See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/330051837

Technology and mental health: The role of artificial intelligence

Article  in  European Psychiatry · January 2019

DOI: 10.1016/j.eurpsy.2018.08.004

CITATIONS READS

9 1,904

3 authors, including:

Chris Lovejoy
St George's, University of London
19 PUBLICATIONS   77 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Factors affecting surgical outcomes View project

All content following this page was uploaded by Chris Lovejoy on 08 April 2019.

The user has requested enhancement of the downloaded file.

 European Psychiatry 55 (2019) 1–3

Contents lists available at ScienceDirect

European Psychiatry
journal homepage: http://www.europsy-journal.com

Viewpoint

Technology and mental health: The role of artificial intelligence

1. Introduction AI may be combined with sensors and smartphone applications

to enable increased monitoring in the community. Cogito, a
Advances in computational capacity, data collection and Boston-based AI company, developed the app “Companion” which
machine learning are contributing to an increasing interest in records behavioural indicators via smartphones. In a study of 73
artificial intelligence (AI), as reflected by a recent surge in funding people with either PTSD or depression, they found that number of
and research. AI has many potential applications within medicine. outgoing calls, count of unique numbers texted, absolute distance
It is being used to automate the national 111 triage service [1] and travelled, dynamic variation of the voice, speaking rate and voice
can detect retinal pathology as effectively as consultant oph- quality were all good predictors of symptoms of depression and
thalmologists [2]. However, before AI can safely have an impact PTSD [5]. As well as behavioural indicators, physiological data may
within psychiatry, there are a number of issues which must be provide further objective measures of mental health. In a group of
considered, such as capacity and consent, data security and patient 251 college students, Sano et al. found skin conductance and
privacy, and clinical governance. temperature to have 78% accuracy for classifying students into
high/low stress groups and 86% accuracy for classifying high/low
2. Diagnostic tools mental health groups [6].
Another aspect of monitoring is medication adherence, which is
One challenge for clinicians in making diagnoses is that patient an issue in all chronic health problems and may particularly be so
interactions only offer a snapshot of an individual’s mental state, in mental health conditions. The estimates of non-adherence to
yet mood disorders are dynamic in nature and fluctuate over time. antipsychotic medications range from 20% to 89% for patients with
At present, psychiatric assessment of patients includes observation schizophrenia or bipolar disorder [7]. Several smartphone
of their mental state and subjective self-report questionnaires, applications aim to improve adherence by giving reminders and
such as the GAD-7 for anxiety and PHQ-9 for depression. These helping patients keep track of their medications. The use of
methods are subjective, difficult to repeat and can be time- machine learning may facilitate continual improvement of such
consuming. AI may enable additional methods, such as audio and applications, including tailoring to the individual, to maximise
video analysis, which have greater objectivity and may have better their effect on medication adherence.
predictive value. IBM research developed a machine learning
speech classifier with 79% accuracy in predicting psychosis onset in 4. Treatment
those with clinically high risk [3]. Another group showed how
computer vision can detect ADHD and ASD with 96% accuracy Artificial intelligence can enable existing treatments to be
based on video analysis of a person’s behaviour [4]. As well as provided through novel means, which may increase availability
assisting with diagnosis, such tools may also help monitor the and effectiveness.
progress of patients, in both inpatient and outpatient settings.
4.1. CBT chatbots
3. Monitoring
Internet-based cognitive behavioural therapy (CBT) has been
All but the most severe psychiatric care takes place in an offered since the 1990s but has been characterised by low
outpatient setting so additional monitoring in the community adherence. The development of CBT chatbots, which mimic normal
would offer significant benefit. Early detection and prevention of conversational style to deliver CBT, may increase adherence and
relapse can have a significant impact on outcomes. offer other advantages. Kirkpatrick et al. found that one such
One aspect of monitoring mood and mental health outside the chatbot, named ‘Woebot’, decreased both depression and anxiety
clinical setting is written records kept by patients, such as mood in college students over a 2 week course [8].
diaries. This is prone to recall bias and compliance issues. Apps The optimal use of such a chatbot is still unclear given the
have been developed which actively prompt users to answer preliminary stage of research. While some early studies suggest
questions on their mood, sleeping patterns and other relevant improvement in anxiety and depression [8], one study showed an
areas, although some of the same issues persist. increase in anxiety and alcoholic consumption among Japanese

