Chapter 26: Technology Acceptance in Healthcare: The Role of the Campus Environment

Traditional approaches to the Technology Acceptance Model (TAM) see the characteristics of the technology itself as the main predictors of acceptance and adoption, namely, the perceived ease of use and the perceived usefulness [1]. Around the same time, the Diffusion Of Innovations (DOI) model introduced the willingness of potential adopters, the implementational context, and communication channels as significant factors beyond the technology itself [2]. Later extensions to TAM have sought to account for differences in adoption by different user types, which moderates perceived ease of use and usefulness, such as the Unified Theory of Acceptance and Use of Technology (UTAUT, [3]). For healthcare specifically, the Health Belief Model (HBM) focuses on patient perceptions alone [4]. While broader stakeholder engagement, including patients, clinicians, technologists, planners, and policymakers, is said to be required to ensure the long-term adoption and sustainability of an intervention, [5] the complexity of the technology and the process of deployment are perhaps even more significant than specific features of the technology itself. Indeed, direct collaboration and cognitive buy-in of patients and clinicians are a prerequisite to rolling out technology or other interventions within healthcare [6].

The BigMedylitics (BML) project provides a unique opportunity to explore the attitudes of those directly affected by the deployment of advanced technology in healthcare settings. In this chapter, I report on two empirical studies that aim to provide additional evidence for how advanced technologies are likely to be received by those stakeholders – patients and clinicians – directly involved in the provision and receipt of care.

Figure 26.1 illustrates the potentially disruptive effect of technology introduction into healthcare (for a more detailed discussion, see [7]). The left-hand panel shows the simple relationship between patients and clinicians which is based largely on trust. In the behavioural sciences, trust involves an acceptance of risk: the patient expects the clinician to cure them, [8,9] whilst appreciating that this may not be possible. This relationship, that is between patients and clinicians, may be guided and monitored, of course, by a relevant authority.

The studies in the BML project (Sections II–IV) sought to introduce technology in different healthcare contexts. For instance, Chapter 9 (eHealth and Telemedicine for Risk Prediction and Monitoring in Kidney Transplantation Recipients), Chapter 10 (Remote Monitoring to Improve Gestational Diabetes Care), and Chapter 11 (Monitoring Wellness in Chronic Obstructive Pulmonary Disease Using the myCOPD App) introduced Remote Patient Monitoring (RPM) solutions for self-monitoring and thereby for patients to engage with their own care. Chapter 14 (Usability of Enhanced Decision Support and Predictive Modelling in Prostate Cancer), Chapter 15 (Monitoring and Decision Support in Treatment Modalities for Lung Cancer), and Chapter 16 (Artificial Intelligence to Support Chooses in Neoadjuvant Chemotherapy in Breast Cancer Patients) provided support to clinicians during diagnosis and treatment. Finally, Chapter 19 (Implementation and Impact of AI for the Interpretation of Lung Diseases in Chest CTs) and Chapter 20 (Innovative Use of Technology for Acute Care Pathway Monitoring and Improvements) focus on hospital operations to improve efficiency.

The right-hand panel of Figure 26.1 shows the effects of introducing these technologies into a healthcare context. Now, the trust relationship that accepts and assumes an element of vulnerability, overseen by a relevant healthcare authority, is replaced on the one hand by reliance on advanced technology and on the other by an ambivalent response by the patient as a member of the public to advanced technology. First, the patients assume that the data used to build the model that the clinician is relying on are relevant to them: the sample used is representative and inclusive. Second, they are unsure about AI technologies per seⁱ [10].

In this section, we summarize the findings from an online survey targeted at the general public; as UK residents, they are UK National Health Service (NHS) patients. So, their perceptions reflect their expectations and experience with a free-at-source health service. A little more than half (53%) do not currently use an app; the others do. The survey was intended to explore the general public’s perceptions of using an app in healthcare, rather than specifics of the technology as highlighted by TAM or user demographics as suggested by the UTAUT. It does, however, seek to explore user decisions in adopting apps in much the same way as the HBM. Further, adopter willingness from the DOI theory is represented to a limited extent in trying to cover both existing app users (Early Adopters) and non-users (perhaps the Late Majority).

