Access to this full-text is provided by Wiley.
Content available from Health Expectations
This content is subject to copyright. Terms and conditions apply.
Health Expectations
ORIGINAL ARTICLE OPEN ACCESS
Healthcare Professionals' Views on the Use of Passive
Sensing and Machine Learning Approaches in Secondary
Mental Healthcare: A Qualitative Study
Jessica Rogan
1
| Joseph Firth
1,2
| Sandra Bucci
1,2
1
Division of Psychology and Mental Health, School of Health Sciences, Faculty of Biology, Medicine and Health, Manchester Academic Health Science Centre,
The University of Manchester, Manchester, UK |
2
Greater Manchester Mental Health NHS Foundation Trust, Manchester, UK
Correspondence: Sandra Bucci (Sandra.bucci@manchester.ac.uk)
Received: 16 September 2024 | Revised: 5 November 2024 | Accepted: 14 November 2024
Funding: This study was supported by the NIHR Manchester Biomedical Research Centre NIHR 203308) and the National Institute for Health and Care
Research research professorship (NIHR300794).
Keywords: artificial intelligence | interviews | machine learning | mental healthcare | passive sensing | qualitative | views
ABSTRACT
Introduction: Globally, many people experience mental health difficulties, and the current workforce capacity is insufficient to
meet this demand, with growth not keeping pace with need. Digital devices that passively collect data and utilise machine
learning to generate insights could enhance current mental health practices and help service users manage their mental health.
However, little is known about mental healthcare professionals' perspectives on these approaches. This study aims to explore
mental health professionals' views on using digital devices to passively collect data and apply machine learning in mental
healthcare, as well as the potential barriers and facilitators to their implementation in practice.
Methods: Qualitative semi‐structured interviews were conducted with 15 multidisciplinary staff who work in secondary mental
health settings. Interview topics included the use of digital devices for passive sensing, developing machine learning algorithms
from this data, the clinician's role, and the barriers and facilitators to their use in practice. Interview data were analysed using
reflexive thematic analysis.
Results: Participants noted that digital devices for healthcare can motivate and empower users, but caution is needed to
prevent feelings of abandonment and widening inequalities. Passive sensing can enhance assessment objectivity, but it raises
concerns about privacy, data storage, consent and data accuracy. Machine learning algorithms may increase awareness of
support needs, yet lack context, risking misdiagnosis. Barriers for service users include access, accessibility and the impact of
receiving insights from passively collected data. For staff, barriers involve infrastructure and increased workload. Staff support
facilitated service users' adoption of digital systems, while for staff, training, ease of use and feeling supported were key
enablers.
Conclusions: Several recommendations have arisen from this study, including ensuring devices are user‐friendly and equitably
applied in clinical practice. Being with a blended approach to prevent service users from feeling abandoned and provide staff
with training and access to technology to enhance uptake.
Patient or Public Contribution: The study design, protocol and topic guide were informed by a lived experience community
group that advises on research projects at the authors' affiliation.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly
cited.
© 2024 The Author(s). Health Expectations published by John Wiley & Sons Ltd.
1of16Health Expectations, 2024; 27:e70116
https://doi.org/10.1111/hex.70116
1 | Introduction
Approximately 10% of the world's population is living with a
mental health difficulty [1]. Those who experience severe
mental health difficulties, such as psychosis, can be at greater
risk of poorer lifelong outcomes [2]. This is partly associated
with the risk of relapse, which can be distressing for individuals
and resultant treatment costs can be expensive [3]. Mental
healthcare that supports long‐term treatment management and
assessment is therefore important to reduce the risk of relapse
and promote wellbeing [4]. However, a substantial gap persists
between those needing care and those with access to care [5]. A
global shortage of mental healthcare providers has cast doubt
on the possibility of ever meeting the demand for mental
healthcare through in‐person care alone [6].
Digital technology could play a significant role in addressing the
challenges faced by services and people who experience mental
health difficulties by offering flexible and tailored approaches to
mental healthcare that are more accessible than those currently
available [7]. Indeed, it has been suggested that digitising
healthcare will support people to live independently, have more
control over their lives and prevent health and social care needs
from escalating [8]. The rapid growth in access to digital
devices, which for the purposes of this article refers to all
modern smartphones, smartwatches and other wearable devices
that have sensors embedded, could augment traditional mental
healthcare and bridge the gap between the need for treatment
and capacity to deliver it [6]. Passive sensing allows for the
continuous, unobtrusive collection of health data and can be
used to understand an individual's activity levels, sleep and
social engagement [9]. A large amount of raw sensor data can
be collected this way and translated into markers of behaviour,
emotions and mental states using machine learning, which is a
branch of artificial intelligence (AI) that can be applied to large
datasets to ‘learn’, identify patterns and draw inferences [10].
Machine learning methods can essentially detect when an in-
dividual's online footprint, generated from apps, social media
and other personal monitoring devices, may indicate an
impending risk of illness, self‐harm or even violence, thereby
automating risk monitoring [11]. Service users can therefore
track and collect clinically relevant data using their mobile or
wearable devices and share this data with health professionals
[12]. This can assist in the detection, diagnosis and treatment of
mental health difficulties [13]. For example, raw sensor data
captured about speech characteristics, location and activity can
be transformed to derive high‐level behavioural markers around
fatigue, sleep disruption and mood, which can be used to
identify clinical states, such as depression [10]. Recommenda-
tions, feedback and reminders can then be delivered in‐the‐
moment to inform users about their current state and how to
improve it [14].
Bucci et al. suggest that theory‐driven digital interventions that
monitor distressing experiences and provide near real‐time
active management strategies could improve the speed and
quality of recovery in psychosis over and above conventional
treatments [15]. Digital interventions can also encourage
autonomous motivation which, according to self‐determination
theory, is facilitated through satisfaction of three basic psy-
chological needs: autonomy, competence and relatedness [16].
When service users have these needs met in healthcare, they
may be more engaged in treatment and experience better health
outcomes [17]. In line with this, digital health interventions
(DHIs) can empower individuals by facilitating self‐
management, increasing self‐awareness and prompting the use
of coping skills [18, 19]. Other potential benefits of digitally‐
enabled care pathways include the collection and provision of
objective measures that can supplement subjective self‐reports,
which can be prone to recall bias [20]. Furthermore, traditional
service delivery methods rely on service users seeking treatment
themselves, but, mental health difficulties can hinder help‐
seeking behaviour. Thus, digital health systems could identify
individuals in need of support and accelerate their access to
treatment [10].
Conversely, previous research has highlighted concerns about
the potential for service users to interpret their behaviour data
negatively, which could result in demotivation and frustration
[18]. Concerns around data security and privacy issues are also
widely discussed, with fears that access to a large amount of
personal data could lead to confidentiality breaches [21]. In
addition, it has been suggested that replacing direct contact
with healthcare professionals with digital support could nega-
tively impact the therapeutic relationship. This is due to
reduced attendance to non‐verbal cues and loss of social con-
nection provided by in‐person appointments [21]. Further
concerns include the potential for clinicians to become reliant
on digital health systems that use AI, which could result in
misdiagnosis or missed diagnosis. This, in turn, could lead
service users to lose trust in clinical decision‐making [22].
Previous research has also highlighted clinicians' concerns that
integrating passively collected data and machine learning al-
gorithms into clinical practice could increase their workload.
Clinicians reported feeling ‘overwhelmed’when presented with
passive sensing data due to the huge volume of data generated
by sensors [23] and concerns about the time involved in re-
viewing this data before and during sessions [24]. Clinicians
have also expressed workload concerns about integrating this
data into practice, the methods for receiving service user data,
and their potential responsibility for monitoring data for risk
issues [25]. A recent systematic review of 175 papers examining
the barriers and facilitators of user engagement with DHIs for
people with severe mental illness [26] identified several key
facilitators. These included recognition of DHIs' advantages, a
clear alignment between the intervention focus and service user
needs, a simple and low‐effort digital interface, human‐
supported delivery and device provision when necessary.
Despite staff concerns about loss, damage or selling of equip-
ment, studies reported only an 11% device loss rate. Barriers
included intervention complexity, perceived risks, user moti-
vation, discomfort with self‐reflection, digital poverty, symp-
toms of psychosis, poor compatibility with existing clinical
workflows, fears that DHIs would replace traditional face‐to‐
face care, infrastructure limitations and limited financial sup-
port for delivery.