http://dx.doi.org/10.1016/j.eurpsy.2018.08.004
0924-9338/© 2018 Elsevier Masson SAS. All rights reserved.
2 C.A. Lovejoy et al. / European Psychiatry 55 (2019) 1–3

workers, which authors attributed to heightened awareness of 7. Clinical governance

pathological thought and drinking behaviour [9].
The regulation of mobile apps is significantly less than the
4.2. Accessibility, de-stigmatisation and personalisation regulation of medical products and treatments. For example, over
47,000 mental health apps were on sale to US consumers in 2015.
In 45% of the world, there is less than one psychiatrist to every The majority of these had not been validated, and of the few that
100,000 people yet over 50% of the world population now owns a had it was primarily by small-scale, short-term pilot studies. The
smartphone. A single AI system can be used by very large use of unvalidated smartphone apps poses risk to patients due to
populations, provided they have access to the required hardware poor quality information and potentially harmful recommenda-
which is increasingly ubiquitous. Such technology could increase tions. Many of the apps available for mental health may not be
accessibility by reducing lengthy and costly travel to centrally- classed as ‘medical products’ as they focus more on self-
located mental health clinics. It may enable over-burdened mental management and lifestyle
health professionals to increase the reach of their services. It could Clinical governance in this area is important to prevent
also help people whose conditions restrict their ability to travel, unregulated use and potential harm. In the UK, the NHS has made
such as GAD, agoraphobia or physical health problems. efforts to introduce this by creating the “NHS Digital Apps Library”
AI may also increase accessibility by circumventing some of the which includes ‘NHS Approved’ labels for those which sufficient
stigma surrounding mental illness. For example, chatbots may evidence of effectiveness and safety.
avoid such stigma as they are not part of a wider social construct It should not be assumed that new technologies will definitely
with all the associated cultural norms and expectations. They are have a positive effect and they can plausibly cause harm. For
more likely to be perceived as non-judgmental, non-opinionated example, while some feel more empowered to control their illness
and overall neutral. by using passive monitoring, others may feel overwhelmed by the
AI may enable greater personalisation of care. For example, additional responsibility, or perceive it as a constant reminder that
more sophisticated analysis of large volumes of data may enable they are ill [10]. Some technologies may benefit certain sub-
better prediction of therapeutic response and side effect profile populations yet be harmful to others and these distinctions must
from different medications. This would reduce the need to perform be borne out by thorough research.
multiple trials of different medications.
8. Conclusion
5. Re-balancing clinician workload
The successful integration of AI into healthcare could dramati-
AI may drastically re-balance a clinician’s workload, giving cally improve quality of care. In psychiatry, new tools for diagnosis,
them more time to interact with patients, thereby improving the monitoring and treatment may improve patient outcomes and re-
quality of care. This is particularly true within psychiatry because it balance clinician workload. While there is great potential,
is a text-heavy field. Psychiatrists spend lots of time reading numerous risks and challenges will arise. These will require
previous notes to build an accurate picture of a patient’s history. careful navigation to ensure the successful implementation of this
Natural Language Processing (NLP) is an area of AI which involves new technology.
analysing human language for meaningful information. NLP could
be used to summarise the most important data from a patient’s Conflicts of interest
electronic health records, providing a succinct summary at the
beginning of a consultation. For example, a timeline of mental state Christopher Lovejoy and Varun Buch are employees of Cera
examinations and the associated treatment regime could be Care. Mahiben Maruthappu is an investor and employee of Cera
produced. NLP could also provide more powerful analysis if Care. Cera Care is a domiciliary care provider conducting research
combined with AI analysis of speech and video to provide a short into how artificial intelligence can be used to improve the care
summary of a patient’s mental state, which could be a useful delivered to elderly people living at home. This work received no
supplement to the psychiatrist’s mental state examination. Such an specific grant from any funding agency in the public, commercial,
algorithm would be objective and not at risk of inter-clinician or not-for-profit sectors.
variability.
References
6. Data security, privacy and consent
[1] Burgess M.. The NHS is trialling an AI chatbot to answer your medical
questions. [cited 2018 Jun 15]. Available from:. http://www.wired.co.uk/
Healthcare data is sensitive and mental health data is article/babylon-nhs-chatbot-app.
particularly so given the high risk of stigmatisation and [2] Fauw JD, Ledsam JR, Romera-Paredes B, Nikolov S, Tomasev N, Blackwell S,
discrimination in the case of data disclosure. For the public to et al. Clinically applicable deep learning for diagnosis and referral in retinal
disease. Nat Med 2018;(August 13):1.
embrace this new technology, which is necessary for it to be [3] Corcoran CM, Carrillo F, Fernández-Slezak D, Bedi G, Klim C, Javitt DC, et al.
effective, the healthcare profession must earn their trust. There Prediction of psychosis across protocols and risk cohorts using automated
have been high profile cases involving misuse of personal data, language analysis. World Psychiatry 2018;17(February (1)):67–75.
[4] Jaiswal S, Valstar M, Gillott A, Daley D. Automatic detection of ADHD and ASD
such as the Facebook-Cambridge Analytica scandal. Similar events from expressive behaviour in RGBD data. December 7, ArXiv161202374 Cs.
within healthcare could have serious consequences. Available from:. 2016. http://arxiv.org/abs/1612.02374.
Mental illness can affect capacity and thus the ability to provide [5] Place S, Blanch-Hartigan D, Rubin C, Gorrostieta C, Mead C, Kane J, et al.
Behavioral indicators on a mobile sensing platform predict clinically validated
consent. Presence of capacity can fluctuate over time. A patient
psychiatric symptoms of mood and anxiety disorders. J Med Internet Res
may initially give consent for passive monitoring, for example, but 2017;19(March (3)):e75.
if they later lose capacity due to worsening mental health then it is [6] Sano A, Taylor S, McHill AW, Phillips AJ, Barger LK, Klerman E, et al. Identifying
unclear whether the patient’s initial consent remains valid. AI objective physiological markers and modifiable behaviors for self-reported
stress and mental health status using wearable sensors and mobile phones:
would also require patients to consent to much greater amounts of observational study. J Med Internet Res 2018;20(6):e210.
data; those who consent to medical notes may not so easily [7] Stentzel U, van den Berg N, Schulze LN, Schwaneberg T, Radicke F, Langosch JM,
consent to video, audio and other forms of data being stored. et al. Predictors of medication adherence among patients with severe
 C.A. Lovejoy et al. / European Psychiatry 55 (2019) 1–3 3

