The corporate giants Apple and Google have their fingers in many pies, so their present interest in healthcare should probably not surprise us. Like many other tech firms, they aim to bring medical services on to a digital platform. These services hope to use artificial intelligence to develop digital therapeutics (‘digiceuticals’) and new diagnostic techniques. Though exciting, these developments must be treated with caution to ensure their safety and medical accuracy.

The move to digitize healthcare comes as doctors struggle to meet demand. While the European Union of General Practitioners (UEMO) recommends general practitioners (GPs) to have up to 25 daily consultations, a survey of 900 GPs in 2017 reported an average of 41.5 and a maximum of 70 daily consultations. Not only does this create an unhealthy working environment for doctors, it also increases the risk to patients by exacerbating the potential for human error. Lightening the workload without compromising patient care is obviously critical.

DeepMind, a branch of Alphabet (the parent company of Google) is seeking a somewhat unorthodox solution: it has provided an app called ‘Streams’ to four UK hospitals, which monitors patient data and sends an instant alert to a clinician if acute kidney injury (an indicator of patient deterioration) is detected. It also ranks patient information from most relevant to least relevant so that nurses can quickly develop a picture of the patient’s state. By saving nurses and clinicians up to two hours of work, the app improves prognosis and gives professionals more time to see their patients, or take a much needed break. However, the app has been met with some criticism. The initial agreement between DeepMind and the Royal Free London NHS Foundation Trust did not adequately protect patient data and had to be severely revised, and the storage platform used by DeepMind to house the 1.6 million patient data records was found to have 11 minor vulnerabilities. At the heart of these new technologies is the struggle to strike a balance between keeping patient data confidential while easily accessing as much data as possible to increase the predictive capacity of algorithms.

Apple, in collaboration with the Stanford University School of Medicine, launched the Apple Health Study in November 2017. Here, an individual’s heart rate and rhythm are measured via a sensor in their Apple Watch, and monitored for irregularities or arrhythmias. If any are found, a notification alerts the participant and they are offered a free video consultation with one of the clinicians taking part in the study. If required, a free ECG test may be sent to the individual to further track their health. By frequently monitoring individuals and giving them easy access to a healthcare professional, the study may increase the chances of discovering an incipient arrhythmia such as atrial fibrillation, one of the main risk factors for stroke.

Apple is also applying this idea to complex, chronic diseases such as Parkinson’s disease (PD). The symptoms of PD vary between individuals, and without frequent specialist appointments its progression can be difficult to track. As such, the app "mPower", developed by Sage BioNetworks in collaboration with Apple’s ResearchKit, requires users to complete three tasks or surveys a day. The creators hope that by providing a closer analysis of symptom progression and its frequency and variation, a treatment to control this variation can be manufactured.

For more general diagnosis, the digital health provider Babylon uses an AI chatbot to deliver medical advice to users via text messages, and may recommend an appointment with a GP either via a video call or a face to face appointment. The AI chatbot was created by inputting millions of data points from doctors and scientists to best match symptoms and background information to possible conditions. The underlying algorithm analyses the contents of the client’s messages, matches them to medical terminology, and uses the latest medical literature to predict new relationships between diagnoses and treatments, and symptoms and outcomes. From this it can draw a reliable link between a symptom and outcome, despite the billions of possible combinations of symptoms, risk factors and diseases.

However, the accuracy of this algorithm has not been independently supported by the Royal College of General Practitioners (RCGP). Babylon has claimed its chatbot has equal accuracy to its human counterparts, scoring 81% on the final exam required for membership to the RCGP. The RCGP disputes this, because they did not provide the questions used. They additionally pointed out that while the service might be able to give an accurate diagnosis during a test, in clinical settings a range of factors such as emotional impact must be considered before delivering diagnoses – which currently only a human physician can accomplish.

The interest in digitalising therapeutics has led to the production of ‘digiceuticals.’ Akili Interactive has developed a video game which presents individuals suffering from ADHD with stimuli that activate areas of the brain exhibiting deficiencies, using adaptive algorithms to provide personalised care to improve attention and memory.

Products targeting symptoms are also in development, such as self-stabilising cutlery for those with PD and exosuits to assist mobility. Whether clients are willing to trust a video game or an app over a pill prescribed by their doctor, however, remains to be seen. Until some solid regulations are in place, it is likely that the public will remain skeptical towards such treatments.

These developments are exciting and more than a little daunting in a society already concerned with the amount of time spent on digital devices. Their benefit to health may be met with suspicion both from the public and professional domain. That said, technology’s potential to lower the risk of medical error and provide easily accessible healthcare is huge.
The digitalisation of medicine is underway, but whether it can be successfully integrated into society remains to be seen.