Are health companies still playing catch up with AI?

A strategic framework for success

When defining a strategy, some approach the task in a truly holistic way and look to implement a fully connected ecosystem of AI across the business and patient network. While an admirable approach, the very fact that they are looking to address all use cases and define a full solution means there’s little chance of securing the required funding as payback.

A more pragmatic approach would be to develop a framework allowing for small steps, implementations and successes to be achieved quickly, unlocking subsequent confidence and investment.

A generalised 5-stage framework plotting maturity could look a little like this:

Auditing

Review business needs, particular use cases, data availability and integrity, feasibility and the types of AI/AI partners and products required to make impact, whether this be driving efficiencies in process or enhancing patient experience and outcome.

Piloting

Active implementation of AI, often through collaboration with consultants/practitioners and on-the-market products that can be tailored to the particular need/use case. Research, testing and iteration throughout.

Scaling

Successes have been achieved and socialised within the organisation. AI is becoming more integral to the business and the business’ future strategy as opposed to pockets within divisions. The internal talent pool is growing and businesses consider developing their own AI IP as well as continuing partnership programmes.

Connecting

The business is able to build on successes across individual service and product areas and provide a more holistic, enterprise AI ecosystem with more sophisticated layers of machine learning helping to feed each other.

Mainstreaming

The AI strategy cannot be separated from the business strategy. AI has become standard business-as-usual and has organisation-wide acceptance, embedded in the culture of the business. Continual cycle of innovation against arising use cases and business and patient needs.

Practical applications and considerations

There are a handful of emerging trends that have already been adopted with certain sectors validating their investments.

The rise of sensors and companionship

This is especially prevalent in situations when supporting patients with mental health conditions, the elderly and those with terminal illness.

SwissRe recently spoke about their shift from claims management to safety management and the application of home sensors that can alert family members and carers to certain behaviours so supportive action can be taken when this occurs i.e. frequent opening of a fridge door indicate growing anxiety

The introduction of robot carers, like Care Angel, is a more sophisticated extension of the above. As well as being successfully piloted as an additional set of eyes and ears for the patient, they have also been used to nudge behaviour changes and medication adherence.

In addition, voice companions can also provide a level of companionship and stimulus to keep patient’s brains active and engaged and, most importantly, to help ward off loneliness - which is arguably more harmful than obesity or smoking and increases the risk of dementia, heart disease and depression.

Symbiotic AI

Rather than the much-feared media hype regarding AI taking our jobs, in health, much like other categories, the reality is proving to be the creation of a more symbiotic relationship between human and machine.

Examples of this include a range of initiatives from Babylon and Bupa-trialing patients to relieve pressure on GPs; to Novartis’s collaboration with the University of Oxford to accurately crunch disparate data and spot disease insights and build predictive models to help support drug development.

Behaviour change and personalised programmatic

Machine learning is also playing a significant role in supporting patient recovery and preventative actions through a series of smart behavioural nudges.

The smokebeat product combines wearables and sensors to differentiate between smoking and other hand to mouth movements and uses predictive modelling to send timely incentives at expected points of weakness.

Woebot is another example of machine learning, tracking behaviour, this time using language. It analyses key words used by people suffering from mental health issues and then responds with suggestions and cogitative support to help the user before further professional help can be sought.

Ethics and patient-centred design

With the pace of technology advancement once again outrunning the government and trade bodies alike, the importance of ethics is paramount. Especially in health.

It’s increasingly important businesses recognise that just because they can, they shouldn’t. The use of AI should be intrinsically linked to corporate values and principles and its use should be transparent to internal and external audiences.

Businesses should also be open with their AI ethics policy and have clearly defined red lines in place. Without this, patient and potentially clinical staff, will be eroded and we will return to the anti-trust headlines some media publications find it so easy to write.

Conclusions

Outside of the above, the role of AI and machine learning is also accelerating across electronic health records, assisted robotic surgery, image processing and disease detection, digital therapeutics and clinic trials to name a few.

So, it’s fair to say, that while health wasn’t the quickest of the blocks with regards to AI adoption, it’s certainly making up for lost time. And as Derek Lowe, drug discovery researcher, recently stated in the New York Times:

“It is not that machines that are going to replace chemists. It’s the chemists who use machines that will replace those that don’t.”

Paul Vallois is the managing director at Nimbletank.