Mental health clinicians' views on using digital devices to col-
lect passively sensed data, which can then be used to generate
machine learning algorithms for insights into an individual's
mental health, may affect its adoption and usage [11]. For this
reason, engaging and involving frontline staff in the design and
2of16 Health Expectations, 2024
rollout of digital approaches to mental healthcare is essential for
ensuring utilisation and successful uptake [27]. Despite this,
there is sparse research exploring the views of mental health
staff on the use of passively sensed data collected from digital
devices and applying machine learning algorithms generated
from these data in mental healthcare. The aims of this research
are twofold, to explore: (i) healthcare professionals' views and
attitudes towards the use of digital devices, passive sensing
technology and machine learning algorithms as applied in
mental healthcare; and (ii) barriers and facilitators to incorpo-
rating digital healthcare systems into routine clinical practice.
Qualitative methods using individual semi‐structured inter-
views will enable in‐depth exploration of this relatively
unexplored area. This method will provide valuable insights
into participants' experiences and perceptions of digital systems
that utilise passive sensing and machine learning algorithms in
mental healthcare [28].
2 | Methodology
2.1 | Design
This study used a qualitative design. Semi‐structured in-
terviews were conducted with clinicians working in sec-
ondary mental health services (services for people with
complex or long‐term mental health conditions that are
beyond the scope of primary care) who approached the
research team and were eligible to take part (N=15). A
predominantly inductive approach to analysis was adopted.
Data was open‐coded, and meanings based on the inter-
pretations made by interviewees were emphasised. How-
ever, the questions asked in the interview, although used
flexibly, meant that deductive analysis was also employed to
ensure that the open coding allowed for the identification of
themes that were meaningful to the research questions
posed. The analysis approach emphasised the importance of
the researcher's subjectivity as an analytic resource and
their reflexive engagement with theory, data and interpre-
tation [29]. Questions were open‐ended with sequencing
dictated by the flow of the exchange. Probes were used to aid
further elaboration of responses. A critical realist episte-
mological position underpinned the research, recognising
that an objective social world exists outside of people's
constructions of it and that one's understanding of the social
world is limited by one's experiences and position within it
[30]. The study design, protocol and topic guide were in-
formed by a lived experience community group that advises
on research projects at the authors' affiliation. The group
advised on topic guide development, including phrasing of
questions and use of terminology and conducting qualitative
interviews. In the development of this study, the consoli-
dated criteria for reporting qualitative research (COREQ)
checklist was consulted with regard to reflexivity and study
design [31]. A working lay description of each of the rele-
vant concepts (e.g., active symptom monitoring, passive
sensing, machine learning) was given to help so that parti-
cipants were working from the same reference point. This
study was approved by the University Ethics Committee
(Ref: 2021‐12959‐20908) and the Health Research Authority
(21/HRA/5696).
2.2 | Recruitment
Participants were recruited from secondary mental health
services in the Northwest of England. Inclusion criteria
were (i) experience working with service users in secondary
mental health services; (ii) currently working in a clinical
role and (iii) agree for the interview to be audio‐recorded.
Service managers of relevant NHS Trusts were approached
and information about the study was sent, including a par-
ticipant information sheet and consent form to circulate to
staff. The first author (J.R.) also offered to attend depart-
mental meetings and contact details were provided, with
participants invited to contact J.R. with any questions. All
participants provided written informed consent and a final
sample of 15 staff members took part.
2.3 | Data Collection
All one‐to‐one interviews were conducted by the first author
(J.R.) between June and November 2022, over an online
platform (Microsoft Teams). Online interviews were offered
to reduce demands on participants' time. Participants were
asked to complete an anonymised demographic form. Inter-
views were audio‐recorded using an encrypted digital
recorderandlastedbetween30and60min(M=45). The
interviews were centred on a topic guide (Table S1), which
was created based on the research questions and existing
research. Open‐ended questions were used to facilitate dis-
cussion around: (a) participant's personal use of digital
devices; (b) digital devices in healthcare, including specific
discussion around passive sensing and machine learning and
(c) impact on role and practice. Early transcripts were dis-
cussed within the research team to reflect on the topic guide
and allow refinement, such as further exploration of areas.
For example, the topic of consent was raised and so prompts
were added to explore this further.
2.4 | Data Analysis
Interviews were transcribed verbatim, anonymised, given an
identifying code and stored securely. Transcripts were read
through whilst listening back to interviews to check for errors.
Data analysis followed the recursive and iterative six‐stage
thematic analysis protocol to facilitate coding and theme‐
identification: (i) familiarisation with the data; (ii) generating
initial codes; (iii) generating themes; (iv) reviewing potential
themes; (v) defining and naming themes and (vi) producing the
report [31]. Interview transcripts were read multiple times by
the author (J.R.) to allow familiarity with the data. Initial codes
were generated to summarise the relevant data content. Codes
were grouped into potential conceptual themes (J.R.) and re-
viewed and refined by the research team (J.R., S.B. and J.F.) to
ensure they were reflective of the original data, related to the
aims of the research and sufficiently distinct to ensure they
reflected differing theoretical concepts. Differences in opinion
were managed through discussion and reviewing the data
together. There was no attempt to determine intercoder reli-
ability, as this can contradict the interpretive agenda of quali-
tative research [32]. Instead, quality assurance of coding, theme
3of16
development and the final write‐up were guided by a tool for
evaluating research quality [29]. Data collection was stopped
when data saturation was reached [33]. NVivo qualitative data
analysis software (NVivo version 12, 2018) was used to facilitate
data management and analysis.
2.5 | Reflexivity
Field notes and reflexive logs were kept throughout the inter-
views. Reflexivity is a process that addresses the subjective
nature of research and increases the quality of research by en-
hancing the understanding of how the researcher's positions
and interests impact all stages of the research process [34]. J.R.
is a Trainee Clinical Psychologist and S.B. is an Academic
Clinical Psychologist, both of whom have experience working in
mental health services. J.F. is a UK Research and Innovation
Fellow. Both S.B. and J.F. are principal investigators in clinical
trials implementing digital health interventions and have ex-
tensive experience in conducting and supervising qualitative
studies.
All authors utilise digital devices in their day‐to‐day lives. It is
important to acknowledge that this may affect the analysis and
interpretation of the data as the areas focused on may be issues
the researchers found personally salient. Steps were taken to
minimise the impact of this. JR ensured that interview ques-
tions were presented in an open and neutral way with no
indication of the views of the research team. Participants were
asked about their thoughts and were encouraged to explore
both positives and negatives. JR also used reflective logs to
consider how participants' responses influenced their own
views and held this in mind throughout data collection and
analysis.
3 | Results
3.1 | Participants
Overall, 15 participants took part in this study (3 males, 12
females). Participants were aged between 30 and 62 years
(M= 45). Participants identified their ethnicity as White British
(n= 13), British Pakistani (n= 1) and Anglo‐Arab (n= 1).
Duration of experience working in secondary mental health
services ranged from 3 years to 35 years (M= 12.7 years). See
Table 1for participant characteristics.
3.2 | Key Themes
From the data, three distinct but inter‐related themes were
identified: (1) positives and negatives of digital devices, passive
sensing and machine learning methods in mental healthcare;
(2) barriers to use in clinical practice and (3) facilitators to use
in clinical practice. Seven subthemes were also identified.
Themes, subthemes and relationships between them are rep-
resented in Figure 1and summarised in Table 2.
Theme 1: Positives and Negatives of Digital Devices, Passive
Sensing and Machine Learning Methods in Mental Healthcare
During the interviews, participants were asked to consider digital
devices, passive sensing and machine learning methods separately
Consequently, Theme 1 is organised into digital devices, passive
sensing and machine learning methods; participants highlighted
both positive and negative aspects within each approach. We
acknowledge, however, the overlap in concepts, as passive sensing
is one function of digital devices like smartphones and smart-
watches. However, we were concerned that issues around digital
TABLE 1 | Participant characteristics.
Participant
number Gender Ethnicity Profession
Time worked in secondary
mental health services
P1 Female White British Clinical Psychologist 3
P2 Male White British Psychiatry Trainee 5
P3 Female White British Mental Health Nurse 20
P4 Female British Pakistani Consultant Psychiatrist 6
P5 Male White British Psychiatry Trainee 4
P6 Female White British Consultant Nurse 6
P7 Female White British Occupational Therapist 12
P8 Female White British Clinical Psychologist 7
P9 Female White British Mental Health Nurse 3
P10 Female White British Mental Health Nurse 17
P11 Female White British Principal Clinical
Psychologist
12
P12 Female White British Mental Health Nurse 35
P13 Male Anglo‐Arab Consultant Psychiatrist 28
P14 Female White British Specialist Practitioner 21
P15 Female White British Senior Counselling
Psychologist
11
4of16 Health Expectations, 2024
literacy might lead participants to conflate their views on different
elements of digital approaches explored in this study. As such, we
aimed to disentangle these elements to understand participants'
views on each approach distinctly.
