mHealth technology for ecological momentary assessment in physical activity research: a systematic review

Objective To systematically review the publications on ecological momentary assessment (EMA) relating to physical activity (PA) behavior in order to classify the methodologies, and to identify the main mHealth technology-based tools and procedures that have been applied during the first 10 years since the emergence of smartphones. As a result of this review, we want to ask if there is enough evidence to propose the use of the term “mEMA” (mobile-based EMA). Design A systematic review according to PRISMA Statement (PROSPERO registration: CRD42018088136). Method Four databases (PsycINFO, CINALH, Medline and Web of Science Core Collection) were searched electronically from 2008 to February 2018. Results A total of 76 studies from 297 potential articles on the use of EMA and PA were included in this review. It was found that 71% of studies specifically used “EMA” for assessing PA behaviors but the rest used other terminology that also adjusted to the inclusion criteria. Just over half (51.3%) of studies (39) used mHealth technology, mainly smartphones, for collecting EMA data. The majority (79.5%) of these studies (31 out of 39) were published during the last 4 years. On the other hand, 58.8% of studies that only used paper-and-pencil were published during the first 3 years of the 10-year period analyzed. An accelerometer was the main built-in sensor used for collecting PA behavior by means of mHealth (69%). Most of the studies were carried out on young-adult samples, with only three studies in older adults. Women were included in 60% of studies, and healthy people in 82%. The studies lasted between 1 and 7 days in 57.9%, and between three and seven assessments per day were carried out in 37%. The most popular topics evaluated together with PA were psychological state and social and environmental context. Conclusions We have classified the EMA methodologies used for assessing PA behaviors. A total of 71% of studies used the term “EMA” and 51.3% used mHealth technology. Accelerometers have been the main built-in sensor used for collecting PA. The change of trend in the use of tools for EMA in PA coincides with the technological advances of the last decade due to the emergence of smartphones and mHealth technology. There is enough evidence to use the term mEMA when mHealth technology is being used for monitoring real-time lifestyle behaviors in natural situations. We define mEMA as the use of mobile computing and communication technologies for the EMA of health and lifestyle behaviors. It is clear that the use of mHealth is increasing, but there is still a lot to be gained from taking advantage of all the capabilities of this technology in order to apply EMA to PA behavior. Thus, mEMA methodology can help in the monitoring of healthy lifestyles under both subjective and objective perspectives. The tendency for future research should be the automatic recognition of the PA of the user without interrupting their behavior. The ecological information could be completed with voice messages, image captures or brief text selections on the touch screen made in real time, all managed through smartphone apps. This methodology could be extended when EMA combined with mHealth are used to evaluate other lifestyle behaviors.


Study selection and inclusion criteria
No exclusion criteria were applied for gender, age or clinical condition, but as regards the 145 language, only full-text articles in English or Spanish were reviewed. Reviews, editorials, 146 protocols and theses were not included. The articles selected by title and abstract met the 147 conditions indicated in Table 1. EMA could be applied to PA or to other variables, but then PA 148 has to be assessed by other methodologies and compared to the main EMA variable. 149 150 *** Include Table 1 here *** 151 152 Data extraction 153 In a first step, duplicate articles from the four databases were deleted using Mendeley. One 154 reviewer (RZ) applied the inclusion/exclusion criteria to all titles and abstracts. Articles meeting 155 the inclusion criteria were selected and when decisions could not be made from the title and 156 abstract alone, the full paper was also retrieved. The selected papers were checked independently 157 by two review authors (RZ, JFL). Discrepancies were resolved through discussion (with a third 158 author where necessary, LC) until reaching consensus. A standardized, pre-piloted form was 159 used to extract data from the included articles in order to assess the study quality and for the 160 synthesis of the evidence. Extracted information included: general information (author, year, 161 country); sample (size, population, age, etc.); details of tools used (measure, purpose/use, type of 162 tool, etc.); methodological protocol (experimental design, response rate, time interval required to 163 completion, and mode of administration, sensor use, etc.) as well as assessing the main variables. 164 Furthermore, the data extraction was carried out by two reviewers (RZ, JFL). 165 166 Risk of bias assessment tools 167 The tool proposed by National Institute for Health and Care Excellence (NICE) for prognostic 168 studies (24) was applied by two reviews (RZ, JFL) to assess the risk of bias (RoB) of the selected 169 non-experimental studies. Following the same procedure, to assess the RoB of the selected 170 experimental and quasi-experimental studies we applied the "Cochrane Risk of Bias Tool for 171 Randomized Trials" (25) and the "Risk Of Bias In Non-randomized Studies -of Interventions" 172 (ROBINS-I) (26), respectively. Since the aim of the systematic review was not to analyze the 173 obtained results, the following items were removed: from NICE 1.3 ("the prognostic factor of 174 interest is adequately measured in study participants, sufficient to limit potential bias") and 1.6 175 ("the statistical analysis is appropriately accounted for, limiting potential bias with respect to 176 the prognostic factor of interest"); from Cochrane "Selective Reporting" and "Other Bias"; and 177 from ROBINS-I "Bias in measurement of outcomes" and "Bias in selection of the reported 178 result". 