In January 2019, the National Health Service published its long-term plan, setting out key ambitions for the next 10 years. One of the most ambitious targets of this plan was in the field of digital technology, with a vision toward increasing care at home using remote monitoring and digital tools [1]. This move will likely be expedited by the need for social distancing brought about by the COVID-19 pandemic. Therefore, with the impetus and growing necessity for distance health care, it is important to consider how this new type of service will meet the needs of patients. A way to ensure that new technologies are usable, acceptable, and tailored toward the patients they aim to support is to ensure that the patients themselves are central to the design and development process. Co-design offers a way to ensure that new technologies and interventions are tailored to patient needs [2]. Indeed, there is growing support for the benefits of co-design in health care [3] but less evidence as to how these approaches can be tailored to the needs of diverse patient populations [4].

Within the context of software development, co-design can be defined as a process that draws on the shared creativity of software developers and people not trained in software working together [5]. To this end, special attention is given to involving end users and ensuring that their input as experts through experience is central to the design process and that their specific needs are understood and met [5-7]. This is in line with existing literature suggesting that integrating patient voice with software development is achievable and can provide valuable feedback to improve the intuitive design and usability of software outcomes [8,9].

In dementia research, patient and public involvement (PPI) and co-design is still a developing field [10], although it has been suggested to confer benefits to research outcomes, researchers, and members of the public who play a part in the process [11,12]. The capacities, needs, and preferences of those living with cognitive impairments can be diverse [13]. Therefore, standard co-design and PPI methodologies often need to be adapted to suit this population. This may be particularly important in the field of digital technology and software development, as studies suggest that there is utility and an appetite for assistive technology for older people and those living with cognitive impairments. However, despite there being motivation for older people to use digital technologies, barriers exist around usability and lack of experience [14-17].

An area in which digital technology can help with the care and management of dementia is through the monitoring of cognitive change and variability, which is an issue that is considered important for this population [18,19]. For instance, many people with dementia experience worsening cognitive and neuropsychiatric symptoms in later periods of the day, a phenomenon known as sundowning [20,21]. Current practice bases the diagnosis of dementia on a combination of clinical history, biomarker detection, and examination, of which cognitive testing is a key part [22]. However, conventional cognitive assessments cannot detect short-term fluctuations in cognition that might be relevant to understanding or managing individuals’ cognitive, functional, and behavioral symptoms.

Recent developments in computerized cognitive testing have made it possible to measure microlongitudinal patterns of cognitive function [23]. However, although these tools have been tested in cognitively healthy older adults [14], they have not yet been used in populations with cognitive impairment. Furthermore, these tasks were designed for use on large, touchscreen tablet devices and have not yet been adapted for use on smaller, more mobile devices, such as smartphones, which are used by an increasing number of older people [23]. Therefore, there is an impetus to adapt such tasks for use on smartphone devices and meet the needs of those living with clinical conditions that affect their cognitive abilities [24].

Despite the diverse and divergent lived experiences of those living with dementia, software apps are rarely designed with this patient population in mind [15]. It is even rarer to find software codeveloped alongside those living with dementia [4]. This can result in poorer quality technology that can be difficult to use for those living with dementia [25]. However, research indicates that those living with dementia have an interest in assistive technology and are capable of using touchscreen technology [17,26]. Therefore, we approached the adaptation of microlongitudinal computerized cognitive tests to the needs of people living with cognitive impairments through an iterative Agile process with patient co-design at its center.

Agile software development focuses on collaboration with users and rapid software deployment [27]. Scaling tests to a mobile device requires regular input and development iterations from end users, with an understanding that direct translation between devices may be unsuitable. The Agile methodology is best suited to such projects where requirements may not be clearly defined at the outset and emerge over time [28]. This is especially relevant in this case, where iterative co-design workshops spaced throughout the development process meant that the final product was not clearly defined early in the process and, instead, emerged based on consultations with experts through experience via regular workshops.

In this paper, we describe how we modified co-design approaches to involve members of the public living with dementia and their care partners in the production of a smartphone app. Although we worked specifically with people with dementia, the principles could be applied to other patient groups who do not find it easy to engage with standard co-design approaches. We also explain the benefits of the Scrum development methodology as a way of integrating user feedback into the design and development process.

The initial prototype comprised a shopping list task and an N-back task (Figure 1).

Much of the participant feedback collected during workshop 1 correlated poorly with the research team’s prior assumptions. Some of the key feedback from workshop 1 and the actions taken to address this feedback are presented in Table 1.

Overall, workshop participants seemed positively disposed to the purpose of the app and said that assuming certain alterations were made, they would be willing to interact with such a program on a subdaily basis.

Building on feedback from workshop 1, the software team undertook a second development sprint, updating the original prototype to incorporate feedback from workshop 1, including removal of N-back task and replacement with shopping list+ task (Figure 2).