Subtheme 1.1: Digital Devices
Participants described both the positive and negative aspects to
service users wearing or carrying digital devices with embedded
sensors that collect mental health data and provide visual
summaries, feedback and intervention strategies to service
users. Positive aspects of digital devices centred on motivation
and empowerment.
Using Digital Devices for Passive Data Collection Can Be Motivating
Participants said that setting goals, receiving alerts and getting be-
havioural insights from passively collected data might motivate
service users to be more physically or socially active. This could be
particularly beneficial for service users who are inactive, as their
inactivity may be maintaining some of their difficulties:
…we have a lot of long‐term patients who have been in
hospital a long time who don't exercise or very rarely get
up or they're not doing much. So, I think it could be
motivational for that patient group.
(Participant 5)
FIGURE 1 | Relationships between themes and subthemes. Solid lines reflect direct links between concepts. Dotted lines reflect indirect links
between concepts. Single‐direction arrows mean single direction; double‐ended arrows mean bidirectional relationship between constructs.
TABLE 2 | Summary matrix of main barriers and facilitators identified.
Barriers to using passively collected data from
digital devices in clinical practice
Facilitators to using passively collected data
from digital devices in clinical practice
Service users Access and accessibility Support from staff
The effect of receiving digitally delivered emotional
and behavioural insights on wellbeing and psychosis
Staff Lack of supportive infrastructure Access to training
Impact on workload Ease of use
Feeling heard and supported
5of16
There was specific discussion around how devices that prompt
engagement in activity and encourage motivation could be
useful for people who have experienced psychosis, as they can
have particularly poor health outcomes:
…people with psychosis, because of maybe the high doses
of medication or the types of medication they're on or the
nature of the illness, they've got a more sedentary lifestyle
and we know from the research that their lifespan can
be reduced by a number of years, can't it? And I think if
we can encourage people to be more active…I think
things like that are really helpful, because not everybody
has somebody there to motivate them to do things.
(Participant 14)
It was further suggested that clinicians could support ser-
vice users to connect with others who are using digital
devices, to form groups in which they can encourage
and motivate one another, which also facilitates social
connection:
You know for patients you could even have little erm self‐
help groups so you could set them up you know walking
groups ‘off you go, go and test this equipment’, you know
‘come back, have a bit of a healthy competition’.
(Participant 6)
Using Digital Devices for Passive Data Collection Can Be
Empowering
Participants discussed that service users accessing their data
and intervention strategies offered to them via devices could be
especially useful because it can be difficult when feeling dis-
tressed to remember coping strategies in‐the‐moment. A
prompt via the device could remind service users what they
could do to reduce their distress:
I suppose sometimes it's difficult to think in the moment
especially if it's a moment when you're feeling more
anxious or there's other stuff going on for you…to think
actually what's going to be helpful for me in this
moment…it might be harder.
(Participant 1)
Participants wondered if reminders of helpful strategies to
manage distress could support service users to take more con-
trol of, and responsibility for, their wellbeing, which could be
empowering:
…there's the proactive ‘I've got some control’element,
which is what we always want for people, is to have some
autonomy and participation in their healthcare. I'm not
having it done to me….
(Participant 12)
Participants also noted that when users self‐manage their
mental health, it could have positive repercussions on services
by reducing overall demand:
We know that mental health services are really stretched
(…) So, I suppose this would help hopefully, for the right
person who has the right plan, them to live independently
and manage their own mental health.
(Participant 11)
Participants highlighted that a potential downside to using
digital devices in mental healthcare is the unintended feeling of
abandonment users might experience from their healthcare
provider if this was used in replace of human contact.
Abandoning Service Users and Creating Inequalities
Participants cautioned that using digital devices might diminish the
‘humaneness’of mental healthcare, potentially increasing feelings
of isolation and abandonment by their healthcare provider if it
results in less direct contact and interaction with clinicians:
…Because the risk of it is that people could feel aban-
doned to devices, a bit like robot world….
(Participant 12)
In addition, participants expressed concern about exacerbating
inequalities in the mental healthcare system. They worried clini-
cians might inadvertently focus more on service users who are
engaging with the digital technology, as they would receive more
information from these users than from those who do not engage:
…I'm getting loads of information about the people that
have consented and a lot less about people who have
chosen not to have that then erm it would be…there's
potential for more focus to go on the people you're getting
the information about and less focus to go on the other
people on your caseload.
(Participant 1)
Concerns were also expressed around creating a sense of
abandonment once a service user was discharged from services
if the device was taken away from them at this point. Partici-
pants felt that there could be potential negative consequences
for a service user's recovery if the device was helping while they
were engaged with their mental health provider, only to find the
device was taken away once they were discharged from services:
Or do people only get access to this whilst they're in sec-
ondary care services? And then discharge…that feels a bit
unfair and not very helpful for promoting recovery.
(Participant 11)
Subtheme 1.2: Passive Sensing
This subtheme summarises participant's views on the use of passive
sensing in mental healthcare; that is, the passive collection of a large
amount of data, which both service users and clinicians have access
to. As in Subtheme 1.1, participants described both the positive and
negative aspects to passively collected data for mental health. The
positive aspect focused on enhanced data collection, while the
negative aspects included challenges with ongoing informed con-
sent and the accuracy of the data collected.
6of16 Health Expectations, 2024
Improved Data Collection Requiring Less Effort
Most participants reported that they often recommend using an
app or paper diary to track mood or behaviour in their clinical
practice. However, they noted that the required effort, time and
motivation can be a barrier for users. Participants recognised
that a digital device, which passively monitors a user's mental
health with minimal effort, might simplify data collection.
…they have to do less for us to collect the data, they don't
have to open an app, click things, answer questions or ring us
or text us. So, it requires less effort on their part if they carry
their phone around that will tell us lots of information.
(Participant 4)
Participants noted that active monitoring might heighten dis-
tress and hypervigilance, as some mood or symptom‐tracking
apps require significant time and effort for data input. In con-
trast, passive sensing is continuous and unobtrusive, potentially
making it less stressful for users:
I've worked with people on sleep before where they keep a
sleep diary, but then the amount of self‐monitoring people
have to do…I feel like that's actually unhelpful for sleep,
because it adds a level of stress to having to actually note the
times and remember and write them down…if the wearables
just doing it and they don't have to think about it, they can
justputitouttheirmindandthatdata'sstillgoingtobe
collected, that would be really helpful.
(Participant 8)
In addition, participants believed that passively sensed data is
more objective than subjective self‐reported data. As such, they
viewed passive data collection as a better method for gathering
information from service users since it reduces reliance on
potentially biased and less reliable self‐reporting of symptoms
and experiences. That is, participants perceived that service
users might not tell the truth, lack insight or struggle with
recollecting their thoughts, feelings and behaviours between
appointments when actively inputting data into an app or paper
diary, something that passively collected data can overcome:
…but yeah, it would be helpful, because obviously all our
clients don't tell us the truth, do they?
(Participant 10)
It would help with their recollection of things as well,
because as a consultant, I probably only see them
infrequently.
(Participant 13)
It was also noted that, from the service user's perspective, col-
lecting large amounts of data in near real‐time can validate their
experience of psychosis over time, as it allows them to see and
demonstrate their symptoms over time:
And it might be helpful for them because they can
actually show like look, I'm not sleeping, like I'm not.
(Participant 9)
Negatives around the use of passive sensing in mental health-
care concerned ongoing informed consent and the accuracy (or
inaccuracy) of data.
Concerns Around Privacy, Data Protection and Ongoing
Informed Consent
A concern raised during discussions on passive sensing was
around invading a service user's privacy, as clinicians might
have access to vast amounts of information that the service user
may not wish to share with their healthcare provider:
…they might not want their care team to know this
information like they've got a right to privacy so erm if
they feel like…it's a bit of an invasion of privacy.
(Participant 1)
Furthermore, concerns were raised about data protection. Parti-
cipants were mindful about the need to consider the destination,
storage, access and responsibility of passively collected data,
especially when the data highlights issues requiring attention:
If we're collecting all this data, we need to be really clear
about what will happen to the data. Will it be safe? Will it
be secure? What actions will we take if there's a problem
with the data?
