Revolution in Health Care: How Will Data Science Impact Doctor–Patient Relationships?
April 2018 | Volume 6 | Article 991
OpiniOn
published: 03 April 2018
doi: XXXXXXXXXX/fpubh XXXXXXXXXX
Frontiers in Public Health | www.frontiersin.org
Edited by:
Enrico Capobianco,
University of Miami,
United States
Reviewed by:
Rimpi Khurana,
University of Miami,
United States
*Co
espondence:
Ivan Lerner
XXXXXXXXXX
†Co-first authors.
Specialty section:
This article was submitted
to Digital Health,
a section of the journal
Frontiers in Public Health
Received: 21 Fe
uary 2018
Accepted: 16 March 2018
Published: 03 April 2018
Citation:
Lerner I, Veil R, Nguyen D-P, Luu VP
and Jantzen R XXXXXXXXXXRevolution in
Health Care: How Will Data Science
Impact Doctor–Patient Relationships?
Front. Public Health 6:99.
doi: XXXXXXXXXX/fpubh XXXXXXXXXX
Revolution in Health Care:
How Will Data Science impact
Doctor–patient Relationships?
Ivan Lerner1*†, Raphaël Veil2†, Dinh-Phong Nguyen2, Vinh Phuc Luu3 and
Rodolphe Jantzen4
1 UMR8156 Institut de recherche interdisciplinaire sur les enjeux sociaux Sciences sociales, Politique, Santé (IRIS), Paris,
France, 2 So
onne Université, UPMC Univ Paris 06, Paris, France, 3 Univ Paris Diderot, So
onne Paris Cité, Faculté de
médecine, Paris, France, 4 Université Paris Est, Faculté de médecine, Créteil, France
Keywords: doctor–patient relationship, data science, artificial intelligence, machine learning, digital health
Over the last decade, a technical revolution has taken place in several industrial sectors, starting
with internet companies. The computerization and interconnection of a wide variety of services
and devices has facilitated the collection and storage of data, which has since increased by several
orders of magnitude. The exploitation of these data has completely reshaped some services, such
as Internet advertising, which has become largely personalized, while
inging with it its fair share
of privacy issues. As data management and analysis have become central to many businesses,
computer scientists have been called upon to provide tools capable of extracting knowledge from
ever-growing, structured and unstructured databases. In this context, a paradigm shift occu
ed
in data analysis as more data became available; with deep learning, data-driven approaches are
nowadays often surpassing domain-specific approaches (1). Indeed, in very diverse predictive
tasks, such as machine translation (2), object recognition (3) or speech recognition (4), general
purpose models such as artificial neural networks have outperformed advanced algorithms devel-
oped by experts with domain-specific knowledge. Additionally, these machine learning algorithms
have often reached experts’ performance level at various tasks, including medical diagnosis (5–8).
However, these great successes have often been achieved at great expense: the acquisition of a large
amount of structured and unstructured data.
Big health-care data are already a reality; academias, industries, insurance agencies, and public
health systems struggle to adapt their infrastructure to a data volume whose size is doubling every
12–14 months (9). Such storage systems are also challenging in terms of accessibility, ownership,
and privacy issues (10). Still, medical uses remain mainly in the field of research, aiming to provide
information about patients’ conditions by analyzing massive amounts of data and assisting with
decision-making. Within hospitals, some data-driven softwares are being developed to identify
patients at high risk of hospital mortality (11), while others predict patient affluence and/or wait-
ing times in emergency departments (12). Outside the hospital, data-driven applications are also
flourishing in various fields, such as telemonitoring systems, implementing advanced prediction of
asthma exace
ations (13), or automatic detection of falls in the elderly population (14). Moreover,
precision medicine aims to provide a personalized recommendation of the optimal treatment for
each patient, relying on the analysis of large heterogeneous datasets, including imaging, genomics,
or various biological values extracted from electronic health records. This framework can be applied
in many areas of medicine, such as radiation oncology (15), psychiatry (16), and infectious diseases
(17). While these developing medical applications will require rigorous clinical validation, many
should find their way into daily clinical practice over the next few decades (Figure 1). How will such
innovations impact clinicians and their relationships with patients?
First, let us explore what was driving this relationship before data science found its way into
health care. Traditionally, the work of physicians is a balancing act between technical expertise and
human interpersonal skills. On the one hand, medical doctors aim to improve their knowledge to
https:
www.frontiersin.org/Public_Health
https:
crossmark.crossref.org/dialog/?doi=10.3389/fpubh XXXXXXXXXX&domain=pdf&date_stamp= XXXXXXXXXX
https:
www.frontiersin.org/Public_Health/archive
https:
www.frontiersin.org/Public_Health/editorialboard
https:
www.frontiersin.org/Public_Health/editorialboard
https:
doi.org/10.3389/fpubh XXXXXXXXXX
https:
www.frontiersin.org/Public_Health
https:
www.frontiersin.org
https:
creativecommons.org/licenses
y/4.0
mailto: XXXXXXXXXX
https:
doi.org/10.3389/fpubh XXXXXXXXXX
https:
www.frontiersin.org/Journal/10.3389/fpubh XXXXXXXXXX/full
https:
www.frontiersin.org/Journal/10.3389/fpubh XXXXXXXXXX/full
https:
www.frontiersin.org/Journal/10.3389/fpubh XXXXXXXXXX/full
https:
loop.frontiersin.org/people/528764
https:
loop.frontiersin.org/people/530954
https:
loop.frontiersin.org/people/531928
https:
loop.frontiersin.org/people/532055
https:
loop.frontiersin.org/people/530992
FiguRe 1 | Revolution in healthcare.
