Content uploaded by Michael Rowe
Author content
All content in this area was uploaded by Michael Rowe on Aug 25, 2019
Content may be subject to copyright.
Content uploaded by Michael Rowe
Author content
All content in this area was uploaded by Michael Rowe on Aug 07, 2018
Content may be subject to copyright.
Opinion OpenPhysio
Artificial intelligence in clinical practice: Implications for
physiotherapy education
Michael Rowe
1
1. Department of Physiotherapy, University of the Western Cape, Cape Town, South Africa
Submitted: 28 May, 2018 | Accepted: 23 August, 2019
Abstract
About 200 years ago the invention of the steam engine triggered a wave of unprecedented development and growth in
human social and economic systems, whereby human labour was either augmented or completely supplanted by machines.
The recent emergence of artificially intelligent machines has seen human cognitive capacity enhanced by computational
agents that are able to recognise previously hidden patterns within massive data sets. The characteristics of this
technological advance are already influencing all aspects of society, creating the conditions for disruption to our social,
economic, education, health, legal and moral systems, and which may have a more significant impact on human progress
than did the steam engine. As this emerging technology becomes increasingly embedded within devices and systems, the
fundamental nature of clinical practice will evolve, resulting in a healthcare system that may require concomitant changes to
health professions education. Clinicians in the near future will find themselves working with information networks on a scale
well beyond the capacity of human beings to grasp, thereby necessitating the use of artificial intelligence (AI) to analyse and
interpret the complex interactions of data, patients and the newly-constituted care teams that will emerge. This paper
describes some of the possible influences of AI-based technologies on physiotherapy practice, and the subsequent ways in
which physiotherapy education will need to change in order to graduate professionals who are fit for practice in a
21st-century health system.
Keywords: artificial intelligence, health professions education, physiotherapy education
Introduction
For thousands of years the trajectory of human progress showed
only a gradual increase over time with very little change in the
material aspects of ordinary people's lives over the course of
successive generations (Morris, 2011). But about 200 years ago
there was a sharp increase in human social and economic
development, which saw the curve of population growth and
productivity veer sharply in an almost vertical direction (ibid.).
Figure 1: World population and social development index
(Morris, 2011). Click on image to enlarge.
The invention of the steam engine started the Industrial
Revolution and brought with it a range of technological
advancements that enabled human beings to generate
enormous quantities of energy, removing the biological
bottleneck that limited the amount of work that could be
performed by human and animal labour. The introduction of
machine power significantly increased our capacity to shape our
physical environment and created the conditions for the mass
production of material goods, improving the quality of life
across all levels of society (Brynjolfsson & McAfee, 2014). As the
steam engine enhanced our ability to do physical work so the
rise of computing machines will create the conditions for
cognitive enhancement that will enhance our capacity for
intellectual work. We are currently at an inflection point beyond
which we will see human cognition first augmented, and then
surpassed, by artificially created intelligence, leading to a
transformation of our social and economic conditions that will
be no less profound than those that were introduced by the
steam engine (ibid.).
Artificial intelligence
To understand more clearly what is meant by artificial
intelligence (AI), it may be useful to consider what we mean by
intelligence and by extension, what we do not mean.
Corresponding author:
Michael Rowe (email | Google Scholar)
Department of Physiotherapy
University of the Western Cape
Cape Town, South Africa
OpenPhysio
Rowe, M. (2019). Artificial intelligence in clinical practice: Implications for physiotherapy education.