(Participant 4)
The importance of capacity and informed consent was a
prominent issue raised throughout interviews. Participants
stressed that service users would need to know exactly what
they are agreeing to when data was being collected passively
over time. The passive nature of data collection can lead users
to forget that their data is being recorded. This raises concerns
about whether consent at regular intervals is necessary when
digital devices are used for passive data collection, especially
when this data informs mental health insights that impact
clinical decisions:
…. the service users must know that that is happening….
That you are collecting the information, why you're col-
lecting it and what it's going to be used for and if it's
going to be shared. I think that those are necessary
conversations.
(Participant 14)
Ongoing consent for passive data collection was deemed nec-
essary, as a service user may become unwell and either with-
draw or lose the capacity to provide consent:
I suppose that's why informed consent is going to be so
important and thinking as well with psychosis that
maybe capacity fluctuates…I think that's quite tricky….
(Participant 11)
The Accuracy (or Inaccuracy) of Passively Collected Data
Another issue raised was around clinicians trusting the data
they receive due to concerns that passive sensing data is not
7of16
always accurate and there may be negative implications if
clinical decisions are made based on inaccurate data:
It sounds good in theory, my only thing is, particularly if
we're using it to make clinical decisions, is to make sure
that it's completely accurate because erm I am a bit
sceptical in terms of how 100% definitely accurate it is.
(Participant 5)
Participants highlighted that data may be inaccurate if service
users behave in a certain way to influence the data their clini-
cian will receive. This could communicate a need, which might
be useful, but could also mask needs:
…they could carry it or not carry it erm and that might be
good or bad. They might use that as an indirect way of
saying look erm (…) something's wrong or they might try
and hide something and pretend everything's ok and
pretend that they're going to sleep or walking around.
(Participant 4)
Participants wondered how they would manage discrepancies
between self‐report and passive sensing data, particularly if the
data is not completely accurate, as dismissing service user's
claims could damage the therapeutic relationship and leave
service users feeling invalidated:
…if the patient disagrees with the tech, we can't just go,
‘nope the tech knows best’, we have to temper that with…
with patient experience because it is so important in
maintaining the relationship, the patient relationship.
(Participant 2)
Participants highlighted how it would be difficult to have these
conversations with service users as it might appear punitive,
perhaps re‐enforcing a power imbalance between clinician and
service user:
…because I'm not their mum. I don't want to go round
telling them off because they've lied to me or you know
not that they've lied, but you know they've not disclosed
the whole truth.
(Participant 10)
Subtheme 1.3: Machine Learning Algorithms
This subtheme centres on participants' views on the use of
machine learning methods in mental healthcare; that is, the idea
that machine learning methods can be applied to passively col-
lected data to identify patterns of potential worsening in an in-
dividual's mental health. As in Subthemes 1.1 and 1.2, participants
described both the positive and negative aspects of using machine
learning methods to develop algorithms that can be used in the
context of an individual's mental healthcare. The positive aspect
centred around the ability of a machine learning algorithm to
identify trends and patterns in behaviours that could increase
service users' and clinicians' understanding of an individual's dif-
ficulties and potentially spot early warning signs of worsening
mental health, thus facilitating early assessment and intervention.
Improved Awareness and Access to Support
Participants felt that recognising trends and patterns in a ser-
vice user's behaviour, relative to their own baseline, could be
useful in practice. This approach could support both the clini-
cian and the service user gain a deeper understanding of their
difficulties and the factors that may be contributing to their
onset and maintenance:
I like the idea that it would be their norm. So, you'd start
by actually looking at what they do over a period, so
hopefully we could capture them when they're relatively
well (…) then you'd have a period of time where you'd
capture data and then when they don't seem so well, you
could then spot that trend, so it'd be compared to
their norm.
(Participant 4)
This approach could help identify and raise service users'
awareness of early warning signs, enabling them to access or
seek support sooner or enabling clinicians to intervene earlier,
thereby reducing the risk of relapse:
…we can identify when to intervene at a much earlier stage,
which has lots of positives, positives for the patient, it could
keep them out of hospital, you know, it could identify a
medication review much earlier or you know simply support
that needs to be put in place.
(Participant 6)
Participants thought that this approach could be especially
beneficial for people who have already experienced an episode
of psychosis. Since relapse can be particularly distressing for
this group, being aware of early warning signs could help
reduce the risk of hospitalisation:
…I think for people with psychosis, it could be really
helpful for them to…you know for having a really good
understanding of relapse and early warning signs and to
notice them. It would be really helpful to have those
notifications that kind of remind them of the things that
would help or to go and see their GP.
(Participant 11)
Participants also expressed concerns about machine learning
algorithms, particularly highlighting the lack of contextual data
incorporated into algorithms, as well as the risk of diagnostic
overshadowing.
Lack of Context
Participants highlighted that mental health is subjective, and
participants questioned whether machine learning relies too
heavily on objective, physiological data, potentially overlooking
service user experience. That is, objectively collected data might
not always align with a user's subjective experience. Partici-
pants suggested incorporating service users' input to better
contextualise the data and ensure its accuracy:
8of16 Health Expectations, 2024
…. the device thinks that your sleep is really different last
night erm. . does this feel right from what your experience
of your sleep was? Because sometimes it's totally wrong
and if it's been totally wrong, you don't want it then
flagging up alerts (…) we want to be led by our service
user's perspective.
(Participant 8)
Concerns were also raised around diagnostic overshadowing,
where behavioural changes might be mistakenly attributed to a
mental health problem. In fact, these changes could indicate a
physical health problem, leading to misdiagnosis or dismissing the
underlying cause, potentially resulting in incorrect treatment:
So, I'll worry that someone's got this diagnosis of erm
depression for example and they're really fatigued and
they're sleeping loads and it's getting flagged up on a
mental health app, ‘oh they must be having a relapse of
low mood’, when actually it's a sign of cancer, a sign of
something else that needs checking out….
(Participant 11)
Theme 2: Barriers to Using Passively Collected Data From Digital
Devices in Clinical Practice
A prominent theme was barriers to using passive sensing in
particular in mental healthcare. Participants highlighted sepa-
rate barriers for both service users and clinicians and therefore
we present these data separately. Participants did not identify
machine learning as a significant barrier in clinical practice.
This may be because they are unfamiliar with the difference
between data capture and data use. Participant responses were
more focused on the application of digital devices and passing
sensing for managing service users' mental health and service
provision, and less focused on machine learning, which
involves developing algorithms and statistical models for com-
plex tasks without explicit instructions.
Subtheme 2.1: Barriers for Service Users
Clinicians identified several perceived barriers to service utilising
digital devices for passively collected data. These included limited
access to digital devices, challenges in accessibility of data and
feedback, and concerns about the impact of passive sensing tech-
nologies and data on wellbeing and symptoms of psychosis.
Access and Accessibility
Participants noted that service users may not have access to digital
devices and if they were given access, there could be financial
implications, such as charging the device or incurring data net-
workcoststoaccesstheinternet.Participantswonderedifusing
digital devices to collect data passively to provide emotional and
behavioural insights would actually be affordable for service users:
Some people don't have the resources, they don't have the
money for internet connection, they don't have the money
for a laptop or smartphone or iPad, whatever it is.
(Participant 14)
Participants also described their concern around providing
service users with a digital device in the event they give the
device away or lose or sell the device due to current cost of
living demands or other pressures in their life that may lead to
such behaviours:
…or even if they lost it. I shouldn't have said maybe ‘sell’
that was wrong of me, but you know what with the
financial crisis that we're all heading towards apparently
erm you know…you never know, do you?
(Participant 10)
Concerns around the stigma of owning and using a digital
device to monitor one's mental health were also raised. Parti-
cipants said that if devices were given to service users, they
would have to be relatively ‘sleek’,‘unobtrusive’and not obvi-
ous that they were for a mental health problem to avoid stig-
matising or revealing that the individual potentially has a
mental health problem that requires monitoring:
…if people in the public cottoned on that people who
generally were wearing this kind of device, if it was one
that was initiated by the NHS, and could go ‘that means
that you're being monitored for a mental health medi-
cation’(….) erm you would have to have it as an
incredibly generic device to avoid people just becoming
aware of, ‘oh yeah, you must be ill’.
(Participant 2)
Participants further wondered about accessibility issues, such as
eyesight, learning disability and dexterity, which may impact a
service user's ability to engage with the technology and use it for
its intended purpose. Age was also discussed, with participants
wondering if older service users may struggle to use digital
devices because they were less experienced with digital tech-
nology and the training needed might potentially be prohibitive:
I fear I'm being a bit stereotypical, but maybe…it's a big
change for maybe an older adult population who never
grew up with this kind of technology.
