The pace of artificial intelligence technology adoption in healthcare varies considerably. Some medical establishments are undertaking small incremental changes; others centers have seen several years of innovation; and a proportion remain tied to the traditional healthcare model of the 1990s. This is the view of Dr. Ameet Bakhai, deputy director of research at the Royal Free London NHS Foundation Trust.
Dr. Bakhai was expressing his views in advance of a major conference that is set to look at artificial intelligence in healthcare: Digital Healthcare Transformation Summit 2017, which takes place in London in December. A key theme is that although there are more advanced machines, from ultra-high-resolution imaging instruments to surgical robots, these tend to remain fully controlled by humans rather than with decisions made by artificial intelligence. AI tends to complementary, helping with results interpretation.
Even with a variable rate of adoption, a study by Frost & Sullivan shows the market for artificial intelligence in healthcare to be projected to reach $6.6 billion by 2021, a 40 percent growth rate based on 2016 figures. One outcome of this, plus the development of healthcare technology in general, is the production of digital data.
One of the reasons for the slow pace of development of medical data appears to be partly due to a series of imperfectly answered medical and ethical questions. These are set out by law firm DLA Piper LLP. These include questions such as ‘who owns any intellectual property rights in the data and any expertise that is provided?’ With this question, an important related issue is whether the medical data collected from a patient is owned by the healthcare institution, pharmaceutical or technology company, or by the individual patient?
A further issue with data sharing, which is also a concern with many wearable medical devices that collect data, is that data from advanced hospital scanners to wearable fitness monitors can be sent to medics or pharmaceutical companies. To what extent does the patient have a say over how their data is used? Moreover, to what extent is data appropriately anonymized or pseudonymised? This latter point also relates to when data is used as part of clinical trials.
A second issue concerns what are the data protection and privacy implications? This is question for technology developers to answer in terms of how data security will be ensured. Related to this is deciding on the types of information to be shared with patients.
The amount of data generated continues to grow rapidly. According to PwC, between 2010 and 2015, the amount of stored patient data increased 700 percent. Of this, 91 percent of the data is unstructured. Unstructured data refers to the information that resides outside of organized databases such as electronic health records and laboratory reports.
A third factor comes down to the power relationship between the medical expert and patient. How does data driven medicines alter the relative bargaining power of the two parties. What happens as patients have access to more data about their condition? A related issue is the power relationship between the technology provider and the healthcare instiution. With a wider use of bespoke and preparatory technology, does a greater balance shift towards the providers of equipment rather than the medics?
The trend towards artificial intelligence in healthcare is welcomed by many working in the medical field. However, the development of technology needs to be coterminous with public policy in relation to data ownership, privacy and security.