Intelligence is a general mental ability that involves reasoning,
planning, problem solving, abstract thinking, comprehending
complex ideas, and learning from experience. It also reflects a
broad and deep capacity for understanding our environment
(Gottfredson, 1997). This definition of intelligence says nothing
about self-awareness, consciousness, emotion or morality, nor
that there is anything essentially human about intelligent
behaviour. In addition, there is nothing in this definition of
intelligence suggesting that non-human intelligent behaviour
should replicate human thinking processes. Indeed, "there are
lots of ways of being smart that aren't smart like us" (Winston,
1992). Today, AI is a cross-disciplinary field of research that
attempts to understand, model and replicate intelligence and
cognitive processes by invoking computational, mathematical,
logical, mechanical and biological principles and devices
(Frankish & Ramsey, 2017). More specifically, it is the study of
the synthesis and analysis of computational agents that exhibit
intelligent behaviour (Poole & Mackworth, 2010). In the field of
AI research, an intelligent agent is something that acts in an
environment where those actions are appropriate for its
circumstances and its goals. In other words, the agent is able to
adapt to changing environments and changing goals, learn from
experience, and make appropriate choices given its perceptual
and computational limitations (ibid.).
The modern field of artificial intelligence
research was born in
1956 at the Dartmouth Workshop where the term was coined
and was initially conceived as a project that might take a few
decades to conclude (Frankish & Ramsay, 2014). However, until
recently progress in AI research has not lived up to the
enthusiastic predictions of its initial supporters. This was
particularly true with respect to the claims made by pioneers in
AI research around the development of artificial general
intelligence. For example, “...machines will be capable, within
twenty years, of doing any work that a man can do” (Simon,
1965) and “In...3 to 8 years, we will have a machine with the
general intelligence of an average human being” (Minsky, 1970),
have turned out to be completely wrong. While we are still not
much closer to a generally intelligent machine, there are three
main characteristics of modern AI research that demonstrate
how and why the development of AI-based systems has seen
significant growth over the past ten years (Susskind & Susskind,
2015).
The first is the ubiquity of cheap computation. The second is the
availability of massive data sets that are necessary for machine
learning algorithms to be trained on. The third is the
development of improved programming techniques and more
sophisticated algorithms (Brynjolfsson & McAfee, 2014). The
combination of these three characteristics have led to an
increase in the ability of AI-based algorithms to reach
conclusions in uncertain contexts that, in many ways,
outperform what is possible by human beings. In addition, it is
not only the computational power of machines that is
noteworthy but the increased power and sensitivity of the
associated hardware that makes the AI of today qualitatively
different than what was possible in the past. Better and cheaper
sensors, fast networks, smaller gyroscopes, and the global
positioning system (GPS) enable AI-based systems to perceive
their physical environment, locate themselves on earth, connect
to and establish relationships with other connected devices, and
understand their own position in space. AI-based technologies
are not only smarter than they were even five years ago but are
embodied and able to move and learn in the real world.
Intelligent machines are therefore no longer constrained to
solving theoretical problems in academic research labs but are
doing real work in the real world.
Today, AI-based research has led to the use of expert systems
that guide clinical decision-making, computer vision algorithms
that outperform human beings in the analysis of CT and MRI
scans, better diagnostics and prediction of patient outcomes,
and enhanced administration and planning in health systems
(Harwich & Laycock, 2018). In addition, AI research is
achieving important advances in the areas of information
retrieval and retention, problem-solving and reasoning, image
recognition, planning, and physical manipulation (Frankish &
Ramsey, 2017). It is worth noting that, broadly speaking, these
are also core aspects of physiotherapy practice. It seems
reasonable to suggest therefore, that much of physiotherapy
practice may become increasingly vulnerable to automation by
AI-based systems. The aim of this paper is to explore the
potential impact of AI-based technologies in clinical practice
and the subsequent ways in which physiotherapy education
might change in order to graduate professionals who are fit for
practice in a 21st century health system.
Artificial intelligence in clinical practice
Health and education systems are increasingly recognised as
complex adaptive systems, characterised by high levels of
uncertainty and constant change as a result of rich, non-linear
interactions that cannot all be tracked (Fraser & Greenhalgh,
2001); Bleakley, 2010). This means that complex systems are
inherently ambiguous and uncertain and that they lack
predictable outcomes or clear boundaries. Over time these
systems become more complex, making it difficult - or
impossible - for individuals or even single disciplines to work
effectively within them (Frenk, et al., 2010). The emergence of
AI-based technologies will play an essential role in augmenting
our cognitive capacities so that we are able to function
effectively in increasingly complex systems. However, there are
some important ways in which this will require future health
professionals to change some of our more fundamental concepts
around clinical practice (Wartman & Combs, 2017).