The HBM seeks to predict patient willingness to engage with an intervention based on the combination of the following factors:

• The perceived vulnerability to a given condition and its impact

• The cost/benefit assessment of adopting the proposed intervention

• The self-efficacy resulting from adopting the proposed intervention.

An anonymous online survey was developed based on assertions associated with the HBM factors, adopted specifically for the use of an app to monitor a health condition (see Chapter 11 – Monitoring Wellness in Chronic Obstructive Pulmonary Disease Using the myCOPD app). These were supplemented with statements derived from a Patient and Public Involvement and Engagement (PPIE) discussion with patients who have a COPD diagnosis. Specifically, reports that patients could feel isolated and a little ignored and the potential for a self-reporting app to provide social contact with other patients.

This resulted in a survey containing 28 items, grouped into three sections: the first dealing with perceptions of participants’ own health, the second about app usage, and the third general attitudes to healthcare. Participants were asked to respond on a four-point Likert scale (Strongly agree to Strongly disagree). Some assertions were reversed to try and avoid participants selecting the same response throughout.

A total of 400 UK residents were recruited via a crowdsourcing platform (Prolific.co) and were paid a nominal amount (£3.00).ⁱⁱ The average time taken to respond to the survey was 4 minutes and 7 seconds. Table 26.1 summarizes the demographics of the survey respondents.

The respondents correspond to current UK census data.ⁱⁱⁱ In addition, 187 respondents reported that they were regular app users, whereas 213 were not.

I focus here solely on the 13 assertions that refer directly to app usage. Using 67% – or roughly two-thirds – provides a threshold to identify significant (dis)agreement among respondents. Namely, percentages above 67% indicate significant support or rejection of an assertion. Table 26.2 summarizes those assertions where responses from both app users and non-app users coincide: Columns 2 and 3 in the table are percentages over 400.

The first two assertions relate to self-efficacy: both app users and non-users agree that healthcare app usage provides a sense of doing something positive. Even though I suggested above that an app might disrupt the healthcare ecosystem (see the righthand panel of Figure 26.1), the general public in the United Kingdom accepts that healthcare apps increase their feeling that apps give them the opportunity to do something positive and take responsibility for their own health. The last three records in the table show general disagreement with the assertions. Neither app users nor non-users believe that a healthcare app is too difficult to use nor too time-consuming. Interestingly, nor do they believe that app usage is a way to replace traditional clinician-provided healthcare: the general public does not believe they are being fobbed off and left to fend for themselves.

Table 26.3 summarizes cases where private citizens did not agree about the importance of the assertion. In all cases, with the exception of the first assertion in the table, responses for both app users and non-users are around the 50% mark: there is no clear (i.e., greater than 67%) indication that citizens either agree or disagree with the statement. Other assertions suggest that privacy and trust are not issues, nor that the potential social-interaction benefit of using a healthcare app (the last two assertions) seems to be so important.

The final app-specific assertions are shown in Table 26.4. For Using an app regularly would identify problems earlier, the app users seem to agree with the statement (70% versus 30% disagreement), whereas the non-users appear to be undecided. This suggests that the app users can see benefits to app usage once they start using them which the non-users don’t yet appreciate. For Using a healthcare app means that I can show a doctor what’s been going on for me if needed, private citizen responses are reversed: app users disagree (85%), whereas non-users agree (82%). Expectations from healthcare app usage are therefore different once apps are being used. Indeed, these two statements suggest that app usage helps the patient identify issues not that they wish to alert a clinician. It is about self-awareness rather than part of the general healthcare context, but only once users have experienced (are using) healthcare apps. This suggests that users will make their own decisions about the benefits they get rather than any particular preconceived usefulness, such as an aide-memoire between consultations (see [11] which provides some evidence of cognitive support for rheumatoid arthritis sufferers).

