Design recommendations of personal health data visualisations on mobile phones for patients with diabetes: a mixed-methods approach

Diabetes is a chronic, metabolic disease characterised by elevated levels of blood glucose (or blood sugar. If uncontrolled, diabetes can lead to serious damage to the heart, blood vessels, eyes, kidneys and nerves (WHO, 2024). To control diabetes, patients need to monitor their blood glucose and then adjust their lifestyle accordingly, both short-term and long-term. Recent advancements in health information technologies (HITs) have enabled patients to collect a wide range of data with less need for manual testing and recording. For example, FreeStyle Libre and Dexcom are popular continuous glucose monitors (CGMs) in the UK, which come with their mobile application, that allows patients to collect blood glucose in time and share data with healthcare professionals (HCPs). Fitbit is used universally to record physical activities. While existing mobile and ubiquitous technologies make it easier for patients to collect a wide range of personal health data, these technologies still do not provide the means necessary for patients to visualise this data in an effective way that can facilitate comprehension of their health status. While there is a rich literature on the use of mHealth and ubiquitous technologies for the self-management of diabetes, the majority of this literature is focused on understanding user needs and requirements regarding the process of collecting personal health data. There is very little evidence or design guidance about the way personal health data, once collected, should be visualised and presented to patients to facilitate comprehension. Design considerations are of primary importance in this case since an overestimation or underestimation of risk, or difficulty understanding health data can lead to negative emotions, lack of motivation to change lifestyles, and poor self-management decisions with adverse consequences for health outcomes. Therefore, the research reported in the present thesis produced new knowledge about the design of personal health data visualisations in this context.

Specifically, this thesis aimed to address the following two main research questions:

• Research question 1 (RQ1). What are the user needs and requirements of patients with diabetes in regard to the presentation of their personal health data?

• Research question 2 (RQ2). What are the effects of different designs of visualisations for blood glucose readings on understanding and how should blood glucose be integrated with other types of health data to present in mobile phones?

To address the research questions a mixed-methods research design grounded on a design thinking approach was employed. Specifically, in the case of research question 1 (RQ1), two studies were conducted to help build empathy, including a web-based cross-sectional questionnaire survey with 150 patients with chronic conditions in the UK (UK), that aimed to understand their behaviours, attitudes and problems faced towards personal health data and visualisations, and a series of semi-structured interviews, with 15 participants, which were focused on the needs of patients with diabetes specifically (Type I and Type II). These two studies helped to define the focus of Research Question 2 (RQ2) which involved two web-based controlled studies and two participatory design workshops. The web-based controlled studies (with 106 and 100 patients with diabetes respectively) aimed to test a series of hypotheses to help understand how to visualise blood glucose readings in reference range number lines and reference range line graphs by examining the effect of two key design attributes of visual perception (colour hue and colour style) on risk perception in situations of lower and higher risk severity (borderline and extreme out-of-range blood glucose). A total of 16 images of reference range number lines and 16 images of reference range line graphs were developed and tested in the two studies. Finally, the purpose of the participatory design workshops (n = 11), followed by a web-based survey that aimed to validate the results of the workshops (n = 21), was to investigate what other types of health data (that influence blood glucose) should be displayed to enhance understanding of blood glucose for patients with diabetes and how to integrate them in mobile phone screens.

The results of the studies showed that patients with diabetes viewed personal health data more frequently compared to other conditions. Patients with diabetes primarily used their personal health data to monitor their progress. However, the task of monitoring was frequently hindered by the difficulty they faced comprehending their data. Specifically, when monitoring their data, the results of the semi-structured interviews showed the emergence of three main themes that described participants’ needs in regard to the visualisation of their personal health data to facilitate understanding. These themes were: 1. Data representation and management, which described the lack of guidance or information on how to interpret visualisations of blood glucose readings especially when these were outside a normal range as well as the difficulty of drawing causal relationships between different types of health data (such as the relationship between and blood glucose). 2. Patient-centred tailoring, which represented the desire of participants to customise the data density, the ability to query and filter data to obtain summaries and the display of more personalised visualisations; 3. Reactions to data visualisations. This theme related to theme 1 and described situations where participants felt overwhelmed in their attempt to understand personal health data, especially without enough support from HCPs. These consequently hampers their motivation to adjust diabetes-related behaviours.

The findings reported from the studies conducted to address RQ1 further defined the questions to be focused. Specifically, the results of the controlled studies showed that, when visualising blood glucose readings using reference range number lines and reference range line graphs, certain combinations of colour hue and colour style may lead to overestimation of risk when blood glucose readings were borderline. The findings showed statistically significant interaction effects between colour hue and colour style on risk perception. The use of red colour regardless of block or gradient colour style can result in the overestimation of borderline risk by patients. Similarly, blue and grey colour hues were likely to lead to an underestimation of extreme danger. While the web-based controlled studies found the effects of visual components on risk perception of blood glucose readings, on the other hand, the participatory design workshops allowed participants to ideate possible ways blood glucose readings could be visualised alongside other types of personal health data. The results showed that participants in the workshops identified four main categories of data of interest to them when interpreting blood glucose readings. These were: 1. Blood glucose-related data (most recent readings, daily average reading, time-in-range, trend and historic data); 2. Insulin-related data; 3; Data related to food, carbohydrate and physical exercise; 4. Artificial intelligence and/or HCP’s recommendations and actionable data. The analysis of participants’ design outputs during the workshops (drawings and outputs of card sorting activities) showed a range of opportunities including timestamped graphs, tailored visualisations and gamified visualisations.

The research reported in the present thesis is one of the first attempts to generate design recommendations in this context about visualising the personal health data of patients with diabetes on mobile phones. More evaluation and product implementation/delivery from the design recommendations are needed to gain a deeper understanding of this problem area. The work reported by the author of this thesis was a step towards this direction.