Firstly, patient management will be based on the interpretation
of massive aggregates of data that are collected from multiple
sources and applications, many of which will be
patient-controlled. We can already see this with wearable - and
soon, ingestible - technology that moves with patients providing
a continuous flow of information on a wide variety of
2 of 6
OpenPhysio
Rowe, M. (2019). Artificial intelligence in clinical practice: Implications for physiotherapy education.
health-related metrics at a scale that health professionals have
never had to work with before (Susskind & Susskind, 2015). The
exponential growth of these interconnected medical devices
continuously generate a volume of data that simply cannot be
analysed, interpreted, or even understood by human beings
alone as they exceed the capacity of our cognitive ability
(Obermeyer & Lee, 2017; Wartman & Combs, 2017; Claiborne,
2018). Clinicians will therefore need to incorporate
AI-enhanced analysis of large patient data sets into their clinical
decision making, where the data resides in multiple locations
and across a variety of service providers, all interacting in ways
that are too complex for individuals to track, or for single
disciplines to manage.
Following from this, a second characteristic of 21st century
health systems is that care will be provided by newly-constituted
care teams where patients interact with providers across a wide
range of professions, some of which are themselves only
beginning to emerge. For example, patients may work directly
with data scientists who will help them analyse and interpret the
enormous amounts of personal and health-related data that
they gather themselves (Topol, 2015). Thus, the role of the
patient will change as they become true participants in clinical
decision-making, possibly even becoming team-leaders as they
delegate responsibility for various aspects of their care to
different providers, as and when they
deem it is necessary.
Thus, the nature of the professional role may need to be
re-evaluated as specialised knowledge becomes increasingly
available outside of the disciplines. This will lead to
decision-making that is distributed across different providers
and services, some human and some machine, with the patient
in control (Topol, 2015; Susskind & Susskind, 2017).
Finally, these interactions will take place in both physical and
digital environments and will therefore be distributed in both
time and space. As clinical encounters move into digital and
online spaces, the vast number of interactions between patients,
their data, and distributed care teams will need to be mediated
by AI-based systems that manage and plan interactions based
on the patient's self-identified goals. Professional scopes of
practice may need to be revised as certain cognitive and physical
tasks, previously the explicit domain of discipline-based
professionals, become be distributed between diverse care
providers with differing levels of expertise. In these new health
care teams clinicians will need to learn how to fluidly pass
control of patient care between themselves, other team
members (which may include data scientists and software
developers, for example) and smart machines (Wartman &
Combs, 2017). Success will increasingly be premised on a teams'
ability to recognise the collective intelligence of the system or
network, rather than privileging the expertise of an individual or
single profession.
It seems clear that 21st century healthcare will be characterised
by a move from producing information, to analysing and
interpreting it on a massive scale and across distributed
platforms. Clinical practice will therefore be enacted in
data-rich systems where information flows will include high
volumes of data that are generated from multiple sources of
differing quality and validity (Wartman & Combs, 2017). It is in
this context that the clinician will need to learn how to position
themselves as they work alongside intelligent machines, not
only because AI will augment our physical and cognitive
abilities but because it will soon be impossible to function at all
without them. As machines gain preeminence in the retention,
access, and analysis of information it has never been more
important for clinicians to recognise the caring aspects of the
profession (Claiborne, 2018). Future clinicians will need to
focus their attention and efforts on a more patient-centered,
higher quality of care (Panagiotis, 2017) and discard the
paternalism that continues to characterise clinical practice
today. We will need to come to terms with the fact that 21st
century health systems will be patient-driven and that care
teams will be loosely connected, cross-disciplinary, and will
include smart machines imbued with artificial intelligence.
