be made a reality. While this is a sensible and seemingly straight forward solution on the face of it, it also represents substantial challenges.

Key challenges to successfully leverage existing systems

We explored some of the key challenges associated with leveraging existing systems, and potential top-level thoughts on ways to overcome them with experts in systems integration and information technology.

Data completeness and quality

Ensuring that the data input into the existing systems is complete and of a high quality is crucial to provide confidence in the results, as this forms the very fundament of OBR. Evidence from the SACT database used in oncology in England (which is widely regarded as a world-leading data collection framework) shows that making data collection mandatory (and potentially also tied to funding) is effective in providing high rates of completion. However, the ‘carrot’ might be equally, if not more effective than the ‘stick’ in achieving and maintaining data completeness and quality: the data collected need to benefit doctors, nurses and frontline staff that collect them and provide learning opportunities, which may do as much to create a favourable data collection practice as potential financial implications.

Data security

A new system that accesses the relevant data where it is, rather than collecting and storing it somewhere new, is associated with a considerable responsibility in terms of the secure handling of sensitive (patient) data. Experts we spoke to hold that the technology to do this is currently available, and end-to-end encryption through blockchain is suggested as a safe and effective way to ensure that the identity of the sender/receiver is secure, and that the transmitted data cannot be tampered with.

Access and governance

Accessing the relevant data entails obtaining permission by the owner of the data, which will be NHS trusts or the registry/database. Ensuring patient confidentiality is a central concern whether the owner of the data is a national registry, or a local trust, and this can have implications for the types of OBR schemes being considered. E.g. a cohort-level scheme, where price will be re-assessed based on the real-world effectiveness in a group of patients, may be less challenging from a governance perspective, as it is less likely that individual patients’ identities can be revealed. On the other hand, a patient-level scheme, where the financial implications are tied to how each individual patient fares on the new therapy, will be more challenging than a cohort-level scheme, especially if the number of patients treated is very low (which will be the case for many cell and gene therapies). A potential approach to overcome this is to only request access to the very data points required for collecting the outcomes in question, and avoiding e.g. demographic data, which would make it easier to identify the patient. Furthermore, we believe that it is important that the data remains in the ownership of NHS and registry stakeholders, and that the data that is utilised supports an analysis that is valuable to these stakeholders, so that everybody has something to gain from the arrangement.

Utilising data stored locally

A central challenge to integrating the data from existing sources is to ensure that the data that is recorded locally in formats that currently require manual retrieval and interpretation, can be accessed in a cost-effective manner. Experts we spoke to point to technologies such as artificial intelligence (AI) as a way to capture data to fill in the existing gaps by pulling e.g. test results, directly from scans, clinical notes and electronic health records (EHR). This avoids creating additional work for clinical staff reduces the risk compromising data completeness, as it does not require additional staff resource.

Sourcing humanistic outcomes

Patient-reported outcomes and experience measures (PROMs and PREMs) and quality of life (QoL) are not a core part of the types of outcomes typically associated with OBR schemes in the past. However, we believe that this is changing, and that over time, such humanistic outcomes will become increasingly important and play a greater role as a measure of treatment success. Given the historically limited application of such outcomes in practice, there is no centralised database or registry for PROMs/PREMs or QoL in England. Furthermore, there are several generic and a great number of disease-specific QoL measures, and so an effort is required to establish which measures are best suited for the purposes of OBR. The generic EQ-5D measure is an obvious candidate for a generic measure, as this is the basis for the QoL utility values used in health economic modelling, whereas it is also useful to capture validated disease-specific QoL measures. Importantly, it will be necessary to consider how technology can aid in collecting these outcomes where possible. There are several companies developing technologies for capturing outcomes directly from the patient (e.g. uMotif, Medopad, Dignio) through applications, and these have the potential to serve as a valuable source of data that requires little or no NHS staff resources.

Conclusion

We believe that the best approach to facilitating OBR in England is by integrating the data that is already collected into different legacy systems into a new platform that can securely utilise existing and additional data sources. This would enable OBR schemes to be implemented without requiring a step-change in the practice of frontline hospital or clinic staff. We believe that compelling such staff to comply with yet another, independent data collection system will very challenging and potentially be a threat to the success of the endeavour all together. Devising a platform that can source the outcomes relevant to the therapy and OBR scheme in question, in a secure way, and that provides confidence in the results to manufacturers, payers/HTA bodies and healthcare staff, it is now a technical possibility that should be explored. Such solutions could come from NHS stakeholders like NHS Digital, or from a trusted third party (one potential example is EY’s Health Outcomes Exchange Platform, which is developed for these purposes).