(Participant 11)
Indeed, it was felt generally that service users who have had less
experience with digital devices may lack digital literacy and
thus experience barriers to engagement:
If people have never used it, automatically they feel on the
back foot because they haven't got that literacy.
(Participant 14)
On the other hand, participants recognised that introducing
digital devices to less experienced service users could provide an
opportunity to teach new skills and upskill service users:
…we should be skilling our patients up. So, if they don't
know, let's use it as an opportunity to teach them, to
show them….
(Participant 4)
9of16
The Effect of Receiving Digitally Delivered Emotional and Be-
havioural Insights on Wellbeing and Psychosis
Participants highlighted the potential impact of receiving
emotional or behavioural insights from passively collected
data on a service user's wellbeing and experience of psycho-
sis, especially if they had been relatively inactive when the
data was being collected. For example, if a service user
received feedback about their inactivity, participants were
concerned that this may further demotivate the service user
and reinforce this inactivity rather than serve as a motivator
to change:
…for some people it could be really demotivating and
possibly negative when they're not achieving what they
feel they should be achieving.
(Participant 12)
It was suggested that an increased focus on data itself could
become a source of anxiety and create a fear of relapse, causing
service users to over‐or misinterpret a natural ebb and flow in
mood as a sign of an impending relapse:
…. if you've got somebody who is really health conscious
and who is really frightened of becoming very unwell
again in the future, they can be maybe over monitoring
themselves. So, they're taking normal fluctuations of
mood for example as that's an early warning sign,
something must be happening.
(Participant 14)
Indeed, finding the right balance between engaging with the
data and becoming hypervigilant about it can be challenging
and difficult to achieve:
…you're between a rock and a hard place, aren't you?
Because you don't want people to be looking at it all the
time and becoming obsessed with it and driven by it erm
to completely ignoring it and not taking any notice of it.
(Participant 12)
Participants also shared their concerns for individuals experi-
encing paranoia, noting that devices that ‘track’behaviour and
provide feedback might be perceived as threatening and cause
distress:
Someone with psychosis who maybe has some kind of
paranoid thoughts or beliefs about people erm I don't
think it would help to have this kind of tracking and
know all these people know about them.
(Participant 11)
Subtheme 2.2: Barriers for Staff
Infrastructure and the impact of passively collected data from
digital devices on workload were identified as barriers to the
implementation of digital technologies and systems into clinical
practice.
Lack of Supportive Infrastructure
Participants described how it would be challenging if guid-
ance regarding digital technology and associated systems
were provided without consideration of how use might differ
between clinical teams or clinical settings. For example, the
integration of passively collected data, which is then used to
develop machine learning algorithms about an individual's
mental state, might be applied differently in inpatient set-
tings compared to community settings. Therefore, these sys-
tems need to be personalised rather than adopting a one‐size‐
fits‐all approach:
I'm a bit of a negative nelly and I think it's just working
for the NHS it's the experience of anything you roll out,
it's just one size. It's never catered to individual teams you
know that we…that we work with. It's sort of just like roll
it all out and make it fit.
(Participant 10)
Participants also highlighted that NHS IT systems are disap-
pointingly ‘slow’and outdated, which may hinder or indeed
prevent digital transformation. They also noted that the ratio of
computers to clinicians is insufficient, potentially limiting
access to the technology needed to use a digital health system in
clinical practice:
…. at work we've got this RiO, we've got the systems and
it's so slow compared to what it could be…you know it's a
bit disappointing the systems in the NHS for me.
(Participant 3)
There's four or five desktops, but there's actually more
staff than that so if we only loaded it onto a desktop
device and they had to login…all the desktops we have
are really slow erm it's really tricky getting erm that kind
of access.
(Participant 4)
Concerns were also raised about device maintenance and en-
suring they continue to work, as well as the insufficient access
to IT support if a device failed and needed repair:
IT within the NHS is dreadful erm without being too
unkind (…) erm but the level of IT that's required versus
what we have is such a disparity er I can see that
something like this…we would end up with a lot of faulty
devices that don't work right or aren't recording or are
recording wrong and no one who knows how to fix them.
(Participant 2)
Impact on Workload
Participants expressed concerns that using digital devices in
practice could increase their workload, including tasks such as
setting up devices, monitoring data and reviewing it with ser-
vice users. As a result, staff may be resistant to adapting their
practice to implement a digital care pathway:
10 of 16 Health Expectations, 2024
Erm but you always get people in teams, don't you? That
are resistant to that change, resistant to doing that little
bit extra.
(Participant 6)
Participants felt that current demands, staff shortages, feeling
under pressure and the overwhelming service demands would
create even higher workloads, which are already unmanageable.
These factors would likely prohibit the integration of digital
approaches in clinical practice:
…erm we're short staffed, we're under pressure anyway.
So, we're a bit in the negative zone at the minute, it's like
well, when the hell am I going to do that? When can I
do that?
(Participant 10)
Concerns were also raised around data flows and how clinicians
would be alerted by an algorithm, for instance, notifying them of a
service user's worsening mental state. This is particularly chal-
lenging when clinicians are already overwhelmed with multiple
calls and emails related to user's clinical needs and care. Partici-
pants were curious as to whether clinicians would be required to
act on alerts immediately and considered the impact this would
have on their workload and clinical priorities:
…how the team receives that information and whether
erm you know if it requires someone to act on that
information like straight away or in the team …and the
kind of the impact that might have on someone's
workload….
(Participant 1)
Participants were also concerned about their responsibility
to act, respond or defend themselves if something in the
data flow indicated risk, especially if they did not have
sufficient time to review the data flow and a risk indicator
wasmissing.Participantsfeltthatifthisweretoincrease
their workload, the outcome would need to justify the time
invested:
…there was an indicator that would concern you…if you
haven't seen that and then something happened, how would
that stand up in a coroner's court?
(Participant 14)
And definitely that it's evident that the pros of getting this
system erm are bigger than the cost to clinicians in terms
of time and effort, because they already have little time to
do anything else.
(Participant 11)
Theme 3: Facilitators to Using Passively Collected Data From
Digital Devices in Clinical Practice
Participants discussed the factors that might encourage both
service users and clinicians to engage with digital devices and
passively sensed data in mental healthcare.
Subtheme 3.1: Facilitators for Service Users
Support From Staff
Staff support was identified as essential for facilitating service
user engagement with digital approaches to mental healthcare.
Participants felt that, as clinicians, they would play a key role in
this process. They highlighted that a trusting relationship and a
strong therapeutic alliance between the service user and clini-
cian would be crucial to encourage buy‐in:
If you can develop that therapeutic alliance with the patient
then…and if they've already seen some benefits of having
involvement from the team (…) then they might be more open
to trying novel types of technology.
(Participant 13)
Participants felt it was important for clinicians to ensure that
service users have control over how they engage with digital
approaches to mental healthcare. They emphasised that the
process should be collaborative, promoting autonomy in their
healthcare decisions:
…the collaboration around what this is for, we're not
doing this to you. So, checking that that's kind of agreed
(…) rather than, ‘we want you to do this’, and really kind
of checking out that it is collaborative.
(Participant 12)
Having made this agreement with service users, documenting this
in their care plan was viewed as further facilitating engagement:
As much as care plans are difficult to enforce, if a patient
knows that part of the plan is that they wear this thing or
they do this thing, they're often better at getting involved.
(Participant 2)
Subtheme 3.2: Facilitators for Staff
Access to Training
All participants highlighted the importance of high‐quality
training. They suggested that training should help clinicians
understand how to use digital devices, navigate associated sys-
tems, interpret data flows and integrate these into their practice:
Clinicians that are using it would need to understand the
system, what the different options within the system were,
how to get it all set up. I feel like there's obviously some
level of training package needed.
(Participant 8)
Participants felt that presenting clinicians with an evidence base
supporting digital approaches to mental healthcare could en-
hance engagement. This might help clinicians understand the
rationale behind using digital technology in mental healthcare:
You could say well you know look there's evidence here
that it really helps, that it's beneficial, there's been a study
11 of 16
done and if you can provide evidence to say that it's…it
has been helpful, I think erm team members would
maybe be more supportive of it.
(Participant 9)
Furthermore, allowing clinicians to try the technology firsthand
would enhance their understanding of a digital system, boost
their confidence and skills, and enable them to effectively share
this knowledge with service users:
It's always down to the training for me. So, the way
around that for me would be to trial staff with it so they
would have…they would know how it works because
they're using it.