Preparing graduates for clinical practice in the 21st century will
not come through iterative changes to our current clinical and
educational paradigms but rather through fundamental reforms
to how we think about professional practice (Susskind &
Susskind, 2017). We would do well to think carefully about
whether physiotherapy is an art practice or a technical craft
remembering that, for much of our professional history, it has
been viewed as the former (Nicholls, 2018). But if the nature of
future practice is such that the technical components of the
discipline are outsourced to intelligent machines, we may find
ourselves in the position of being well-trained, competent, and
irrelevant. We should ask how to adapt physiotherapy education
so as to deepen and strengthen the human-based components
that are difficult for AI-based systems to replicate, as well as to
integrate the technological and data literacies that are needed in
order to understand and work with smart machines.
Physiotherapy education in the intelligence
age
Education has always been one way for human beings to adapt
to social and economic disruption; this is how we upgrade
ourselves. But as machines get smarter - and the pace of change
accelerates - the relative value of a professional degree is
reduced (Susskind & Susskind, 2015). Accordingly, the ability to
access professional education continuously throughout our
working lives will become increasingly important. We should
stop thinking of physiotherapy education as time-limited degree
programme that people graduate from once in their lives and
reconsider it as a platform for lifelong learning that provides
learners with customisable modules they can access when they
need to (Aoun, 2016). This will be critical as more people return
to higher education during their careers, driven by the need to
stay ahead of technological change. In addition, it will not be
enough to focus only on training future healthcare
professionals, as the pace of AI implementation in health
3 of 6
OpenPhysio
Rowe, M. (2019). Artificial intelligence in clinical practice: Implications for physiotherapy education.
systems will rapidly outpace the retirement of practising
clinicians. In other words, we face a situation in which current
physiotherapists may find themselves unable to practice
effectively as they struggle to communicate and collaborate with
AI-based systems. Neglecting professional development may
have been less serious in times of incremental change but in an
era of profound transformation that requires the retraining of
many thousands of clinicians, it will be problematic (Hodges,
2018). It is therefore essential that we develop improved
systems for continuing professional development that not only
aim to provide current practitioners with the knowledge and
skills necessary for clinical practice in an intelligence age but to
cultivate a mindset of flexibility that enables them to adapt to a
rapidly changing workplace (Susskind & Susskind, 2015). As a
side note it is by no means self-evident that this curriculum will
be housed within an accredited higher education institution. In
fact, as specialised knowledge is democratised and available
outside discipline-specific programmes it seems more likely that
continuing education will be provided elsewhere (ibid.).
This is likely to present a significant problem for physiotherapy
training. It should not be controversial to suggest that the
profession is traditionally conservative and that undergraduate
education continues to privilege the accumulation of knowledge
and skills rather than the agency that students need to navigate
uncertain and changing futures (Barradell, 2017). The current,
largely memorisation-based programme must transition to one
that integrates three fundamental literacies across the core
physiotherapy curriculum; data literacy, technological literacy,
and human literacy (Aoun, 2017). As clinicians become single
nodes (and possibly not even the most important nodes) within
information networks, they will need data literacy to read,
analyse, interpret and make use of vast data sets. As they find
themselves having to work more collaboratively with AI-based
systems, they will need the technological literacy that enables
them to understand the vocabulary of computer science and
engineering that enables them to communicate with machines.
Failing that, we may find that clinicians will simply be
messengers and technicians carrying out the instructions
provided by algorithms. Finally, as machines take over the
computation and reasoning tasks previously performed by
human beings clinicians will need the human literacy that is
beyond the reach of machine learning algorithms, helping them
to develop skills in empathy, teamwork, creativity, design,
ethics, and entrepreneurship (ibid.).
