Lopes et al. ¹³ developed a SapoFitness application, which even in the earlier days of smartphones, back in 2011, had an overall positive reception from their users as a tool for weight monitoring and assessment. Martin et al. ¹⁴ developed SmartLoss, an automated mHealth system for weight loss consisting of a smartphone application and a dashboard. SmartLoss leveraged mathematical models to predict weight change and check if the user’s actual weight measurements fell within an adherence zone based their anthropometric data. The system automatically retrieved weight data from a smart scale and activity data from fitness devices like the Fitbit. The application allowed users to receive automatic and personalised feedback as well as to track their weight and steps done. We extend the work described above with additional data monitoring modalities to better understand behaviour, modern cross-platform application, and a health dashboard for managing fitness goals and data labelling to complement automatically collected data. We also include gamification elements in our smartphone application. With our approach, we intent to study conditions for patient adherence and satisfaction.

Klein et al. ⁹ introduced eMate system to support patients with HIV, diabetes or cardiovascular disease by monitoring patient behaviour and providing cooperative coaching. In order to support patients, eMate collected behavioural data, including medication adherence through smart pillbox, adherence to healthy food choices through in-app questionnaire and adherence to fitness goals. Subsequently, eMate estimated what caused the patient to not adhere to the proposed health goals with the help of questionnaires so the patient could receive personalised motivational messages to help them change their behaviour.

Scheerman et al. ²³ introduced WhiteTeeth, a mobile application to nudge adolescents with fixed orthodontic appliances to improve their oral hygiene. Adolescents would take a selfie on which they could click on the already coloured areas with any dental plague to provide feedback on the actual status of dental plague they may be having. The adolescents would receive positive reinforcement if they brushed their teeth or achieved their dental hygiene goals as well as reminders and timers to brush their teeth. In case the adolescent did not meet their dental health goals, they were instructed to create a coping plan so they can improve as coping is a valuable resource for long-term lifestyle change. ²⁴

SitCoach by van Dantzig et al. ⁶ included a mobile application to suggest active breaks from sedentary activities to office workers. The physical activity was measured by on-device accelerometer sensor and the office workers were reminded to take a break via the vibration and auditory signals from their smartphones.

In the present work, we included nudges for patient behavioural change in the form of “tips of the day” provided by the patient’s health coaches. The recommendations provided by the app aligned with the in-person behaviour change advice provided by the health coaches.

In recent years, gamification has come to light as a promising approach to promote healthy lifestyle changes, particularly in the context of health improvement and tracking. ^{4, 15} By incorporating game-like elements into health interventions, researchers aim to increase user engagement, motivation, and adherence to healthier behaviours.

Pollak et al. ²¹ explored the potential of mobile games to encourage healthy eating habits. Their research highlighted how game-based interventions can make the process of adopting better dietary choices more engaging and enjoyable for users. By leveraging the widespread use of smartphones and the appeal of mobile gaming, there is a potential to reach a broad audience and integrate seamlessly into daily routines.

Orji et al. ¹⁷ developed LunchTime, a slow-casual game designed to promote long-term dietary behaviour change. Chiu et al. ⁵ introduced the Playful Bottle, a mobile social persuasion system designed to motivate healthy water intake to tackle another important aspect of a person’s weight loss journey. Another approach to behaviour change encouragement was presented by Soares et al. ²⁵ with their Sustainable App for encouraging sustainable choices. The Sustainable App introduced a virtual pet plant whose mood changes depending on the points collected while choosing sustainable choices.

In this work, we introduce a multi-goal concept, where patients receive specific goals according to agreements made with their health coach. Based on their goal achievement, patients receive points to be spent for the interaction with a virtual pet in the EghiFit application. Patients could receive feedback in the EghiFit app from their health coaches who review the health dashboard data (i.e., tips of the day).