2
Lerner et al. Data Science Impact on Healthcare
Frontiers in Public Health | www.frontiersin.org April 2018 | Volume 6 | Article 99
diagnose diseases more accurately and recommend optimal treat-
ment for each specific health condition. On the other hand, they
strive to be more empathetic toward patients, taking into account
their psychosocial background and cultural beliefs. Today, most
doctors know that engaging in a dialog with each patient is criti-
cal in delivering adapted information, improving adherence to
treatment, and ensuring an understanding of their condition.
However, in the mid-twentieth century, with the biomedical
evolution reaching its peak, medicine became increasingly
efficient but also more technical. In addition, the trust between
doctors and patients has always faced multiple challenges,
including language and cultural ba
iers, potentially threatening
alternative medicine, or more recently the decreasing importance
of family practice (18). In light of these evolutions, data science
is not going to be the first phenomenon to challenge physicians’
ability to adapt and rethink their profession (Figure 2).
This rapid increase in technical knowledge regarding medi-
cine and biology
ought, as a corollary, the development of
methodically conceived standardized guidelines for physicians
to apply. In an effort to push evidence-based medicine onto the
field, health authorities worked with experts to offer, promote,
and soon enforce the strict application of these guidelines by
more and more professionals (19). This phenomenon aimed to
increase health-care quality, equity, and security to patients, is
sometimes criticized for being too rigid to apply individually
and for reducing the autonomy of physicians (20). In France
for instance, it is not yet mandatory to follow such guidelines,
ut physicians may be held responsible for not being able to
justify their deviation from them. Some physicians have now
traded this decision-making privilege with much more implica-
tion in research, public health, the elaboration process of these
guidelines, or simply the exploration of new areas of expertise in
their field. And all these new tasks, including the ones that came
along with the standardization of health care, have taken up
an increasing amount of personal and collective efforts, leaving
ever less time to dedicate to the improvement of doctor–patient
elationships.
Alongside these changes, a cultural movement inspired by
the social sciences and humanities appeared in North America
in the late 1970s, focusing on the importance of human skills in
clinical practice (21). One trigger may have been the realization
that patient adherence to treatment was much lower under real
conditions than in clinical trials, sometimes making efforts to
improve treatment efficacy i
elevant (22, 23). As end-of-life
situations became more and more medicalized, patients’ expecta-
tions of care attitudes often contrasted with physicians’ healing
ehaviors. This led to the development of palliative care (24, 25),
with a special focus toward patients’ reported experience. Patient
empowerment through stronger patient organizations has also
led to more patient-centered care (26, 27). Additionally, atten-
tion to patient choice, consent, psychosocial context, and cultural
eliefs has become increasingly important in research (28, 29)
and clinical practice, so much so that it became a key element
in the concept of Evidence-Based Medicine. Medical schools
then started training their students in medical ethics, patient
communications, therapeutic education, and na
ative medicine
(30, 31). In most developed countries, laws have been amended to
ecognize patients’ right to refuse treatment or to be informed of
their medical conditions, thus striking a blow to the long-lasting
paradigm of medical paternalism. How will these cultural move-
ments respond to the a
ival of medical data-driven applications,
a new set of technologies aimed at further “dehumanizing” clini-
cal practice?
In light of these recent changes in health care, one may
anticipate the impact of data science and its medical applica-
tions on clinical practice and doctor–patient relationships.
At first glance, it is tempting to see such disruptive new technolo-
gies as a factor for an evermore dehumanized medicine, where
doctors–patient relationships would come down to sensors and
computer screens. Based on remote monitoring signals provid-
ing detailed clinical information, machine learning algorithms
could, for example, display risks of various patient outcomes for
each hypothetical therapeutic strategy. Ultimately, physicians
may no longer require direct interaction with their patients to
accurately assess their clinical and even psychological status.
Conversely, one can argue that the progressive automation of
various tasks of clinical practice could free up more time for
physicians to invest in an improved doctor–patient relationship.
Today, most physicians spend a large amount of time trying to
detect potential drug interaction, searching for specific events
in the patient’s medical history, and surveying repeated and
various lab results. This precious medical time could arguably
e better spent with the patient, discussing therapeutic choices,
assessing treatment comprehension, or detecting psychological
vulnerability. Moreover, continuously used monitoring devices,
if co
ectly deployed, could very well enhance doctor–patient
elationships by extending it beyond the walls of the physi-
cian’s office. Take the case of type 2 diabetes: today’s protocols
ecommend for a regular follow-up consult in a predetermined
https:
www.frontiersin.org/Public_Health
https:
www.frontiersin.org
https:
www.frontiersin.org/Public_Health/archive
3
Lerner et al. Data Science Impact on Healthcare
Frontiers in Public Health | www.frontiersin.org April 2018 | Volume 6 | Article 99
FiguRe 2 | Evolution of the physician’s role in healthcare.
sequence, depending on stage and complications; yet, based on
multiple inputs (e.g., gly cemia controls, self-reported symp-
toms, urine test strips, etc.), predictive algorithms could very