The integration of these literacies within and across the core
curriculum will help future therapists develop the creative
mindset and mental flexibility to invent, discover and produce
the original ideas that are necessary to supplement the
computation and reasoning abilities of artificial intelligence
(Aoun, 2017). A curriculum that is integrated in this way will
help future clinicians learn to collaborate with high-performing
algorithms while at the same time accentuating our uniquely
human strengths. In contrast, a curriculum that is driven by
content is oriented to the present and not the future, leaving
little room for imagination, professional artistry and capability
(Barradell, 2017). We will therefore need to reevaluate certain
taken-for-granted assumptions about what we consider to be
core to the curriculum. Will we continue teaching students to
interpret X-rays when algorithms are already better at image
recognition? Will students need to know about disease
progression when expert systems have the sum total of all
human knowledge embedded within them? Should students
learn research skills when search, filtering, aggregation and
synthesis algorithms do a better job of summarising a body of
work? Or, will students need to know how to interpret
algorithmic decisions and more importantly, know when to
ignore them? These are only a few examples of the kinds of
questions that physiotherapy educators might consider as we
move into AI-supported health systems.
It may be necessary for professional programmes to integrate
data science, deep learning, and behavioral science into their
undergraduate curricula in order that health professionals are
able to develop, evaluate, and apply algorithms in clinical
practice (Obermeyer & Lee, 2017; Hodges, 2018). Without this
integration we may find that we have generations of health
professionals and educators who are unable to speak the
language of 21st century healthcare. In addition, we should
consider moving away from the simple transfer of knowledge
and skills that characterise many curricula and instead aim to
build students’ capacity for creative problem-solving and
ideation. In short, we will need to develop creativity capacity
and personal human connection instead of routine cognitive
skills. Physiotherapy educators must strive to provide the things
that smart machines cannot; depth of disciplinary expertise and
practice wisdom, personal learning pathways, and an emotional
connection to students as part of a relationship-centred
approach to teaching and learning. While AI-based systems may
ultimately take over the mundane tasks of managing
the
learning process, educators will still need to help students
identify meaningful goals, address the emotional aspects of
learning, and develop closer relationships with students in order
to better support and motivate them.
Conclusion
Artificially intelligent systems are driving changes in the health
system that will have profound effects on how health care is
enacted in the 21st century. These disruptive changes will force
all healthcare professions to reevaluate how fit for purpose they
are in an intelligence age that is characterised by smart
machines, massive data sets of vast complexity, and
fundamentally different relationships with patients and
algorithms. We should probably take seriously the notion that
the health professions of yesterday can - and maybe should - be
at least partially replaced by more appropriate alternatives,
including AI-based systems and cheaper alternatives. Intelligent
algorithms are already smarter than us within certain narrow
domains and successful clinical practice in the future will
require that we understand how to interpret the advice of
4 of 6
OpenPhysio
Rowe, M. (2019). Artificial intelligence in clinical practice: Implications for physiotherapy education.
machines, when to hand over control to them, and when to
ignore them.
The physiotherapy curriculum should adapt in order to
integrate data, technological, and human literacy across all that
is considered to be core to the profession, as clinicians who are
unable to communicate with AI-based technology are likely to
become increasingly irrelevant. Educators should acknowledge
that the curriculum must serve people for the duration of their
careers and not only offer time-limited undergraduate and
postgraduate degrees. Clinicians will need to access the
curriculum several times over the course of a career as they
work to stay ahead of the changes in practice wrought by an
increasingly competent variety of intelligent algorithms. We
would do well to focus on adapting physiotherapy education and
clinical practice for a radically different future, one in which we
learn how to excel at the things that computers find difficult to
replicate. Human connection will be key to success in an
intelligence age and we must take every opportunity to enhance
our capacity to care for each other, to learn effectively over the
course of our lives, and to develop creative solutions for the
problems that matter to us.