(Participant 3)
Ease of Use
Participants noted that for digital approaches to be effectively
integrated into clinical practice, they must be user‐friendly,
simple, efficient and straightforward. This ensures they do not
disrupt workflows or take time away from other duties and
responsibilities:
If it was something that just kind of worked very quickly
and was minimal input from us, but if you kind of have
to set something up for every service user and then you
know, that was very time consuming I think that would
be a problem.
(Participant 7)
Furthermore, clear and detailed guidelines and procedures
must be established for clinicians. This will ensure clinicians
know how to manage alerts and notifications about a service
user's health and the appropriate follow‐up required:
So, if certain alerts come through, this is what you would
do with it, you know like SOPS. You know you'd have to…
you'd have to go through everything.
(Participant 3)
Feeling Heard and Supported
Participants felt that innovations such as digital approaches in
healthcare often follow a top‐down process, without consulta-
tion with frontline staff or consideration of clinicians' views.
There was a sense that this approach can be disappointing,
disheartening and dismissive, ultimately devaluing the potential
impact of such innovations on clinical practice:
Well, we never are consulted with anything. It all gets
done upstairs and fed back down and…so, yeah, it's a bit
just kind of this is how it is, and we have to get on with it.
(Participant 10)
Participants highlighted their desire to be meaningfully con-
sulted during the development process and before implemen-
tation. Involving clinicians in this way would facilitate their
engagement with digital technologies and systems, as it would
allow them to inform and shape the system and feel that their
input is valued rather than tokenistic:
I would say in terms of consultation and getting advice
from teams before its even started and not the kind of
half, ‘oh yeah, we really value your opinions’, when
actually you know decisions are already made. So, proper
meaningful engagement with patients and staff teams
about how it's going to be done.
(Participant 11)
Participants also wanted opportunities to provide feedback on
digital systems after their integration into clinical practice,
perhaps through a built‐in mechanism. This would allow cli-
nicians to report if tools are not functioning as intended or
suggest adaptations to make them more useful, which in turn
might increase their willingness to engage with digital tech-
nologies and systems:
I do find a lot of stuff is written without actually checking
with the clinical staff and it's given to us as a kind of ‘here
you go’and we're like well that's useless, but we can tell
you how to adapt it, (…) and actually having a direct line
of communication. I would be a bit worried about you
kind of going ‘well here you are, here's something, just get
on with it….
(Participant 4)
4 | Discussion
We investigated healthcare professionals' views and attitudes
towards the use of digital devices, passive sensing technology
and machine learning algorithms as applied in mental health-
care and the barriers and facilitators to incorporating digital
healthcare systems into routine clinical practice. Three main
themes were developed from the data: (1) positives and nega-
tives of digital devices, passive sensing and machine learning
methods in mental healthcare; (2) barriers and (3) facilitators to
using passively collected data from digital devices in clinical
practice.
Participants discussed that passive sensing technology could
empower service users by giving them more control and
responsibility for managing their daily mental health. Feeling
empowered in this way could positively impact both service
users and the broader mental healthcare system, as it may
motivate service users to adopt healthier lifestyles and enhance
health literacy. Indeed, an evaluation of the impact of digital
inclusion in healthcare in the United Kingdom found that users
who visited their GP less, had more confidence in using online
health information and wider benefits were predicted for
healthcare services and society as a whole [35]. Conversely,
concerns were raised that delegating aspects of clinicians' roles
to digital devices might make service users feel abandoned by
technology, potentially harming the therapeutic relationship
[36]. This issue is especially relevant for older people, those
lacking technological literacy and access to digital devices as
digital approaches to mental healthcare may marginalise these
12 of 16 Health Expectations, 2024
service users and exacerbate disparities [37]. To avoid digital
exclusion, an evidence‐based digital inclusion strategy may be
needed within mental health services [38]. Combining digital
approaches with face‐to‐face interactions can prevent service
users from feeling abandoned [27].
Clinicians generally felt that passive sensing technology allows
for the unobtrusive collection of a significant amount of data
without burdening service users. This is in line with previous
research that has found passive sensing technology is less dis-
ruptive than keeping diaries or completing questionnaires [39].
Participants echoed concerns that have been raised previously
around passive sensing data [40]. However, research suggests
that using objective data alongside self‐report measures can
improve the accuracy of detecting mental health difficulties, as
self‐report can be biased for various reasons; combining them
with objective data provides a fuller understanding of a person's
challenges [41]. Furthermore, participants spoke positively
about service users and clinicians receiving alerts regarding
behaviour change, as this could allow earlier access to support,
something previous research has also highlighted as valuable
[24]. This approach is particularly useful for individuals who
experience psychosis, as it can help reduce the likelihood of
relapse and hospitalisation. This not only benefits the individual
but also decreases demand and costs for services. Indeed,
research shows that passive data is a better predictor of relapse
than survey data alone [42]. However, participants expressed
some discomfort around the collection of this data. There was a
sense that passive sensing technology has the potential to
invade the privacy of service users, as they will have less per-
ceived control over the information they share with a health-
care provider. Concerns were also highlighted that those who
experience psychosis may be prone to delusions and paranoia.
Thus, a device that ‘tracks’their behaviour might be seen as
threatening [43].
Consent emerged as a prominent issue in the interviews. Par-
ticipants highlighted the need for service users to fully under-
stand what they are agreeing and consenting to and noted that
this process must be ongoing due to potential changes in
capacity over time. Research has shown that service users are
generally comfortable sharing passively collected data with
their clinician, but would not be comfortable sharing this data
more widely (e.g. on social media sites) [14]. Therefore, as part
of the consent process, the topic of data privacy and data pro-
tection is important to discuss, ensuring that service users know
where their data goes, how it will be stored, who will have
access to it and how long it will be stored for. Participants
emphasised the need to offer choice and address barriers when
obtaining consent, as some individuals may feel more confident
with digital approaches than others. Fostering good relation-
ships with service users, providing relevant information and
clear reasons for change, and ultimately allowing service users
to decide what is best for them are effective ways to encourage
service user autonomy in their mental healthcare. This is in line
with self‐determination theory, whereby supporting service
users' autonomy, competence and relatedness is important to
enhance adherence and health outcomes [16].
In relation to clinicians utilising digital devices that passively
collect data and machine learning algorithms in practice, concerns
were raised about the impact on clinician's workloads, with other
studies highlighting the importance of this issue [18]. Participants
felt that if digital devices and associated systems required extensive
setup, monitoring and administrative work, along with receiving
spontaneous alerts throughout the day, this would become
unmanageable. Participants further shared a sense of scepticism
that they would be given sufficient access to the technology and
expressed that current systems would not be fit‐for‐purpose. En-
suring access to, and ease of use of, technology and systems will
therefore be important, otherwise, this is likely to result in frus-
tration and reluctance to engage [44]. Training is also important as
some participants noted that they were older, lacked digital liter-
acy and/or struggled with change. Indeed, digital technology and
associated systems will only be useful if clinicians understand
them and feel comfortable using them [23]. Although training was
viewed as an additional time burden, participants highlighted
ways to make it more effective. They suggested that staff try the
technology themselves to gain firsthand knowledge to better
support service users. Educating staff on the evidence base was
also considered beneficial for understanding the advantages for
practice and service users. Generally, it is important for training to
build confidence and trust in systems [27].
In this study, there was an overarching sense that digital devices
that enable passive sensing might result in new ways of doing
things for both healthcare professionals and service users [45].
This was seen positively, as it recognises that digital technology
is now integral to daily life and represents a step towards
mental health services ‘catching up’with physical health ser-
vices. Further, participants identified ways in which they could
use digital technology in practice to complement what they do
or try new ways of working. Concerns appeared to be around
uncertainty [40], with participants highlighting the importance
of standard operating procedures, particularly around mana-
ging risk and data privacy [46]. Overall, healthcare professionals
were open to any innovations that could improve the lives of
service users. However, participants highlighted the importance
of not abandoning service users to technology, and ensuring
clinicians are not left to navigate these systems without support.
Involving staff at every stage of development and deployment is
crucial [23]; designing tools and systems with clinicians, rather
than for them, helps humanise digital mental healthcare [45].
The study sample was heterogeneous, including staff of differ-
ent ages and professional backgrounds, which allowed for a
variety of views and attitudes to be captured. As staff worked
across two large NHS trusts with different populations,
including older adult populations, this allowed for a broad ex-
ploration of views and novel insights. However, we acknowl-
edge potential bias in this study as staff who chose to take part
may have held particularly strong views about the topic. In
addition, participants' responses were specific to the clinical
population they work with and the area in which they work
(Northwest of England), and views/attitudes may be different
elsewhere. However, participants were encouraged to consider
the topic generally across mental healthcare. Although our
study design, protocol and topic guide were informed by a lived
experience community group that advises on research projects
at our University, it might have been useful to consult multi-
disciplinary staff members during the development of the topic
guide and to conduct a pilot interview to identify limitations.