From the survey completed by a representative sample of the general public (in the United Kingdom), there is general agreement that healthcare app usage provides patients with a sense that they are engaged in and taking responsibility for their own health (self-efficacy). Apps are not difficult or intrusive, nor are users particularly concerned about trust and app reliability, and do not see potential social connectivity as a motivator. Importantly, though, app users differ from non-users in that adoption seems to be about making the individual aware of their health status for existing users, whereas non-users are unsure. Further, current app users do not see healthcare apps as an aid in interactions with clinicians, whereas non-users seem to believe that apps would provide useful information to the clinician. The survey sheds some light, therefore, on the complex expectations of patients when offered a healthcare app. Perceived ease of use does not appear to be an issue. Perceived usefulness depends to some extent, though, on the existing app usage experience of patients.

In the previous section, the focus was on patient perceptions of healthcare app usage. Here, I turn to consider other stakeholders within the ecosystem as represented in the BML consortium. Partners were drawn from not only different disciplines but also different roles. This would allow a practical view on the complexity of introducing advanced, AI-enabled technology into healthcare, much as set out in the NASSS framework.

A previous three-round Delphi study involving around 10 experts (12 in Round 1 and 8 in Round 3) focusing on the adoption of advanced technologies identified a number of key areas [12]. Although not specifically targeted at healthcare technology, the types of issues raised are pertinent to the domain. These were used therefore to derive 30 assertions across four areas as described in Table 26.5. 

Participants were asked to respond on a four-point Likert scale as to whether they agreed or disagreed with the statements.^iv

The survey was distributed to partners in the BML consortium. A total of 47 responses were received, but after the initial review, one had to be removed since the respondent did not rate 10 of the 30 assertions, leaving 46 responses in total.Table 26.6 summarizes how participants described themselves.^v

Tables 26.5 and 26.6 provide two sources of variability in responses. A two-way ANOVA (Category x Role) was performed on the ratings to establish whether there were any significant effects due to category or role. Category accounts for some 50% of the variance in responses (from the partial η² p-value):

F(3, 1121.97) = 373.930, p < 0.001(η²_p = 0.501)

The greatest differences in opinion, therefore, relate to the area Requirements, Design and Responsibility, Ethics and Governance, or Transparency. Further,

F(4, 103.846) = 25.962, p = 0.006(η²_p = 0.085)

In total 8.5% of the variance is attributable to the self-reported role. So, role has some effect also. The interaction between category and role—that is, how different roles in Table 26.6 respond differently to each of the categories in Table 26.5—is

not significant:

F(12, 82.854) = 1.012, p = 0.440(η²_p = 0.069)

In the following, I focus only on responses to individual assertions within each category. Pooling the responses provides an overview of the kinds of concerns that the stakeholders represented in the BML consortium perceive related to advanced technology being introduced into healthcare. Figures 26.2–26.5 summarize the responses received from BML consortium members. Each figure should be interpreted as follows: first, there is the assertion that participants were asked to rate agreement with (on a four-point Likert scale). “Strongly Agree” responses are shown to the right in pale orange, “Agree” in yellow, “Disagree” in green, and “Strongly Disagree” in blue. Overlaid is a rectangle that represents 25% agreement on the left-hand edge and 75% agreement on the right-hand edge when reading left to right. The opposite is true if reading from right to left (25% on the right-hand edge and 75% on the left-hand edge). The hashed centre line represents the 50% mark.