References
Aoun, J. (2016). Higher education for the AI age: Let’s think
about it before the machines do it for us. The Washington Post,
27 October..
Aoun, J. (2017). Robot proof: Higher Education in the Age of
Artificial Intelligence. MIT Press.
Barradell, S. (2017). Moving forth: Imagining physiotherapy
education differently. Physiotherapy Theory and Practice, 33(6),
439–447.
Bleakley, A. (2010). Blunting Occam’s razor: Aligning medical
education with studies of complexity. Journal of Evaluation in
Clinical Practice, 16(4):849-855.
Brynjolfsson, E. & McAfee, A. (2014). The second machine-age:
Work, progress, and prosperity in a time of brilliant
technologies. W.W. Norton & Company, New York.
Claiborne, J.S. (2018). Anticipating and Training the Physician
of the Future. Academic Medicine. Preprint.
Frankish, K. & Ramsey, W.M. (2014). The Cambridge Handbook
of Artificial Intelligence. Cambridge University Press, University
Printing House, Cambridge, CB2 8BS, United Kingdom.
Fraser, S.W. & Greenhalgh, T. (2001). Coping with complexity:
Educating for capability. BMJ, 323(7316):799-803.
Frenk, J., Chen, L., Bhutta, Z.A., et al. (2001). Health
professionals for a new century: Transforming education to
strengthen health systems in an interdependent world. Lancet,
376(9756):1923-1958.
Gottfredson, L.S. (1997). Mainstream science on intelligence.
Intelligence, 24(1):13-23.
Harwich, E., & Laycock, K. (2018). Thinking on its own: AI in
the NHS. The Reform Research Trust, Charity no. 1103739.
Hodges, B.D. (2018). Learning from Dorothy Vaughan: Artificial
Intelligence and the Health Professions. Medical Education 52
(1): 11–13.
Minsky, M., in Darrach, B. (1970). Meet Shaky, the first
electronic person. Life Magazine, November.
Morris, I. (2010). Why the West rules - for now. The patterns of
history and what they reveal about the future. Profile books, 3A
Exmouth House, Pine Street, London, EC1R 0JH.
Nicholls, D. (2018). What’s the difference between a technician
and an artist? Critical Physiotherapy Network, 9th May.
Obermeyer, Z. & Thomas, H.L. (2017). Lost in Thought - The
Limits of the Human Mind and the Future of Medicine. New
England Journal of Medicine, 1209–11.
Panagiotis, K. (2017). Man Versus Machine: The Future of
Medicine. Academic Medicine 92 (5): 578.
Simon, H. (1965). The Shape of Automation for Men and
Management (1st edition). Harper & Row.
Susskind, R. & Susskind, D. (2015). The Future of the
Professions: How Technology will Transform the Work of
Human Experts. Oxford University Press, Great Clarendon
Street, Oxford, OX2 6DP, United Kingdom.
Topol, E. (2015). The patient will see you now: The future of
medicine is in your hands. Basic Books, 250 West 57th Street,
New York, NY 10107.
Wartman, S.A. & Combs, C.D. (2017). Medical Education Must
Move from the Information Age to the Age of Artificial
Intelligence. Academic Medicine. Preprint.
Winston, P. (1992). Artificial intelligence (3rd edition). Pearson.
5 of 6
OpenPhysio
Rowe, M. (2019). Artificial intelligence in clinical practice: Implications for physiotherapy education.
Peer review reports
David Nicholls
(david.nicholls@aut.ac.nz)
Auckland University of Technology, New Zealand
Reviewed: 16 July, 2018
Citation: Nicholls, D. (2018). Review - Artificial intelligence in
clinical practice: Implications for physiotherapy education.
OpenPhysio.
Ken Ma
sters (kmasters@squ.edu.om)
Sultan Qaboos University, Muscat, Sultanate of Oman
Reviewed: 16 July, 2019
Citation: Masters, K. (2019). Review - Artificial intelligence in
clinical practice: Implications for physiotherapy education.
OpenPhysio.
6 of 6
