Reflections on the WHO’s “Recommendations on Digital Interventions for Health System Strengthening”
A couple of weeks ago the WHO published its public consultation on the Draft Global Strategy on Digital Health (I wrote about my thoughts on it here) and now it is has released a new Guideline: “recommendations on digital interventions for health system strengthening.”
Taking a living guidance approach, the document will be updated overtime (much like the NHS Code of Conduct for data-driven health and care technologies) and aims to help policy-makers and other stakeholders make informed decisions about investments in digital technologies by providing a critical evaluation of the evidence on a number of digital health interventions that are contributing to universal health coverage. Each of the ten digital health interventions are assessed from the perspective of:
And given one of the following recommendations:
- recommended: the intervention or option should be implemented (only mLearning digital health interventions were given this recommendation
- not recommended: the intervention or option should not be implemented
- recommended only in specific contexts or conditions (9 out of 10 of the reviewed digital health interventions were given this recommendation emphasising the fact that implementation has to be context specific)
- recommended only in the context of rigorous research: there are important uncertainties about the intervention or option
As the ten digital health interventions included are very much at the ‘basic’ end of the digital health spectrum and primarily based on simple mobile phone applications, it would be easy for people like me who spend their time ‘horizon scanning’ to be quite dismissive of this guideline. However, the WHO has a wicked challenge on its hands when it comes to providing guidance on digital health implementation and adoption as, whilst there are countries like us (the UK) that are starting to explore the ‘digital frontiers’ of Augmented Reality and Artificial Intelligence, there are also countries for whom SMS exchanges between clinicians and patients would be revolutionary. It is, therefore, understandable that all documents (including this one) it produces on this topic aim towards the lowest common denominator. However, given the Guideline’s very clear focus on critical evaluation of the evidence of effectiveness of each of the digital health interventions reviewed it is still worth reading as a reminder that: (1) getting the basics right is of fundamental importance; and (2) there is still limited evidence that the ‘basics’ work, or what the unintended consequences of their use might be, and this should be borne in mind as we layer on top of these.
It is also worth reading for the emphasis that the Guideline puts on:
- Systems Level Analysis
The Guideline goes to great pains to stress that digital health interventions are not ‘silver bullets’ but instead inherently linked to the wider health system and must, therefore, be seen as complementary to other existing services, rather than as replacements — particularly whilst the evidence of their effectiveness is lacking.
As the general narrative surrounding digital health tends to be excessively hyperbolic, either painting digital health interventions as the panacea of evidence-based personal medicine or as equity-destroying de-humanising devices, it’s somewhat refreshing to see the WHO stress that most of the time digital health interventions will fall somewhere between these two extremes, and may be pushed in either direction depending on the wider context. This not only encourages policy-makers to take an objective and evidence-based approach, but (even though it doesn’t state that it does this) taking a systems-approach highlights how the impacts of individual digital health interventions, themselves seemingly micro in scale, can aggregate to have far wider reaching impacts at a macro-scale. This is something that is often missed in standard impact assessments.
What is still missing from this approach is a take on the aggregate impact on individuals. It is one thing to stress that the system might be fundamentally changed, and some groups of people might be marginalised, by the cumulative impacts of digitisation, and another entirely to provide suggestions on how to monitor the cumulative impacts on individual people. For example, in the UK, there is a great ‘hope’ that digital health interventions will help transform the care of the elderly, helping them to maintain their independence and reduce the burden (cost) on the state. Impact assessments or acceptability analyses will look at each individual component in what is often depicted as an IoT home where the person is constantly monitored for changes in their vital signs so that medical professionals can be alerted if anything changes to assess the impact on the rights of the individual (e.g. personal autonomy) but will very rarely look at the impact on those same rights from an additive perspective. The use of an Alexa to help someone turn the lights on and off might enhance their autonomy (if you’re temporarily ignoring the potential of privacy infringement) on its own, but living in a house with 50+ IoT devices all carrying out similar functions might undermine this entirely.
2. Digital Environment
Continuing in the same vein, the Guideline also stresses that the feasibility of implementation will depend very much on the wider digital environment. This is vital and one of the reasons why we talk of creating an ‘ecosystem’ for the development, deployment and use of digital technologies. If you take one of these elements out, it is likely that the whole digital enterprise will fall over. Indeed all of these are elements in the tech vision published by the Department of Health and Social Care back in September 2018.