Feedback from this workshop and actions taken are listed below in Table 2.

In line with feedback from the first workshop, no major objections were raised to the usability and acceptability of the MyMindCheck app. Indeed, 1 attendee who stated at the beginning of the workshop that she did not use mobile phones was particularly fast to pick up both tasks and noted at the end of the workshop that she had enjoyed testing the app.

Workshops 3 and 4 aimed to test the software alterations implemented as a result of workshop 2 and to trial new functionality, including prompts and alarms. Feedback from these workshops and the actions taken to address this feedback are listed as follows:

Although most participants complied with the assigned in-app tests, the most common reasons for noncompliance were

To address these concerns, the software team implemented the following modifications:

One tester also noted in her diary that, for the first 2 days of testing, the phone did not register her responses, and therefore was timing out on the tests. Similar issues had surfaced with other testers to a lesser extent in some of the preceding workshops. On the basis of this feedback and previous observations, it was noted that some participants were holding the response buttons rather than tapping them, perhaps reflecting a level of unfamiliarity with mobile technology among this group. Therefore, to address this issue the software was modified to identify both on-press and on-hold events as valid answers.

Testers were confident using both tasks, although they noted that they found the shopping list+ task harder and that it required more concentration. Testers with experience caring for someone with dementia stated that they believed that these tasks could be completed by someone living with dementia, assuming they had support from a care partner.

We used a co-design approach to develop the MyMindCheck app involving three key groups of stakeholders: the research team (including clinical input), the target user group (people living with dementia or individuals with direct experience of caring for others with dementia), and the software development team. As patient involvement and co-design in dementia research is in its relative infancy across Europe [10], this study will make an important contribution toward a model of best practice for related research and provide an exemplar for others wishing to adopt and modify this approach. Our report conforms to the Guidance for Reporting Involvement of Patients and the Public-2 international reporting guidelines for PPI [36]. Therefore, findings from this study will be comparable and address concerns raised by some researchers regarding the lack of consistent reporting in co-design research [37], especially in regard to those living with dementia [4].

By adopting Scrum in this context, we were able to realize the benefits of an iterative co-design approach, with the software evolving throughout each of the 4 workshops. In addition, our use of prototype designs in the first 2 workshops provided the team with a low-cost opportunity to receive feedback and evaluate the idea before commencing software development.

Participants in the workshops gave positive feedback about the experience, showed strong engagement during the sessions, and provided constructive comments on the app. Notably, points raised in early workshops did not resurface in the week-long test undertaken by a different participant group at a later stage. This could be because users in the week-long test focused on different aspects of the app. However, it could also suggest that our approach was effective in addressing design issues at an early stage of development.

Although the co-design methodology enabled the team to iteratively develop the app, we still had to overcome several challenges. For instance, it was agreed early on that embedding of an end user (in this case, someone living with a dementia diagnosis into the Scrum team), as per co-design best practice [38], would not be feasible because of the burden that regular meetings could place on those living with a dementia diagnosis and their care partners, as well as the power differentials and communication difficulties associated with involving lay members in technical discussions [39]. Instead, we took the pragmatic approach of running workshops throughout the project to garner regular feedback from the user group and provide end-user representation through a proxy (in our study, the proxy was a public engagement officer who developed and facilitated all co-design workshops). Ideally, the project would have benefited from more regular contact with the end-user group. However, given the vulnerable nature of this group, our approach seemed to be an appropriate compromise, given that this was a fast turnaround, intensive development project.

There were some logistical challenges in running the Agile development using a co-design process. Specifically, recruiting participants for workshops required multiple interactions with community groups to garner interest in the project and plan suitable times and venues for workshops. Therefore, it was necessary to set the date for each workshop several weeks in advance. However, software development does not always run according to the plan, as it is not possible to estimate development tasks with a high degree of accuracy. Therefore, there is a risk that a date could be set for a workshop only for the software not to be ready in time. We mitigated this risk by setting a date for each workshop, which allowed a suitable leeway for any unexpected delays. We also worked with an experienced and established Scrum team, meaning that the estimates could usually be provided with a reasonable level of confidence.

Given the need for health research, particularly the development of health technology, to be approached in a patient-centered manner [37], we developed a methodology that combines Agile software development with integrated patient co-design. This approach facilitated meaningful user involvement in a manner that was easily manageable by our project team, who were working on a short timescale with budget constraints, a challenge experienced by many developers [40].

We also highlighted several instances where input provided by people with lived experience of dementia helped our team to identify and address usability issues early in the development process, speeding up delivery and reducing software development waste. Our experience evidences how co-design can benefit the software development process and be sustainably tailored to the needs of diverse patient populations [4].

The next step for the MyMindCheck app is to undertake a large pilot trial and adaptations to apply to other health conditions with fluctuating cognitive states. Patient groups will continue to be involved throughout this future work to ensure that the developed software is fit for its purpose.