Development and Validation of the User Version of the Mobile Application Rating Scale (uMARS)

1Mobile health (mHealth) apps have the potential to provide around-the-clock access to evidence-based health information, education, and treatment to end users on a global scale. There are currently more than 165,000 mHealth apps (free and paid) publicly available, yet the accuracy of the health information contained in these apps is not scrutinized by regulatory bodies, which could compromise user health and safety. Concerns about the quality, efficacy, reliability, and security of mHealth apps are also often raised. While meta-analytic studies have demonstrated the efficacy of mHealth apps targeting physical activity and weight loss, the evidence base for other types of mHealth apps is poor at best.

2In response to these issues, we developed the Mobile App Rating Scale (MARS) to provide researchers, professionals, and clinicians with a brief tool for classifying and assessing the quality of mHealth apps. The 23-item MARS contains 4 objective quality subscales—engagement, functionality, aesthetics, and information quality—and a subjective quality rating. The MARS has demonstrated high levels of interrater reliability for evaluating the quality of mHealth apps on well-being and mindfulness. However, training and expertise in mHealth and the relevant health field is required to administer it. This paper describes the development and pilot testing of a simpler, end user version of the MARS (uMARS) and provides preliminary evidence for its internal consistency and test-retest reliability.

3The original MARS was simplified through the following process. The professional version was first reviewed by 2 researchers to remove complex terminology from its items and response scales. Three items requiring professional expertise, pertaining to evidence base, app goals, and accuracy of app description, were removed. Readability of the MARS and the draft uMARS was then determined using the Flesch Reading Ease test, which has a score range of 0-100, with higher scores indicating easier readability. This measure also provides the estimated US school grade required for reading comprehension.

4The draft uMARS was then pilot-tested with 13 young people, to ensure they understood the item content and response scales. The measure was embedded in prototype testing sessions of 2 mHealth apps: Ray’s Night Out and Music eScape. Ray’s Night Out uses a harm-minimization approach to increase young people’s alcohol knowledge and awareness of their drinking limits; Music eScape teaches young people how to identify and manage affect using music. Both are available on the iOS Apple app store.

5Eligible participants were Australian residents aged 16 to 25 years, who had access to an iPhone 4 or later model. The Ray’s Night Out group comprised 1 male and 8 females with a mean age of 20.7 years (SD 1.6). The Music eScape group comprised 3 males and 1 female, with a mean age of 21.5 years (SD 1.9). After testing the apps and rating them with the uMARS scale, participants were asked the question “Do you have any comments or suggestions about the uMARS rating scale?” to identify any unclear or difficult items.

6The uMARS (Multimedia Appendix 1) provides a 20-item measure that includes 4 objective quality subscales—engagement, functionality, aesthetics, and information quality—and 1 subjective quality subscale.

7The reliability of the uMARS was evaluated as part of a randomized controlled trial (RCT), testing the efficacy and quality of Music eScape. The RCT sample comprised 164 Australians aged 19.8 years on average (SD 2.51); 34 males. The highest level of education completed by 59.8% of the sample was secondary school, and 24.4% had completed a bachelor’s degree or higher. Most participants (57.9%) were students and 35.4% had full-time, part-time, or casual employment.

8Participants were randomly allocated via a Web-based research management tool developed at the Queensland University of Technology to receive immediate or 1-month delayed access to the Music eScape app. Young people were asked to use the app as much as they liked over a month, using their own iPhones. The current iOS version at the time of the trial was iOS8. Participants received weekly text messages reminding them to do so. App ratings were collected using the uMARS at 1-, 3-, and 6-month follow-ups in the immediate access group. In the delayed access group, uMARS ratings were collected at 2-, 3-, and 6-month follow-ups (ie, after 1, 2, and 5 months of app access). At each assessment point, participants were asked if they had used the app since the last assessment, and only those who reported some use were included in analyses.

9The internal consistencies of the uMARS subscales and total score were calculated using Cronbach's alpha. For the purpose of analysis, the “N/A” answer option for items 13-16 of the information subscale was recoded as “system missing,” as this option represents a qualitatively different response.

10Test-retest reliabilities were calculated for the subscales and total scores of the uMARS after 1 month of app use and at 3 months post baseline (ie, a test-retest period of 1-2 months), and over 3 months (ie, between assessments at 3 and 6 months post baseline). Interclass correlation coefficients (ICCs) were used, as they provide weighted values of rater agreement and assess proximity rather than equality of ratings. To calculate the ICCs, a random-effects average measures model with absolute agreement was utilized. Data were analyzed with SPSS version 23 (SPSS Inc, Armonk, NY, USA).

11Results of the Flesch-Kincaid readability tests are in Table 1. Scores indicated that the uMARS was written in plain English and that its required reading level was approximately grade 8.

12No suggestions for further scale improvement were made. Seven of the 13 participants who pilot-tested the scale left the comments or suggestions item blank, 4 wrote “no,” and 1 wrote “Well done. Good questions. Well explained.” Another wrote “I thought it was shorter/there is a brief or revised version of it?”

13A total of 152 of the 164 (92%) participants completed the survey after 1 month of app use. Of these, 19 indicated they never used the app, and were excluded from analyses. For the remaining 133 participants, the total uMARS score had excellent internal consistency (Cronbach alpha =.90). Internal consistencies of its subscales were also very high (engagement alpha =.80; functionality alpha =.70; aesthetics alpha =.71; information alpha =.78; and satisfaction alpha =.78).

14Test-retest reliabilities are presented in Table 2. A total of 113 participants completed the scale after 1 month of app use and at 3 months post baseline (ie, a test-retest period of 1-2 months), and 74 completed both the 3- and 6-month surveys (giving a 3-month test-retest period). All included participants had used the app at least once since the previous survey. The uMARS total score demonstrated good levels of ICC of.66 and.70 over 1- to 2-month and 3-month periods, respectively. Levels for all subscales scores were similarly high.

15This study developed and tested an app user version of the original MARS to assist app developers and researchers with evaluating the quality of mHealth apps. The uMARS (Multimedia Appendix 1) provides a 20-item measure that includes 4 objective quality subscales—engagement, functionality, aesthetics, and information quality—and 1 subjective quality subscale. One further subscale, consisting of 6 items is added to measure users’ perceived impact of the evaluated app. The study demonstrated that the uMARS had excellent internal consistency for the full scale and good levels for all subscales. It is reassuring that even after a 3-month delay between ratings, test-retest reliability of the total score was good, and test-retest reliabilities of its subscales were fair to good, with the engagement and subjective subscales being particularly robust.

16These results indicate that the uMARS provides a reliable measure of app quality in target users. Replication of the current results with multiple types of mHealth apps is required to provide additional confidence in its performance. Tests of its sensitivity to improvements in app quality and an examination of its ability to predict outcomes of mHealth apps are also needed. As the uMARS may potentially have applications beyond mHealth, tests of its performance in other domains are also indicated.

17Current indications are that the uMARS will offer an unprecedented ability to readily obtain rich information from users about mobile apps. The scale can be used to obtain user feedback on the quality of mobile apps during the development and testing process, which may result in overall improvements in their quality.