It is worth noting too, that this is a fairly ‘build it and they will come approach.’ Or in other words, it’s completely technologically agnostic. Focusing on providing guidance on how to ensure all the foundational pieces are in place so that when a digital health intervention comes along that can deliver on a real need, it can easily be implemented. Indeed, this is one of the reasons why technologies often ‘leapfrog’ in countries (or cities, villages etc. it works on any scale) that at first seem less technologically advanced because: (1) there is less likely to be legacy infrastructure that needs to be dealt with so the technology stack is shorter and less fragile; and (2) from an adoption perspective, the benefits of (e.g.) a push notification compared to a letter seem bigger than the benefits of a push notification compared to an SMS and thus the place that seemed less tech-savvy goes from letters to app-based notifications in one jump, whereas the ‘faster’ adopter takes several more steps to get there. This is one reason why it’s important to take a proportionate approach to regulation, so that such benefits can be realised easily when the risks are relatively low.
Given more space and the opportunity for greater reflection, the Guideline could also have benefitted from further developing this discussion on the environment to include consideration of how the environment also affects the future development of digital health interventions. In the current version, the focus is very much on how it is important to assess whether the environment is ‘ready’ for the implementation of a digital health intervention that is ready now. It does not consider how, each of these elements can become ‘locked’ in such a way that future technologies are increasingly constrained. This is true both from a technological perspective (if data standards are not in place, then interoperability is not possible and then neither is data aggregation big data analytics or AI), and from a social impact perspective (if funding matches the priorities of only one group of people and becomes a perverse incentive in the system, then cumulative disadvantage will result and those that the system does not incentivise development for will become increasingly marginalised). In short the digital environment is also the digital health ecosystem’s infraethics (Floridi 2014) and can either encourage morally good or morally bad outcomes at scale and this should be reflected in every single element of the system design.
3. Acceptability
In addition to reviewing the available evidence for effectiveness of each of the ten digital health interventions included in the Guideline, it also includes a systematic review of available qualitative evidence of factors affecting acceptability. Whilst, it’s very positive to see this having been attempted, it’s not particularly successful. Because the Guideline is (necessarily) generic the factors deemed to positively or negatively influence acceptability are too high level to really be of any particular use in evaluating whether or not the digital health intervention will be acceptable. Indeed, in each case these factors can largely be summarised by the factors included in the technology adoption model: ‘perceived usefulness & ease of use.’
Detailed case studies of ‘successes’ and ‘failures’ rather than the systematic review approach might have worked better, as these are less likely to be affected by publication bias, and more likely to shed light on how factors such as alignment with user need, involvement in design and the ‘emotional’ aspects of what a technology means (imagined affordance), affect acceptability.
Indeed, in an ideal world (and this is very much not the purpose of the document so this shouldn’t be seen as a criticism) the focus would be on social preferability rather than acceptability. Acceptability is important to avoid utter disasters which can negatively impact public trust and set the digital transformation agenda back considerably, but for digital health interventions to really contribute towards universal health coverage the meaning of ‘universal’ also needs to have a qualitative aspect to it. In other words, ‘universal’ health coverage should be universal in terms of its coverage of different cultures and values etc. and this won’t be achieved be only aiming for technologies that are acceptable.
4. Design
Finally, thrown in at the last minute, is a passing comment on the fact that implementations should be guided by design principles. Given that design has such a significant impact on the way that a digital health intervention is used and interpreted by different people, and thus is a significant determining factor in the overall impact the intervention has at a societal level, it feels a bit like a missed opportunity that these principles aren’t given more airtime. This could perhaps be counteracted by taking a case study approach as outlined above, or by linking these design principles to the factors affecting the acceptability. However, if it were me, I would probably include commitment to human-centred design as another key element in the components of the enabling environment to give it it’s proper status.
Overall, this Guideline is a welcome cut through the noise about the potential of digital health interventions to ‘magically’ change everything for the better overnight and to act as a much needed reality check that:
- digital health interventions are additions not replacements
- evidence for effectiveness is sorely lacking and more comprehensive evaluation is needed and fast
- the success or failure of digital health implementation is context dependent
- digital health interventions are not introduced into vacuums but into existing environments that need to be respected.
It will be interesting to see how this continues to develop.