For the cloud-based behaviour and context analysis system, we consider three user types: patients, researchers and health coaches. The patients receive personalised health goals and tips in the EghiFit application as set by the health coaches. Researchers are responsible for cloud server setup and data analysis and thus have access to part or all of the patient data. Health coaches are responsible for patient management, including onboarding and off-boarding, setting patient goals, reviewing patient performance, and providing personalised feedback. Coaches could give feedback to the patients either in person or through modifying goals or tips. All data communication between the components and the cloud server are secured and controlled by access credentials. Figure 1 shows an overview of our approach.

3.2.1 EghiFit application

We utilised an array of technologies to collect, analyse, and leverage health-related data from the enrolled patients. The data collection was achieved through a smartphone application, termed EghiFit, which is compatible with both Android and iOS devices. The application serves as a comprehensive health tracking tool, capable of monitoring various aspects of a patient’s daily life. Patients were instructed to use a smartwatch alongside the EghiFit application so the vital health information could be acquired.

EghiFit keeps track of patient’s physical activity through total amount of steps taken in a day and heart rate information which provide valuable insights into patient’s overall physical activity and cardiovascular health. Sleep monitoring is another important feature as data on sleep patterns and quality can relate to metabolism and unhealthy eating habits. ^{3, 19} Additionally, the application collects local weather information, which can be correlated with other health data to identify potential environmental influences on weight loss outcomes.

Except for automatically acquired patient data, EghiFit supported a meal journal for patients to log their food intake. Food items with barcodes could be scanned by patients to ease the meal entry burden. The patients were instructed to log their meals for at least three days in a row each week. The first meal logging day of the week was shifted by one day compared to the previous week to have more comprehensive data on patient eating habits across workdays (Monday to Friday) and weekends.

A cross-platform approach was used to implement the EghiFit application, based on the Flutter software development kit. Fitness data was collected from both major health platforms, Google Fit and Apple Health.

3.2.2 EghiFit gamification

To enhance patient engagement and motivation, we incorporated several gamification elements into the application. Patients could set personalised health goals during in-person sessions with their health coach. These goals serve as concrete targets for patients to strive towards, promoting active participation in their weight loss journey. As patients achieve their goals, they received points to be spent for the interactions with their virtual pet in the EghiFit application. The gamification feature allowed patients to feed, give water to, and play with their virtual pet. The virtual pet interaction was intended to create a positive reinforcement loop that encourages patients to maintain healthy behaviours. Furthermore, we designed a game level scoring within the application to give patients a concrete measure of their progress. Alongside interacting with their virtual pets, patients also received personalised tips of the day with detailed action suggestions on a healthy lifestyle and insights into how to lose weight for the specific day.

Patients received two goals per goal type for any study day. Goal types were all quantitative and included amount of steps, number of meals logged and amount of tips implemented. While the first goal had a comparably lower quantity and gave one point upon completion, the second goal required a higher quantity and gave two points upon completion.

The EghiFit application included an overall game score, where patients could gain one level if they achieve the second goal for all goals given for one day. Patients lost one game level, if at least the first goal level was not achieved for any of the goals set on a study day and remained on the same level if all the goals were achieved up to at least the first level. Besides the overall game score, the virtual pet had a daily point score of needs met (zero, one or two points), where the score depended on the patient’s goal achievement and if the patient used the acquired points to interact with their pet.

3.2.3 BEXOME cloud server

The data collected through the smartphone application was securely transmitted and stored in a cloud environment. The cloud system consisted of two main components: a back-end for data storage and processing, and a front-end interface for managing patient enrolment. Moreover, the back-end implemented the game logic and point scoring, while the EghiFit application provided the view on the patient’s data and progress. Our design intended to ensure a simple and straightforward introduction of new features without requiring patients to update the EghiFit application. The cloud server was implemented using the Django framework.

3.2.4 Health dashboard

To provide real time insights into patient’s progress, a web-based dashboard to be used by the health coaches was developed. The health dashboard allows health coaches to visualise the data collected from patients so they can identify trends, patterns, and potential areas of concern. Health coaches can also use the dashboard to label notable events within the patients’ data, which is particularly valuable for providing context to the collected data. The dashboard also empowers health coaches to adjust patients’ goals based on their progress and to provide the patients with personalised tips of the day. This flexibility ensures that the goals remain challenging yet achievable, maintaining patient motivation throughout the study.