To illustrate, in Figure 26.2, for the assertion “There’s too much data available now for humans to be able to process and understand”, “Strongly Agree” (pale orange) and “Agree” (yellow) were selected by more than 50% of respondents: together, these two boxes exceed the hashed line in the middle. For the assertion, advanced technologies help people do their jobs better, the two “agree” boxes (pale orange and yellow) exceed the 75% mark: so over 75% of respondents agreed with the statement. Any boldface black assertions are those where more than 75% of respondents agreed with it; items in boldface red (see Figure 26.4), then more than 75% of respondents disagreed with the assertion.

From Figure 26.2, the response to the assertion I trust the person I get to talk with understands the technology they’re using suggests that those using the output from advanced technologies may not always understand that technology. This may reflect that these stakeholders do not expect clinicians, for instance, to be able to understand the technology they rely on. Other than that, especially given the number of cases where there is overwhelming agreement (those items in bold), it indicates the perception that advanced technologies—however, they are defined— are seen to be essential for the future.

From Figure 26.3, taking both the assertion that stakeholders do not agree with as well as those with overwhelming agreement, it is clear that partners in BML believe that responsibility with advanced technologies is shared across many stakeholders. This includes those developing the technologies, such as data scientists, those reliant on them, such as clinicians, and those affected by them, the patients. Whereas previously, technology would simply be delivered and expected to fulfil its function, advanced technologies require ongoing collaboration from all those stakeholders.

Turning to Figure 26.4, it is clear that a new type of ethics is required (as highlighted in [12]). Moreover, the definition and oversight associated with such governance is the responsibility of multiple stakeholders: different agents within the healthcare context must collaborate for effective governance once the ethical framework has been established.

Finally, as regards transparency (Figure 26.5), there is general agreement that understanding how a decision is reached – such as a prediction in many of the studies in BML – is not only desirable per se but would potentially lead to greater insights and advances. This is not solely about ensuring defect-free operation (respondents disagreed with the assertion “Understanding how a technology works means we can make sure a problem doesn’t occur”) but also about deriving additional benefits.

In addition to the ratings summarized in the figures here, some participants provided free-form comments as follows:

• “Technicians and “other people” need to find or develop a common language to be able to discuss the pros and cons of AI”. This highlights the need for different disciplines to collaborate on the basis of a shared understanding or a common language.

• “Advancing technology may become a new field in which multi-disciplines work together”. Taking the perspective of collaboration from the previous comment forward, this recognizes the importance of projects like BML to encourage cross-disciplinary work, and, of course, to share experience.

• “We need to start viewing the world as a socio-technical system where humans and technologies are networked together and inseparable [from] each other”. This comment highlights the complexity of the ecosystem around and dependent on advanced technologies. It is essential (as highlighted in the survey responses themselves) to rethink how all actors and stakeholders need to be and can be involved or at least be appropriately represented.

The BML project survey on advanced technologies in healthcare derives from the practical experience of the project partners but has thrown up some common themes recognized in the area of responsible AI (see [12], but also the DARPA project and related work, e.g., [13], but also [14]). In the context of technology acceptance, the survey outlines the concerns of the main stakeholders. As such, it not only provides further material for the definitions of complexity in adoption models like NASSS but also highlights some of the communication problems – like a common language, and the need for multi-disciplinary approaches – identified in DOI. The BML consortium has therefore provided a significant use case for considering how technology can and should be introduced into healthcare.

This leads to a set of 18 recommendations regarding the introduction of advanced technologies into healthcare. The first set of five relates to the ethical treatment of all those affected by the technology.

1. There is a clear need to engage in further discussion about the ethical running of advanced technologies in healthcare.

2. Advanced technology testing should include an ethics audit along with standard testing.

3. Advanced technology deployment and operation needs to plan for and resource ongoing monitoring, especially regarding the ethical treatment of those affected by the technology.

4. An acceptability framework should be developed, including checks for bias and how to remedy it.

5. Advanced technologies require a new set of ethical norms, developed in consultation and by agreement with all relevant actors and stakeholders.

   Note that the general public did not express any concerns with privacy (often conflated with ethics) or trust in the technology and how it may work for them as individuals. The next seven recommendations relate to the involvement of all stakeholders and facilitating communication between them.