b
psychiatric disorders: findings from the baseline assessment of a randomized Department of Psychiatry, Springfield University Hospital, South
controlled trial (Tecla). BMC Psychiatry 2018;18(May 29):155. West London and St George’s Mental Health NHS Trust, United
[8] Fitzpatrick KK, Darcy A, Vierhile M. Delivering cognitive behavior therapy to
young adults with symptoms of depression and anxiety using a fully Kingdom
automated conversational agent (Woebot): a randomized controlled trial.
JMIR Ment Health 2017;4(June (2)):e19. Varun Buch
[9] JMH-standalone effects of a cognitive behavioral intervention using a mobile Mahiben Maruthappu
phone app on psychological distress and alcohol consumption among Cera Care, London, United Kingdom
Japanese workers: pilot nonrandomized controlled trial | Hamamura | JMIR
Mental Health. Available from: http://mental.jmir.org/2018/1/e24/.
[10] Glenn T, Monteith S. New measures of mental state and behavior based on data * Corresponding author at: Harewood House, Springfield Uni-
collected from sensors, smartphones, and the Internet. Curr Psychiatry Rep versity Hospital, 61 Glenburnie Road, London, SW17 7DJ, United
2014;16(December (12)):523.
Kingdom.
E-mail address: chris.l@ceracare.co.uk (C. Lovejoy).
Christopher A. Lovejoya,b,*
a
Cera Care, London, United Kingdom
Received 23 August 2018

Available online xxx

View publication stats