Asthma self-management, air quality and technology

In the UK, 4.3 million people of all ages, are living with asthma. Asthma is a variable condition, triggered by viral infections, exercise, allergens (including pollens), air temperature, as well as a range of indoor/outdoor air pollutants [1,2]. People with asthma need to know how to reduce exposure to triggers, recognise deterioration, and adjust their treatment to respond to changes in their asthma status. They need to balance the restrictions of avoiding triggers with the benefits of being physically active and enjoying meaningful activities and lifestyle, contributing to their mental wellbeing as well as protecting lung function. Supported by their healthcare professionals (HCPs), self-management is a highly effective strategy for reducing the risk of attacks and improving control [3].

The health risks of poor air quality and its effect on asthma patients is increasingly being recognised [1,2,4]. Technology can enable informed decision-making and support patients to balance the individualised risk of trigger exposure with participating in the paid work, leisure activities and exercise that they value.

Internet-of-things (IoT) environmental sensors can detect an increase in known risk factors (e.g. environmental pollen) and, combined with other asthma self-management devices (e.g. smart spirometers and smart inhalers), can form a connected system that can provide early warning of an attack. With this prediction, the AI can provide timely advice to patients on adjusting their medication to prevent the attack progressing. Where appropriate, the system can also connect patients to their clinicians for further advice.

Digital health, including ehealth/mhealth, are increasingly seen as able to support effective self-management. Information and communication technology (ICT) models are gradually transforming from standalone systems, that provide healthcare information on a dashboard, into two-way, AI-enabled systems with multiple connected sensors to provide holistic self-management support in self-diagnosing, monitoring triggers, and attack prediction [5,6]. This digital transformation evolved rapidly during the COVID pandemic; as connected systems allowed home surveillance of COVID infections avoiding the risk of viral spread during a visit to the clinician [7]. However, the question remains how such data/AI-enabled digital health applications maybe valued by the end users (that is, the patients), and how to increase engagement and use of these technologies [8].

The role of patient involvement in digital health care research and development

Despite promotion of patient and public involvement (PPI) [9,10], in most digital intervention, patients’ involvement is typically recognised as a tester to evaluate the usability/health outcomes of a digital service, with the aim of supporting the researcher to generate evidence for further development and roll out. Patient involvement at the conceptual stage of shaping innovation is limited despite potential advantages of early involvement [11]. Firstly, providing co-design opportunities for patients to shape a technology pre-development ensures the resulting intervention is of interest and useful to them. Secondly, research participants are typically positive about ready-made interventions and studies of digital technologies rarely garner negative comments [12]. PPI can consult patients about future developments, when patients may be more willing to provide critical and innovative feedback.

Finally, PPI enables researchers to hear the patients’ voice not as research participants but as colleagues helping to shape research. Reflecting this, five of the authors on this paper are the PPI colleagues who contributed to the workshops. Despite promotion of this collaborative process [13], and recognition of the benefits, there is little published work reporting on the role of PPI in shaping the future direction of a digital intervention, and the design of future research [14].

Patient involvement in the Connected for Asthma research

The Connected for Asthma (C4A) research programme has involved patients (and other stakeholders) at every stage of the research phase, to co-create a digital intervention to support asthma self-management.

From our previous PPI discussions, we know patients and clinicians want to monitor their asthma and their triggers and want timely feedback to support their care, within a connected ‘super-app’ that can connect them with their clinicians. They believe AI could be helpful to provide advice in the connected system, though the AI has to be trustworthy and reliable in order to encourage adoption [15]. We summarised their preferred use of AI in three user cases to be discussed in the workshop.

In this paper, we report PPI advisory workshops conducted within the C4A study as an exemplar of empowering patients’ role in shaping innovation towards a digital intervention that would wish to adopt and use in real life.

Ethics statement

The University of Edinburgh Medical Research Ethics Committee confirmed that this was a consultation exercise with PPI colleagues designed to shape a research agenda, so that ethical approval and formal consent was not required (REC Reference: 22-EMREC-057). Five of the PPI colleagues are co-authors, and two chose not to be named. We followed the standards of the National Institute for Health Research (NIHR) for Patient and Public Involvement and Greenhalgh et al ‘s PPI framework to shape the work [13,16].

Workshop design

We used the key learnings from our previous studies to design the workshops [17]:

• Most patients have heard of AI but few know what it is, and the range of its capabilities.
• Patients have different ways to manage their asthma, and they develop their own self-management routine and that influences if/how they would like to use technology to look after their asthma.
• Building on existing proven technologies, and research identifying patients’ most ‘wanted’ features in related to the Al and environmental data [17], we determined three potential use-cases that may have future utility.

Aim of the workshop

To shape the technology features and research topics of the next digital intervention in the C4A portfolio. To inform which of the use-cases had utility, we aimed to consult PPI colleagues on how they would prefer to use AI including indoor/outdoor environmental data to support their self-management and more generally, the workshop aimed to explore how they PPI wish to interact with AI?

