Value-based healthcare is gaining popularity as an approach to increase sustainability in healthcare. It has its critics, possibly because its roots are in a health system where part of the drive for a hospital to improve outcomes is to increase market share by being the best at what you do. This is not really a solution for improving population health and does not translate well to publicly-funded healthcare systems such as the NHS. However, when we put aside dogma about how we would wish to fund healthcare, value-based healthcare provides us with a very useful set of tools with which to tackle some of the fundamental problems of sustainability in delivering high quality care.
What is value?
Defined by Professor Michael Porter at Harvard Business School, value is defined as a function of outcomes and costs. Therefore to achieve high value we must deliver the best possible outcomes in the most efficient way, outcomes which matter from the perspective of the individual receiving healthcare and not provider process measures or targets. Sir Muir Gray expands on the idea of technical value (outcomes/costs) to specifically describe ‘personal value’ and ‘allocative value’, encouraging us to focus also on shared decision making, individual preferences for care and ensuring that resources are allocated for maximum value.
This article seeks to demonstrate that the role of data and informatics in supporting value-based care goes much further than the collection and remote analysis of big datasets – in fact, the true benefit sits much closer to the interaction between clinician and patient.
Data collection – costing and outcomes
Costing of healthcare for value should be done for the whole patient journey. This is important as it is not possible for value to be created in a service alone – it has to be assessed in terms of the outcomes delivered relative to the investment in all possible interventions for a particular population, whether this is a finite episode of care such as a cataract pathway or the costs attached to a population living with a chronic disease such as Parkinson’s disease. In the latter case, we can take a ‘year of care’ approach to the costing, recognising that there are subpopulations with varying needs within the overall caseload. It is very important to identify these cohorts of patients so that their unmet needs can be characterised and quantified. All too often we adopt a ‘one size fits all’ service for all groups resulting in no-one’s needs being entirely met and creating a really unsatisfactory experience for patients and their carers. Outcome data, including patient-reported outcomes, can inform this.
Improving value for patients through improving outcomes and containing costs can only be achieved through flexible approaches to meeting these needs, by avoiding over- and under- intervention , both of which contribute to poorer outcomes and experience of care.
Two main methods of costing are described and a blend of the two is probably required to gain a full picture.
The first is patient-level costing defined by the HFMA as ‘allocating costs, where possible, to an individual patient. Assigning costs to individual patients provides opportunities for a much greater understanding of how costs are built up. The systems that gather this information are known as patient-level information and costing systems (PLICS).’
The second method is Time-driven activity-based costing (TDABC) again defined by the HFMA as ‘a costing method used by some to improve the accuracy of cost estimates for processes and interventions. It requires organisations to estimate the staff, equipment and time for each step of a process, the total costs associated with the staff involved and the time a patient will spend at each step of the process.’
A blended approach of PLICs , TDABC and cohorting is particularly helpful as, although we often talk about clinical activity in terms of disease pathways, peoples’ experience of healthcare is frequently non-linear and complex. We therefore need to obtain a feel for overall programme spend – this is challenging but essential if we are to allocate resource properly in order to improve outcomes for the long term. This is something we often fail to do given that decision-making about resource allocation is often driven by targets rather than need, and is one of the main reasons why we have failed to increase investment in primary care despite the rhetoric.
What is an outcome? Too often in healthcare what we describe as an outcome simply isn’t at least not from the perspective of a patient. ‘An outcome can be defined as a milestone, endpoint or consequence which matters to a person.’ Complete outcome datasets typically contain four domains:
1. Case mix variables
2. Treatment variables
3. Clinically reported outcomes
4. Patient-reported outcomes
Generally all four domains must be brought together to achieve robust analysis of a big dataset that is properly risk adjusted -the approach and methodology adopted by the International Consortium for Healthcare Outcomes Measurement (ICHOM) is very useful here. However, big data analysis still comes with a health warning and a whole load of assumptions built in, as it does not tell us anything about the individual preferences and goals of those individuals contributing to the data.
