Understanding adherence to the recording of ecological momentary assessments in the example of tinnitus monitoring

Miro Schleicher,Jorge Simoes,Winfried Schlee,Vishnu Unnikrishnan,Thomas Probst,Patrick Neff,Myra Spiliopoulou,Rüdiger Pryss

doi:10.1038/s41598-020-79527-0

Abstract

The recording of Ecological Momentary Assessments (EMA) can assist people with chronic diseases in monitoring their health state. However, many users quickly lose interest in their respective EMA platforms. Therefore, we studied the adherence of users of the mHealth app TrackYourTinnitus (TYT). The app is used to record EMA in people with tinnitus. 1292 users, who interacted with the app between April 2014 and February 2017, were analyzed in this work. We defined “adherence” based on the dimensions of interaction duration and interaction continuity. We propose methods that are able to predict the (dis)continuation of interaction with the app and identify user segments that are characterized by similar patterns of adherence. For the prediction task we used the data of the questionnaires MiniTF and TSCHQ, which are filled in when the users enter TYT for the first time. Additionally, time series of the eight items of the daily EMA questionnaire were used. The distribution of user activity pertaining to the adherence dimension of interaction duration revealed a very skewed distribution, with most users giving up after only 1 day of interaction. However, many users returned after interrupting for some time. Some of the MiniTF items indicated that the worries of users might have lead to an increased likelihood of returning back to the app. The MiniTF score itself was not predictive, though. The answers to the TSCHQ items, in turn, pointed to user strata (more than 65 years of age at registration), which tended towards higher interaction continuity. As the registration questionnaires predicted adherence only to a limited extent, it is promising to study the activities of the users in the very first days of interaction more deeply. It turned out in this context that the effects of interaction stimulants like personalized and non-personalized tips, pointers to information sources, and mechanisms used in online treatments for tinnitus (e.g., in iCBT) should be further investigated.

Highlights

The recording of Ecological Momentary Assessments (EMA) can assist people with chronic diseases in monitoring their health state
A remarkable finding on the use of EMA in smartphones was reported by Probst et al.[8]: they compared the sociodemographics of patients of an outpatient tinnitus clinic to the sociodemographics of the users of an EMA-based app, and found that this app reached different strata with respect to age and time since tinnitus o nset[8]
We use the EMA service of the mHealth app TrackYourTinnitus (TYT)[7] as an example, and we study the potential of machine learning methods in characterizing users who are more/ less likely to stop interacting with the app

Summary

Introduction

The recording of Ecological Momentary Assessments (EMA) can assist people with chronic diseases in monitoring their health state. As the registration questionnaires predicted adherence only to a limited extent, it is promising to study the activities of the users in the very first days of interaction more deeply. We concentrate on active engagement with a mHealth app, for the recording of longitudinal Ecological Momentary Assessments (EMA)[1]. They pointed out that the study of chronic pain depends mainly on self-reported pain intensity assessments, which fits to the principles of EMA. They identified 62 quantitative EMA research projects, with a total of 105 scientific publications. A remarkable finding on the use of EMA in smartphones was reported by Probst et al.[8]: they compared the sociodemographics of patients of an outpatient tinnitus clinic to the sociodemographics of the users of an EMA-based app, and found that this app reached different strata with respect to age and time since tinnitus o nset[8]

Methods

Results

Conclusion