appeal to many people who are already acclimated to the use of apps to fulfill personal needs
(e.g., Amazon, Uber, etc.). However, the potential dangers of sharing personal health
information over such networked connections is a concern.
3.1.3 Examples of privacy and transparency
The rapid evolution and adoption of AI applications in remote access health care has a strong
presence outside of the US. One such example is deep learning from the company DeepMind
Technologies [66]. In 2016, DeepMind launched several initiatives in the health care arena
under its DeepMind Health Division [67].
DeepMind is working with UK National Health System (NHS) Hospital Trust Foundation [68] to
develop AI applications involving patient electronic health records from multiple London
hospitals. It is indicative of the fast moving pace of these AI in health care applications that the
most informative information on the plans of DeepMind Health comes from investigative news
articles written within the last few months of this study [69]. These have focused on issues of
transparency, privacy, and health ethics issues relating to delivery of AI products in health care
[70]. NHS patient data was provided without informed consent to Google to test an app
designed to help monitor kidney disease. Some argue that this violated UK regulations that state
“patient records without explicit consent can only be used for direct care delivery.” However,
this controversy has not slowed DeepMind’s continued and expanding partnership with
NHS [71].
From a US government perspective, the work of DeepMind Health using UK NHS data should
be viewed as an early large scale “experiment” that will reveal many real-world issues that arise
in the application of AI to health care. It should therefore be tracked closely. For example, the
problem discussed above with data access transparency may have led DeepMind to an
accelerated application of “blockchain-style” technology for securing and tracking data
access [72]. Basically, blockchain methodologies use a distributed database consisting of
continuously updated (augmented) “blocks” which contain a linked list of all previous
transactions [73]. In the case of health care, this encompasses all previous records of access to
an individual data record including information about how the data was used and any additions
or changes to the data record [74,75].
A second technology application that has emerged from DeepMind Health has many blockchain-
like aspects [76,77]. Instead of blockchain, the DeepMind data audit system uses an approach
based on Merkle Trees [78], a type of hash tree that allows secure verification of the contents of
large data structures. DeepMind hopes to prototype the verifiable data audit system by the end of
2017 for eventual use in its Royal Hospital health care software environment [79]. While the
audit system will be prototyped within the confines of the NHS organizations, DeepMind sees
the possibility that the audit system itself might become available to individuals in the public so
that they can access and verify their actual electronic health records.
Given the rapid growth of similar activities in other countries, the U.S. government should track
developments in foreign health care systems, looking for useful technologies and also technology
failures.
24