6. Multi-stakeholder involvement is essential to ensure that all perspectives are understood and can be factored into the exploitation of advanced technologies.

7. Where advanced technologies are to be deployed, all main actors (those directly involved with the technology) and all other stakeholders (those affected by the technology) should be consulted.

8. All actors (and stakeholders) need some visibility and oversight of advanced technology deployment; it is not enough to have a separate certification authority.

9. All those affected by technology need to be considered and to be engaged, including assuming responsibility for technology once a choice to use it has been made.

10. There needs to be a code of conduct established for all relevant stakeholders.

11. Technology in healthcare is one component within a broader, complex network. Focus needs to be given to how technology can affect or even disrupt existing relationships.

12. To support and facilitate cross-disciplinary understanding, there needs to be a common language in place so that all relevant stakeholders can engage and contribute.

    This is entirely consistent with models such as DOI, NASSS, and NPT, especially in recognition of the complex ecosystem that healthcare relies on. The online survey with the general public showed some discrepancies in expectations around what an app could provide. This is precisely the sort of area in which good communication and a person-centred approach to design are important. Indeed, the next two recommendations cover design:

13. There needs to be a new way of thinking around the development, testing, deployment, and ongoing monitoring of advanced (AI-enabled) technologies.

14. Advanced technologies should be designed and deployed from a human-inthe-loop perspective.

    This is precisely what the BML project sought to achieve: with prospective studies involving key stakeholders to identify the benefits of technology for the ecosystem as a whole.

15. Moving forward, there needs to be more focus on how advanced technologies will affect society and individual people.

16. It is important to consider if and how advanced technologies may affect significant current relationships.

17. How advanced technologies work needs to be understood in the context of the ecosystem where they are deployed.

18. Advanced technologies offer much potential beyond immediate needs. If they are explainable as well as functionally adequate, this will lead to greater potential benefits.

The final set of recommendations highlights the broader effects of advanced technology as it is introduced into healthcare. The BML project partners were aware of this, not least as highlighted in their comments. However, the general public is already demonstrating what they derive from using healthcare apps as reported in the online survey. PPIE-type engagement moving forward is one way to maintain oversight of how apps are used. For instance, current users identified that app usage would help them identify issues early. However, at the same time, they were less clear that they intended to use that information to share with clinicians. This might suggest that a more complex relationship is developing in the healthcare context (see, for instance, [15]).

In this chapter, I have cited empirical evidence gathered during the BML project to identify the challenges associated with introducing advanced technologies into healthcare. The focus was mainly on self-reporting apps used by patients as part of their health regime, though the survey with BML partners also provides insight into broader advanced technology issues. I started with the traditional view that for a technology to be adopted, it should be perceived as easy to use and useful. Traditional models such as TAM and even UTAUT fall into this category. However, especially in healthcare, other researchers have highlighted the complexity of healthcare intervention and technology acceptance. Innovation in general requires willing and innovative participants and appropriate communication channels (DOI). However, there is also a significant need for multi-stakeholder involvement and an appreciation of the complexity of the technology, but also the ecosystem into which it will be deployed (NASSS and NPT). Taken together, though, the surveys reported here demonstrate an awareness of these challenges as well as generate some recommendations as to how to address them. Most importantly, perhaps, is that the survey carried out with the general public seemed to suggest that healthcare app users as well as non-users are not so concerned with reliability and privacy. Instead, they want to feel able to engage in their own healthcare: perceived selfefficacy is essential, alongside all of the ethical and design issues with advanced technologies. Potential healthcare app users want to use and derive benefits from those apps. They do not simply want to maintain the existing status quo. As such, they are perfectly capable of engaging with other stakeholders: they need to be part of the conversation in terms of the NASSS framework, and they have already developed cognitive participation as described in NPT from their own understanding of healthcare apps.