3.3.1 Study description

Ten patients of a weight loss intervention program were included in the study. Patients received detailed information about the study purpose and signed a consent form before being included. Out of a total of ten patients, eight successfully completed the onboarding process. Nearly all of the patients lost weight during the course of the study. Detailed patient statistics can be seen in Table 1.

Patients were enrolled during the course of three months. Each week, patients would meet with their health coaches who would give them personalised advice based on their measured body mass, information from the dashboard and patient feedback. Any concerns or technical challenges would be resolved as soon as possible. The patients were instructed to adhere to protein rich, low carbohydrate diet and to avoid sweets and unhealthy, energy packed foods.

3.3.2 Onboarding process

Detailed instructions on how to operate the EghiFit application were given to patients, including how food intake is logged, as well as how to interact with the virtual pets. Patients were also given detailed instructions on how to setup their smartwatches, connect them to their mobile device and how to install necessary 3rd party applications, if not already available on their device. Patients could use their own smartwatch, if available, and were provided with a new smartwatch otherwise.

In order to use EghiFit, patients needed to register in the application by scanning a unique QR code provided to them by the health coaches. After the first initialisation, the patient’s smartphone was registered in the BEXOME cloud server back-end and only then the patient could use the EghiFit application.

3.3.3 Interviews

To gain insight into how patients used the EghiFit application, we conducted interviews with the patients during the last two weeks of the weight loss intervention program. In the interviews, we asked patients about their experience using the EghiFit application.

Total number of days with successful recordings per data modality and per patient is shown in Figure 2. Some patients used the EghiFit application and followed the usage instructions diligently, while other patients had considerably reduced data than the study duration suggested, even missing some data modalities.

Sleep and heart rate information required the patients to wear the smartwatch during day and night and take breaks only for recharging periods during the day. As it is visible in Figure 2, most of the patients did not use their smartwatch correctly during the course of the study. We got feedback from the patients that they do not find wearing the smartwatch during the night comfortable or desirable which led to insufficient sleep data collected. Heart rate data had a similar fate with heart rate data lacking for half of the patients.

As steps data were automatically acquired and meals manually, there are more days of recordings for step counts as opposed to meal information for Patients 1, 4, 5, 6 and 8. Although meal logging required manual effort, most of the patients have at least 20 days of recordings.

Subsequently, we analysed workday and weekend data separately. Figure 3 shows average steps, average calories and average tips done in the EghiFit application scaled by the maximum value for the data modality across all patients. In other words, steps during workday and weekend are comparable with one another and also across patients, same for calories and number of tips.

Steps data of Patients 3, 7, and 8 had errors due to an error in the smartphone setup. The highest number of average steps were scored by Patient 2 (2533 avg. steps/day) and Patient 4 (5712 avg. steps/day). Patients 1, 4, 5, and 6 did more steps during the workday as opposed to the weekend. All of the patients have logged their meal intake for at least some of the days with some patients not logging meals during the weekend. Patients 4, 5, 6, and 8 had strong motivation to use the tips system while other patients had relatively less tips done.

Reported meal data binned into specific times of the day provides insights into patients’ habits. Most of the patients were not having late night meals, every patient except Patient 4, as shown in Figure 4. Some patients, on average, had imbalanced meals depending on the time of the day with the patient having eaten significantly more calories in a specific time period. On the other hand, other patients reported more consistent amount of calories eaten throughout the day.

During the course of the study, patients could select exact date and time of their reported meal with the current date and time being preselected. Figure 5 shows that most of the patients did not choose to use the option of selecting exact values in the application, instead, the patients have opted for keeping the preselected values in the EghiFit application. Primarily the Patient 5 and Patient 6 had a significant amount of backlogging done, going back more than one day with Patient 1 having done so on only a few occasions.