13 of 16
Rather, the topic guide in particular was refined during the
research process to ensure sufficient exploration of views. For
example, the topic of consent was a salient point raised during
all early interviews. Prompts and follow‐up questions were
added to facilitate further exploration of this area and capture
additional views, such as if/why consent to the collection of
passive sensing data might feel different to the collection of data
that is actively collected.
During the interviews, clinicians primarily reflected on what
mental health service users might perceive in using digital devices
that passively collect data emotional and behavioural data. Whilst
our intention was to also explore the use of machine learning
algorithms in mental healthcare, participants reflected on this less,
perhaps because they are unfamiliar with this approach and its
current application in service provision. Research exploring service
users' views directly is now needed to gain further insight. Fur-
thermore, the views gathered in this study are hypothetical,
reflecting attitudes before using digital systems and technologies in
mental healthcare. Future research should explore actual accept-
ability, specifically clinicians' views after they have experience
using these approaches in their practice [47].
Smartphone ownership among service users is high, making
widespread roll‐out feasible. For example, a recent survey in the
United Kingdom by members of our group found that 91.6% of
psychosis service users own a smartphone. However, while
wearable ownership is increasing, it remains lower (30.1%; Zhang
et al., in press). Consideration must be given to ensuring people
have the necessary equipment and engage with the technology
long enough to collect sufficient longitudinal data for machine
learning to develop accurate algorithms. Participants identified a
number of potential benefits to using digital systems and tech-
nologies in mental healthcare. Staff noted that many people use
devices such as smartphones and smartwatches in their day‐to‐day
lives, and it makes sense to incorporate data generated from digital
devicesintopractice.Varioususesofthedatawereidentified,
including as a means of reflection, education, prompting conver-
sations, and monitoring medication side effects. Furthermore,
participants noted how serviceuserscouldusethesedevices
themselves to engage in healthier lifestyles, set health‐related
goals, monitor and manage their mental health and set up support
groups. Real‐time prompts to engage in coping strategies and
highlight the risk of relapse were considered particularly useful for
empowering service users, leading to broader positive implications
for services.
5 | Conclusions
As summarised in Table 3, our findings show the importance of
consulting staff throughout the entire process, from design to
TABLE 3 | Recommendations based on study findings.
Impact on the therapeutic relationship
Start with a blended approach, combining face‐to‐face contact with digital device usage. This can
help reduce feelings of abandonment and provide opportunities to address problems together.
Consider ways to reconcile discrepancies between self‐reported and objective data to preserve the
value of personal experience and the therapeutic relationship.
Impact on staff Involve staff at every stage, from development to deployment. Providing comprehensive training
that addresses concerns is essential. Ongoing and accessible support is crucial, especially for those
with low digital literacy.
Allowing staff to test the technology and review evidence of its practical benefits will enhance
their engagement.
To ensure staff are well‐prepared, standard operating procedures are needed and must be clear,
especially for managing risk issues.
To minimise significant increases in workload, systems should be fit‐for‐purpose and have
efficient data flows.
To ensure the safety and security of service user data, digital devices and systems must comply
with stringent data management procedures that are clearly defined to help staff feel confident
when obtaining informed consent.
Impact on service users
Ensure devices are user‐friendly and provide information in accessible formats, such as audio and
video, not just text.
Regularly review service users' experiences with the devices to evaluate their impact on wellbeing
and address any issues/difficulties that may arise. This also provides an opportunity to re‐evaluate
consent if circumstances change.
Equity Make plans for those who choose not to use digital tools to ensure equitable healthcare.
To address digital exclusion for service users, consider loaning phones or wearable devices to
those without access. Develop plans for managing lost, stolen or broken devices, and inform
service users about any associated costs, such as charging and using Wi‐Fi usage.
14 of 16 Health Expectations, 2024
implementation, and providing opportunities for feedback once
devices and systems are introduced. Training is crucial, with
several suggestions made to ensure it is engaging and mean-
ingful for staff. Staff reflected on past experiences with new
systems, recalling feelings of overwhelm and neglect, high-
lighting the need for sufficient training and ongoing support to
be in place. Furthermore, systems must be fit‐for‐purpose and
must not introduce new or additional barriers. Digital systems
must also adhere to strict data management procedures to en-
sure data security and safety, with clear guidelines to help staff
feel comfortable and confident when obtaining informed con-
sent from service users.
Author Contributions
Jessica Rogan: methodology, data curation, investigation, validation,
formal analysis, project administration, writing–original draft, writing–
review and editing. Joseph Firth: methodology, supervision, writing–
review and editing. Sandra Bucci: conceptualisation, methodology,
investigation, validation, supervision, funding acquisition, writing–
review and editing.
Acknowledgements
We are grateful for the staff who took part in the study. This study was
supported by the NIHR Manchester Biomedical Research Centre (NIHR
203308).
S.B. discloses the support for the publication of this work from a
National Institute for Health and Care Research research professorship
(NIHR300794).
Disclosure
The views expressed are those of the author(s) and not necessarily those
of the NIHR or the Department of Health and Social Care.
Conflicts of Interest
S.B. is a director and shareholder of CareLoop Health Ltd, a University
of Manchester start‐up to develop and market digital solutions for
mental health problems, currently in schizophrenia and postnatal
depression.
Data Availability Statement
The data that support the findings of this study are available on request
from the corresponding author. The data are not publicly available due
to privacy or ethical restrictions.
References
1. Our World in Data, https://ourworldindata.org/mental-health.
2. NHS. Implementing the Early Intervention in Psychosis Access and
Waiting Time Standard: Guidance. Published Online 2016.
3. S. Allan, S. Bradstreet, H. J. McLeod, et al., “Perspectives of Patients,
Carers and Mental Health Staff on Early Warning Signs of Relapse
in Psychosis: A Qualitative Investigation,”BJPsych Open 6, no. 1
(2020): e3.
4. M. Sheikh, M. Qassem, and P. A. Kyriacou, “Wearable, Environ-
mental, and Smartphone‐Based Passive Sensing for Mental Health
Monitoring,”Frontiers in Digital Health 3 (2021): 662811.
5. “World Mental Health Report,”accessed November 3, 2024, https://
www.who.int/publications/i/item/9789240049338.
6. J. Torous, S. Bucci, I. H. Bell, et al., “The Growing Field of Digital Psy-
chiatry: Current Evidence and the Future of Apps, Social Media, Chatbots,
and Virtual Reality,”World Psychiatry 20, no. 3 (2021): 318–335.
7. C. Hollis, S. Sampson, L. Simons, et al., “Identifying Research Pri-
orities for Digital Technology in Mental Health Care: Results of the
James Lind Alliance Priority Setting Partnership,”Lancet Psychiatry 5,
no. 10 (2018): 845–854.
8. “A Plan for Digital Health and Social Care,”accessed November 3,
2024, https://www.gov.uk/government/publications/a-plan-for-digital-
health-and-social-care/a-plan-for-digital-health-and-social-care.
9. E. Garcia‐Ceja, M. Riegler, T. Nordgreen, P. Jakobsen,
K. J. Oedegaard, and J. Tørresen, “Mental Health Monitoring With
Multimodal Sensing and Machine Learning: A Survey,”Pervasive and
Mobile Computing 51 (2018): 1–26.
10. D. C. Mohr, M. Zhang, and S. M. Schueller, “Personal Sensing:
Understanding Mental Health Using Ubiquitous Sensors and
Machine Learning,”Annual Review of Clinical Psychology 13, no. 1
(2017): 23–47.
11. P. A. Tiffin and L. W. Paton, “Rise of the Machines? Machine
Learning Approaches and Mental Health: Opportunities and Chal-
lenges,”The British Journal of Psychiatry 213, no. 3 (2018): 509–510.
12. C. Hollis, R. Morriss, J. Martin, et al., “Technological Innovations in
Mental Healthcare: Harnessing the Digital Revolution,”British Journal
of Psychiatry 206, no. 4 (2015): 263–265.
13. A. Thieme, D. Belgrave, A. Sano, and G. Doherty, “Machine
Learning Applications: Reflections on Mental Health Assessment and
Ethics,”Interactions 27, no. 2 (2020): 6–7.
14. A. Trifan, M. Oliveira, and J. L. Oliveira, “Passive Sensing of Health
Outcomes Through Smartphones: Systematic Review of Current Solu-
tions and Possible Limitations,”JMIR mHealth and uHealth 7, no. 8
(2019): e12649.
