JMIR Publications

ABSTRACT

Background:

Physical inactivity remains one of the leading risk factors for death worldwide. Due to increasing sedentary behaviour, vehicle-based transport and insufficient physical workload, the prevalence of physical activity decreases significantly with age. To promote sufficient levels of participation in physical activities, the project prototype Fit-mit-ILSE was developed with the goal to make adults aged 55 years and above fit and fit for the use of assistive technologies. The system combines Active and Assisted Living (AAL) technologies and smart services in the so-called ILSE-app. This app supports training at home, promotes outdoor activities and offers eLearning courses on a tablet or on a 3D camera system.

Objective:

The clustering of health and fitness app user types, especially in the context of AAL projects, has so far been defined by experts through one-dimensional cluster thresholds based on app usage frequency. The aim of this work was to investigate and present data-driven ways of clustering app user types using the example of the ILSE-app function Fit at home.

Methods:

During two phases of field trials, ILSE-app log data was collected from 165 participants. Using this dataset, three data-driven approaches were described for clustering. First, the common approach of user type clustering based on expert-defined cluster size and thresholds was replaced by a data-driven derivation of the cluster thresholds using the Jenks natural breaks algorithm. The second approach varied by statistically calculating the lower threshold and the number of clusters. Finally, a multidimensional clustering approach using the Partitioning Around Medoids algorithm was explored to include the consideration of usage pattern data.

Results:

Applying the Jenks clustering algorithm to the mean usage per day and clustering the users into four groups showed that most of the participants (38.2%) used the Fit at home function between once a week and every second day. In both one-dimensional clustering analyses, more men were in the “poor”-user groups than women. In addition, the younger participants were more often identified as “moderate” or “high” users than the older participants, who were mainly classified as “poor” users. The “moderate” and “high” user groups met the WHO functional balance and strength training recommended frequency of three or more days per week for adults aged 65 and older. In addition, the multi-dimensional approach identified four different users groups that differed in terms of time of usage, gender and region.

Conclusions:

The application of the three different clustering approaches showed that data-driven calculations of user groups can replace expert-based definitions and provide objective and time-efficient thresholds for the analysis of app usage data. Furthermore, multi-dimensional clustering can identify characteristics of user groups when more data is available and applicable to characterising user types.