PPI colleagues’ invitations

We invited PPI colleagues from the Asthma UK Centre for Applied Research (AUKCAR) to join a workshop via a post in the Centre’s Twitter. The post contained information such as the aim, time/location of the workshop and reimbursement. Information was also sent to previous C4A study participants who had given consent to be contacted about future activities.

Procedure

To accommodate availability, we carried out four advisory workshops/discussions with seven PPI colleagues, who were adults with asthma, with variable experience of using technology to support their asthma.

In each workshop, a researcher (CYH) led the discussion supported by up to two PPI facilitators (LGR, and/or EK, and/or SE). We ran face-to-face workshops on the University premises with an online option for PPI colleagues unable to travel to the university to participate. To stimulate ideas about their self-management routine and how AI and environmental data could support their daily self-management tasks, we sent PPI colleagues a table to act as a framework to structure the thoughts and experiences we wanted to discuss in the workshop (see S1 Table, self-management table).

Each workshop ran for one hour and consisted of an introduction, two exercises to stimulate discussion and a final ‘wrap-up’ (See Table 1). During the workshop, the PPI facilitators noted the key points of the discussion, and raised questions if, for example, they felt an explanation would be helpful for any technical concepts to ensure full understanding for contributing PPI members. The discussions were recorded to aid notetaking and enable key points to be summarised, but the conversations were not transcribed and analysed thematically as would be expected in qualitative research. This is in line with the principles of working with colleagues [13,16].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. The workshop procedure.

https://doi.org/10.1371/journal.pdig.0000521.t001

Reporting

We adopted NIHR Guidance for Reporting Involvement of Patients and the Public 2(GRIPP2) checklist to report the PPI involvement in manuscript (see Table 2) [18]. All participants were sent the report so that they could review the conclusions and add any additional points.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Reporting Involvement of Patients and the Public 2(GRIPP2) checklist.

https://doi.org/10.1371/journal.pdig.0000521.t002

Discussion on exercise 1: How can AI, indoor and outdoor environmental data be useful to help you managing asthma in your daily routine?

Feedback on exercise 2: the three potential use cases of AI, indoor and outdoor environmental data

The illustrations used to introduce the use cases in the workshops/discussions are illustrated in Figs 1–3, along with the positive and negative comments and suggestions.

1. a) A connected system, including an app with smart devices to detect how environmental triggers may be affecting asthma control and give advice to reduce potential triggers.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. User case 1: An asthma connected system.

https://doi.org/10.1371/journal.pdig.0000521.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Use case 2: An open asthma map.

https://doi.org/10.1371/journal.pdig.0000521.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Use case 3: Data visualisation.

https://doi.org/10.1371/journal.pdig.0000521.g003

PPI colleagues expressed that the smart devices would be useful to see if their asthma is deteriorating or uncontrolled at night through measuring sleep activity. It is important that there is a mechanism to transfer the monitoring data to GP/asthma nurses.

AI could monitor cough, wheeze while patients were sleeping or record night-time use of a reliever inhaler and suggest if treatment needs to be adjusted.

Concerns: PPI colleagues stated that data privacy is a huge issue. It is important to know what personal data will be shared, how it will be stored and who would have access to the data. Individuals should have control over how much data they are willing to share. Some PPI colleagues commented on a worry about the accuracy of the smart measurement devices.

1. b) An open asthma map to understand how the environment may affect individual and in the population; and support patients to plan their days and activities with objective environmental data.

The PPI colleagues thought that the more they knew about their local environmental pollution levels, the better. It would be useful to use objective data such as indoor pollutant level in the city and buildings, and patient-reported data to identify causes of asthma symptoms. They would also find it useful to know when exposure to triggers is high e.g., areas of tree pollen. One volunteer suggested that this information could be used to campaign for healthier air quality, as well as to enable individual asthma patients to avoid high polluted areas.

The idea of a ‘google maps’ style of air quality monitoring was discussed, where air quality could be reported in real time was suggested. PPI colleagues suggested it would be useful to include data on how many asthma attacks have happened in a certain area.

Some PPI colleagues suggested there was too much information on the map. Simple highlighting of polluted areas would be preferable as they were only interested in what areas to avoid.

One volunteer was concerned about the personal observations of the environment which were seen to be subjective. Another volunteer suggested representing individual experiences as a ‘thumbs up’ or a ‘thumbs down’. It was also unclear how comments would be captured and there is also an issue with the data protection of people posting comments and on social media.

Another volunteer questioned the practicality of implementing this intervention as a large amount of data would be needed to ensure the readings provided by other asthma sufferers were accurate. They also believed that would be challenging to extract the indoor environmental data from the buildings (e.g. the shopping mall shown in this use case). The costs of collecting, storing, and analysing such a large dataset were considered too large for the limited use proposed.

1. c) Data visualisation on indoor and outdoor air ventilation/pollution levels is a common analysis in building design and city planning. This technology could be used to show the real time pollutant data in individual homes and the city.