Typically, aggregated PROMS data have tended to be used to look at clinical effectiveness but this misses their usefulness in wider applications. A fuller appreciation of these applications is essential to inform the technology needed to support their capture and maximise the opportunities for healthcare improvement. ie to give us all of the user stories. These nuances of application of outcome data fall broadly into two groups:</d
a) Longitudinal tracking of chronic disease
b) Episodic care, particularly surgical pathways
c) Longitudinal tracking of chronic disease
d). prioritisation of issues by the individual
One of the first things to be observed when capturing patient-reported outcomes in this context is the development of the ability for patients to prioritise and rank the most important issues to be addressed in a particular consultation from their perspective. This is a useful aide memoire for the conversation and improves the patient experience, facilitating a two way exchange of knowledge, expectations and goals. This also tends to facilitate the broaching of issues which are more sensitive and difficult to talk about.
supporting shared decision making
Tracking PROM data can also be useful as a more objective assessment of the impact of an intervention such as a new medication.
As data accumulates we will have the ability to utilise ‘real world’ outcome data in reflecting back to patients information better tailored to their own context (rather than that of an idealised population studied in a randomised controlled trial). This contextualisation aids decision making, enabling people to make choices more likely to help then reach their goals and to think about the trade offs between two different courses of action.
support new models of care
As described in the section on costing, we frequently identify cohorts within a population living with a condition. Each cohort has varying levels of need. As we shall explore later in the informatics section, outcome measures (combined with the correct IT functionality) can form an important part of developing new approaches to more flexible models of care eg virtual monitoring.
triggers for key decision points
Longitudinal tracking of outcome data reveals trajectories of disease progression over time and can therefore act as a trigger or prompt for key clinical decisions such as when to discuss anticipatory care planning or intervene to prevent hospital admission.
At a service and programme level, outcome data, aggregated data allows for the identification of population needs. The characterisation and quantification of those unmet needs aids service planning through the identification of the previously mentioned cohorts and, crucially, the allocation of resource. Value for patients cannot usually be delivered by a single service and needs a system-wide approach so the marrying up of costing and outcome data is important here. we can then tailor services more properly to need eg in separating out those with new diagnoses, those who are stable on maintenance therapy, and those with complex and high level needs.
Larger datasets can be further triangulated with costing and process data for benchmarking purposes to inform efficiency and effectiveness, also important for monitoring patient safety.
For a finite episode of care outcome data can also inform shared decision making, particularly important in preference-sensitive clinical scenarios eg where an invasive intervention may be undertaken for symptom control.
PROMS data is already in use to look at clinical variation and quality improvement. Aggregate data of this sort does come with a health warning – even when adequately risk-adjusted it does not tell us about patient preferences or goals. In other words, in data terms, a relatively poor outcome score may actually have been the best thing in the world from the perspective of the individual receiving the treatment…if it helped them meet their limited goals.
Assessing value across a surgical pathway also assumes that this was ‘the right thing to do’. ie that the same outcome could not have been achieved through different management approaches. High quality does not necessarily always equate to high value.
Informatics and IT
It is probably becoming obvious that the role of informatics in supporting value-based healthcare goes far beyond the simple collection of patient-report outcome measures and the subsequent analysis of big datasets (though these are important features).
As ever, much of the success of the informatics comes down to early and intensive consideration of the needs of the users of the technology – patients and their clinicians in the main. What problems are we trying to solve? People need confidence that their data is safe and to understand how it will (and will not) be used. People need to feel that participation in completing an outcome questionnaire sent to them at home is essential to their direct care and not spam communication! The way we develop this two way communication between people and their clinical teams is absolutely critical to the development of new models of care and to enable people more flexible access to their clinicians and their medical records.
On the clinical side, there are now clear recommendations on the standards which should be applied to medical records and the linking of clinical information, as well as standards for information governance. Both of these topics are mentioned because of their importance, but are beyond the scope of this article.
Finally, there is an urgent to develop and expand our analytical capability in healthcare so that we measure, combine, analyse, present and utilise data effectively. Otherwise we run the risk of ‘not being able to see the wood for the trees…’ and let’s not forget that this data does not belong to us . It belongs to our patients and need to be treated with the utmost respect.
National Clinical Director for Value-based Healthcare. GP. Honorary Chair, Swansea Med School. @RslewsiSally