The lack of backlogging done by the patients is an important insight into meal logging and general effort required to log the meals. The patients resort to doing the least amount of individual steps in the application while still adhering to doing the meal logging as per the instructions the patients were given.

Figure 6 shows percentage of the total days the patient used the EghiFit application where they achieved a certain goal. One of the hardest goals to achieve was the amount of steps taken per day, even though the amount of data collected for that data modality is high due to ease of setup and use. Only Patient 4 had more than 50% of the days achieving a steps goal. Most of the patients had at least 40% of the days meal logging goal achieved even though it is a manual effort, compared to tips where it only took a click of a button on the EghiFit application interface. In comparison to manual meal entry goals, the second goal of completing tips of the day was achieved less often for all of the patients except for Patient 4. When we asked Patient 4 about their experience with the tips of the day, they were very positive, saying that the tips were very motivating. Thus, they used it diligently, while other patients found logging their meals more attractive.

After achieving their goals, patients could use the points they have collected to interact with their virtual pet by playing with it, feeding it or giving it water. Up to 60% of rewarded points remained unused in the Patient 2 case. Percentage of days with at least one unused rewarded point is shown on Figure 6. While Patient 2 had a significant amount of unused points which indicates their lack of conviction in the pet reward system, Patient 2 was not so successful in achieving their goals either which might indicate lack of motivation. The most motivated patient was Patient 4 with more than 50% of the time having achieved more difficult step count goal and they were also motivated to log meal items and completing tips of the day around 80% and 90% of the time respectively. Patient 4 used all of their points to interact with their pet.

Figure 7 shows a quantised view of how many times per day every patient opened the EghiFit application. Timelines across study participants were aligned to the first day of use of the application. The frequency of opening the application at the beginning of the study was comparably high and reduced with time. EghiFit captured all of the patient’s attention for at least one month. Patient 6 in particular stopped using the EghiFit application, but decided to continue with the usage relatively regularly after a short break. The most impressive usage can be seen from Patient 4 who used the BEXOME system in the span of at least 120 days.

Out of the patients, who participated in our study, three agreed to be interviewed (Patients 1, 4, and 6). In the interviews, all three patients praised the tip system, emphasising that the tips were appropriate for their abilities (not too hard nor too easy) and motivated them to engage in healthy behaviours. For example, Patient 1 stated: “Yes, I thought that [the tip of the day] was good too. Also things that you can implement quite well. I think there were tips like’Have you done your ten squats today?’. So that’s something where I say, no matter where I’m sitting, i.e. in the office, and then he says ten squats, then I’ll just do them.”

Here we provide in-depth analysis of selected data. Total steps taken timeline for Patient 4 and Patient 6 are shown in Figure 8 for side by side comparison. Two lines for both step count goals are shown as well as the total recorded steps per each day of the study. Patient 4 had successful steps recordings for nearly every day of the study, while Patient 6 had more spotty data with missing days and improbable values (around 0 steps). Health coaches recognised the Patient 4’s efforts and increased their step goals to keep them challenged and engaged. Patient 4 completed the required steps for most of the days. Patient 6 had their step goals set relatively consistently throughout the study with them failing to meet the step goals for most of the days.

Patient 4 was very motivated and achieved a total game level 37 at the end of the study, the progress timeline can be seen in Figure 9. All of the other patients ended the study at level 1. These findings indicate that gamification and interaction elements presented in EghiFit application have a potential to capture patient’s attention and be their partner in their fitness journey. When we asked Patient 4 about their experience with the app, they mentioned that they particularly appreciated the app’s ability to provide an overview of their diet and exercise habits, which motivated them to engage in healthy behaviours.

To meet privacy and security requirements, ethics approval was obtained for our research. The ethics approval for the study was granted by Ethics committee of the Saarland Medical Association in Germany.

Written informed consent for publication of the patients’ anonymised details and data was obtained from the patients.