Most PPI colleagues preferred to have a simpler way to display the polluted areas of their home and city. One volunteer pointed out that everybody knows about ‘sick building syndrome’ but ‘who knows what causes sick buildings?’. They appreciated that anything that we could learn to improve life without hurting anyone has to be a good thing. They thought the top right illustration of air circulation in the home was clear and could help.

Informing the approach to interacting with AI in our next digital intervention

There was general agreement that AI does not replace humans. PPI colleagues wanted AI to support data collection, remind them about self-management tasks, teach them about asthma triggers, identify risk, and empower them to confidently look after their asthma. Echoing concerns from previous studies [17,20,21], they were clear that:

• Clinician support is needed to check the AI decision and be available for consultation when required. Patients trust that AI can support their self-management if clinicians are involved in the decision loop.
• The patient has to actively teach the AI (as opposed to only providing data to train the AI). A mechanism is needed for patients to teach the AI what is right or wrong about their asthma. For example, AI will ask for feedback on their decision, patient will give a positive credit score to the AI on right decision and negative for a wrong decision.
• The patient must make the final decision whether to follow the advice of the AI–and to tell the AI if they like their advice or not.

In our next digital intervention, we will consider the AI techniques in generative AI Large Language Models (LLM), Large Action Model neuro-symbolic models, and Human-in-the-loop (HITL) ensemble learning model to set up the AI [22,23,24]; with an aim to incorporate the patients and clinicians in the AI decision loop to support asthma self-management, and we will test and verify the user’s trust, satisfaction, and asthma outcomes of such system.

Informing the technology features in our next digital intervention

The PPI colleagues had clear preferences about the features they wanted to see in future technology:

• Correlating asthma symptoms and environmental data to identify potential triggers and providing self-management advice personalised to their daily routine,
• There was interest in illustrating areas with poor air quality on a real-time asthma map or mapping poor air quality at home—though avoiding over-complex displays. The experience of other people with asthma in the area could be useful, but comments would need to be moderated.
• Seasonality was noted as important, and data collection should extend over a seasonal cycle to see changing triggers and associated patterns of asthma symptoms.
• Connectivity was essential–transferring monitoring data and facilitating necessary contact with clinicians, connecting with the smart devices in the home.

PPI colleagues’ voiced concerns about data privacy and security of the system, especially worries about gathering information in public areas, and the permissions required to link the C4A system with indoor environmental data from existing building management systems used in most public buildings (shopping malls, offices, universities, hospitals). Worrying that personal data may be collected by third parties without any consents, resonated to the ethical concerns from stakeholders in many of the AI healthcare applications [25].

In our next intervention, we will incorporate these features into our existing C4A connected app in phases, involving PPI colleagues to contribute their thoughts on prototypes to ensure the interface is built to be useful, and simple to understand and use. There are multiple asthma sensors meeting regulatory standards to which we can link to build a connected system [17], and we will link where possible with existing environmental systems [26]. The system must comply with the General Data Protection Regulation (GDPR) and to connect with the smart build environment system that has implemented the data security strategies to safeguard patients’ personal data [27].

Informing research topics

Patients want to avoid triggers that affect their asthma, but a threshold is required to alert or highlight an area as having poor air quality. There are existing guidelines and regulations defining acceptable air quality thresholds that the system could adopt. For example, WHO set the acceptable threshold of 24-hour mean PM2.5 at 15ug/m3 in 2021[4]. Various building and ventilation regulations use different thresholds [28]. Some patients tend to read the actual air quality reading on the device and use their past experience to decide whether they should take avoiding action. Furthermore, as discussed in the workshops, it may not be possible to take avoiding action if the patient lives in a polluted area. McCarron et al research suggests personalised air quality data and greater public engagement with these data will together encourage patients to minimise their exposure to pollutants [29].

In our ongoing research, we will investigate the best practical and safe thresholds for people with asthma, and use these to inform AI alerts, and what practical and safe self-management advice should be offered. There will also be the potential to develop public engagement activities that could encourage people with asthma to avoid indoor and outdoor air pollutants [30].

Health equity

PPI colleagues recommended that low-income patients should be provided with free smart devices. This recommendation has two implications for the research team. The cost of the device must be incorporated in the budget for future grant proposals, and when planning the implementation, we need to collaborate with cross-disciplinary experts in health equity to explore how a service can be equitable, ethical, open and sustainable.

Liability

PPI colleagues recommended involving clinicians to check the AI decision and are not confident to follow unsupervised AI advice. The extent to which clinicians, manufacturers and patients are liable for incorrect advice given by AI will need to be resolved. There are many reasons for incorrect advice: an incorrect algorithm, inaccurate calibration of a devices/sensor, software ‘bugs’, or incorrect interpretation by the user (patient/clinician)[31]. In our next intervention, we will need to identify the user interface characteristics that relate to the safe use/generation of advice, by involving the clinician in the decision loop. We need to map and test the interaction model between AI, patient and clinician conforming to standards set by IEC 62366.