15. S. Bucci, C. Barrowclough, J. Ainsworth, et al., “Actissist: Proof‐of‐
Concept Trial of a Theory‐Driven Digital Intervention for Psychosis,”
Schizophrenia Bulletin 44 (2018): sby032–sby032, https://doi.org/10.
1093/schbul/sby032.
16. R. M. Ryan and E. L. Deci, “Self‐Determination Theory and the
Facilitation of Intrinsic Motivation, Social Development, and Well‐
Being,”American Psychologist 55, no. 1 (2000): 68–78.
17. R. M. Ryan, H. Patrick, E. L. Deci, and G. C. Williams, “Facilitating
Health Behaviour Change and Its Maintenance: Interventions Based on
Self‐Determination Theory,”European Health Psychologist 10, no. 1
(2008): 2–5.
18. V. De Angel, S. Lewis, K. M. White, F. Matcham, and M. Hotopf,
“Clinical Targets and Attitudes Toward Implementing Digital Health
Tools for Remote Measurement in Treatment for Depression: Focus
Groups With Patients and Clinicians,”JMIR Mental Health 9, no. 8
(2022): e38934.
19. S. Bucci, R. Morris, K. Berry, et al., “Early Psychosis Service User
Views on Digital Technology: Qualitative Analysis,”JMIR Mental
Health 5, no. 4 (2018): e10091.
20. W. Guo, T. J. Key, and G. K. Reeves, “Accelerometer Compared
With Questionnaire Measures of Physical Activity in Relation to Body
Size and Composition: A Large Cross‐Sectional Analysis of UK Bio-
bank,”BMJ Open 9, no. 1 (2019): e024206.
21. C. Blease, C. Locher, M. Leon‐Carlyle, and M. Doraiswamy, “Arti-
ficial Intelligence and the Future of Psychiatry: Qualitative Findings
From a Global Physician Survey,”Digital Health 6 (2020):
2055207620968355.
22. M. I. Thenral and A. Annamalai, “Challenges of Building, Deploy-
ing, and Using AI‐Enabled Telepsychiatry Platforms for Clinical Prac-
tice Among Urban Indians: A Qualitative Study,”Indian Journal of
Psychological Medicine 43, no. 4 (2021): 336–342.
15 of 16
23. E. Rodriguez‐Villa, A. R. Rozatkar, M. Kumar, et al., “Cross Cultural
and Global Uses of a Digital Mental Health App: Results of Focus
Groups With Clinicians, Patients and Family Members in India and the
United States,”Global Mental Health 8 (2021): e30.
24. A. Ng, R. Kornfield, S. M. Schueller, A. K. Zalta, M. Brennan, and
M. Reddy, “Provider Perspectives on Integrating Sensor‐Captured
Patient‐Generated Data in Mental Health Care,”Proceedings of the
ACM on Human‐Computer Interaction 3, CSCW (2019): 1–25.
25. S. Byrne, A. Tohamy, B. Kotze, et al., “Using a Mobile Health Device
to Monitor Physiological Stress for Serious Mental Illness: A Qualitative
Analysis of Patient and Clinician‐Related Acceptability,”Psychiatric
Rehabilitation Journal 45, no. 3 (2022): 219–225.
26. E. Eisner, S. Faulkner, S. Allan, et al., “Barriers and Facilitators of
User Engagement With Digital Mental Health Interventions for People
With Psychosis or Bipolar Disorder: Systematic Review and Best‐Fit
Framework Synthesis,”JMIR Mental Health. Published Online (2024).
27. S. Bucci, N. Berry, R. Morris, et al., “‘They Are Not Hard‐to‐Reach
Clients. We Have Just Got Hard‐To‐Reach Services.’Staff Views of
Digital Health Tools in Specialist Mental Health Services,”Frontiers in
Psychiatry 10 (2019): 344.
28. C. McGrath, P. J. Palmgren, and M. Liljedahl, “Twelve Tips for
Conducting Qualitative Research Interviews,”Medical Teacher 41, no. 9
(2019): 1002–1006.
29. V. Braun and V. Clarke, “One Size Fits All? What Counts as Quality
Practice in (Reflexive) Thematic Analysis?,”Qualitative Research in
Psychology 18, no. 3 (2021): 328–352.
30. P. K. Edwards, J. O'Mahoney, and S. Vincent, Studying Organiza-
tions Using Critical Realism: A Practical Guide (Oxford, UK: OUP, 2014).
31. V. Braun and V. Clarke, “(Mis) Conceptualising Themes, Thematic
Analysis, and Other Problems With Fugard and Potts' (2015) Sample‐
Size Tool for Thematic Analysis,”International Journal of Social
Research Methodology 19, no. 6 (2016): 739–743.
32. V. Braun and V. Clarke. Successful Qualitative Research: A Practical
Guide for Beginners. Published Online (2013).
33. B. Glaser, “Strategies for Qualitative Research,”The Discovery of
Grounded Theory. Published Online (1999): 2–6.
34.L.A.Primeau,“Reflections on Self in Qualitative Research: Stories of
Family,”American Journal of Occupational Therapy 57, no. 1 (2003): 9–16.
35. NHS. Why Digital Inclusion Matters to Health and Social Care.
Published Online 2022, https://digital.nhs.uk/about-nhs-digital/
corporate-information-and-documents/digital-inclusion/digital-
inclusion-in-health-and-social-care.
36. S. Chancellor and M. De Choudhury, “Methods in Predictive
Techniques for Mental Health Status on Social Media: A Critical
Review,”NPJ Digital Medicine 3, no. 1 (2020): 43.
37. A. Ho and O. Quick, “Leaving Patients to Their Own Devices? Smart
Technology, Safety and Therapeutic Relationships,”BMC Medical
Ethics 19 (2018): 18.
38. D. Robotham, S. Satkunanathan, L. Doughty, and T. Wykes, “Do We
Still Have a Digital Divide in Mental Health? A Five‐Year Survey Follow‐
Up,”Journal of Medical Internet Research 18, no. 11 (2016): e309.
39. V. P. Cornet and R. J. Holden, “Systematic Review of Smartphone‐
Based Passive Sensing for Health and Wellbeing,”Journal of Biomedical
Informatics 77 (2018): 120–132.
40. L. Balcombe and D. De Leo, “Digital Mental Health Challenges and
the Horizon Ahead for Solutions,”JMIR Mental Health 8, no. 3 (2021):
e26811.
41. S. M. Maharjan, A. Poudyal, A. van Heerden, et al., “Passive Sensing
on Mobile Devices to Improve Mental Health Services With Adolescent
and Young Mothers in Low‐Resource Settings: The Role of Families in
Feasibility and Acceptability,”BMC Medical Informatics and Decision
Making 21 (2021): 117.
42. A. Cohen, J. A. Naslund, S. Chang, et al., “Relapse Prediction in
Schizophrenia With Smartphone Digital Phenotyping During COVID‐
19: A Prospective, Three‐Site, Two‐Country, Longitudinal Study,”
Schizophrenia 9, no. 1 (2023): 6.
43. B. Greer, K. Newbery, M. Cella, and T. Wykes, “Predicting Inpatient
Aggression in Forensic Services Using Remote Monitoring Technology:
Qualitative Study of Staff Perspectives,”Journal of Medical Internet
Research 21, no. 9 (2019): e15620.
44. J. T. Kelly, K. L. Campbell, E. Gong, and P. Scuffham, “The Internet
of Things: Impact and Implications for Health Care Delivery,”Journal
of Medical Internet Research 22, no. 11 (2020): e20135.
45. B. Spadaro, N. A. Martin‐Key, and S. Bahn, “Building the Digital
Mental Health Ecosystem: Opportunities and Challenges for Mobile
Health Innovators,”Journal of Medical Internet Research 23, no. 10
(2021): e27507.
46. F. M. Dawoodbhoy, J. Delaney, P. Cecula, et al., “AI in Patient Flow:
Applications of Artificial Intelligence to Improve Patient Flow in NHS
Acute Mental Health Inpatient Units,”Heliyon 7, no. 5 (2021): e06993.
47. N. Berry, F. Lobban, R. Emsley, and S. Bucci, “Acceptability of
Interventions Delivered Online and Through Mobile Phones for People
Who Experience Severe Mental Health Problems: A Systematic
Review,”Journal of Medical Internet Research 18 (2016): e121, https://
doi.org/10.2196/jmir.5250.
Supporting Information
Additional supporting information can be found online in the
Supporting Information section.
16 of 16 Health Expectations, 2024