179 180 Strategy for data synthesis 181 We provide a narrative synthesis of the findings from the included studies structuring it around 182 the methods and the procedure related to EMA in PA research. The main information is also  Risk of bias assessment 279 NICE tool was applied to assess the RoB of 92.1% of studies (70 longitudinal prospective 280 studies), the Cochrane tool was applied to 5 experimental studies (6.6%), and ROBINS-I tool 281 was applied to 1 quasi-experimental study. Figure 3 shows a summary of the results of RoB 282 assessment. 283 About half of the longitudinal prospective studies were evaluated as high RoB in three NICE 284 criteria (representativeness, outcome adequately measured and potential confounders 285 appropriately controlled for), and 29% of studies did not provide enough information about loss 286 to follow-up of participants and/or the amount of missing data. Half of the experimental and 287 quasi-experimental studies showed a high RoB regarding the allocation concealment of 288 participants and the amount of missing data in the outcome measure, and half of the studies did 289 not provide enough information about this last RoB. The aim of this study was to systematically review the scientific publications on EMA relating to 295 physical activity (PA) behavior in order to classify the methodologies and to identify the main 296 mHealth technology-based tools and procedures that have been applied during the first ten years 297 since the emergence of smartphones. We have found enough evidence to use the term mEMA 298 (mobile-based EMA) when mHealth technology is being used together with the EMA 299 methodology for monitoring lifestyle behaviors such as PA, in real time and in natural 300 environments. A total of 76 studies (in 74 articles) on the use of EMA and PA were included for 301 the synthesis. The majority of studies were carried out on healthy adults, lasted around one week Manuscript to be reviewed 342 there is still a long way to make the most of the capabilities of this technology for applying EMA 343 to PA behavior. 344 Thus, based on our systematic review, we have found enough evidence to firmly propose the use 345 of the term mEMA (mobile-based EMA) when mHealth technology is being used together with 346 the EMA methodology, although this term had already been used previously (94). More 347 specifically, we agree with defining mEMA as "the use of mobile computing and communication 348 technologies for the ecological momentary assessment of health and lifestyle behaviors". As an 349 example, some researchers have combined mHealth and EMA methodologies to study alcohol, 350 tobacco and drugs consumption (106), depression and anxiety (107), eating disorders and obesity 351 (108), or nutrition and PA behaviors (109). Thus, mEMA methodology can facilitate monitoring 352 of healthy lifestyles under both subjective and objective perspectives, using tools such as written 353 diaries or self-reports on a touch screen, messages on social networks, and motion or 354 physiological sensors, all managed through smartphone apps (7,110). The selected studies included heterogeneous samples from children to older adults, men and 358 women, healthy and clinical participants. Regarding the age of the participants, the studies were 359 carried out on young-adult samples, including children and adolescents (25%), university 360 students (26%) and adults (45%). In general, this variety of population age is receptive to the use 361 of new technologies like smartphones as well as using this technological advance daily 362 (111,112). This fact could explain the high rate of technology used for EMA assessments in our 363 review (76.3%). Conversely, only 3 studies were found using EMA and PA in older adults 364 (39,61,63). Studies in the elderly are an important challenge for science, due to the increase in 365 life expectancy (113). Thus, we encourage scientists to incorporate older adults and elderly 366 samples in the future EMA projects. It has been pointed out that older people are reluctant to 367 participate in studies that promote electronic forms of data collection (114). However, 368 technology is becoming an increasingly used strategy in health research (115), because, among 369 other reasons, it may offer a great level of accuracy in PA research using EMA (116). Thereby, 370 mHealth research has the capability to adapt the advances in technologies to elderly and 371 overcome the initial rejection. In fact, the 3 studies found with older adults used smartphones and 372 accelerometers (61,63), and only one of them reported difficulties with the battery management 373 and the size of the response scales on the screen (39). It is noteworthy that the average age of the 374 participants in the two studies without technological issues was 60.1±7.1, which is not even 375 considered old population in most countries; whereas the average age was 68.7±5.5 years in the 376 only study with few technological difficulties, which could be considered pre-old age (117). This 377 fact could explain the positive acceptance of technology by the older adult samples analyzed in 378 the 3 selected studies. 379 As regards gender, we found more studies carried out with women (60%) rather than men. 380 Previous literature has indicated that women could be more adherent than men to participate in 381 research projects and this could be a possible explanation for the gender differences found (118). 382 Interestingly, in general women were less likely to participate in both PA and exercise behaviors 383 (119). Therefore, future EMA and PA studies should take into account possible gender 384 differences and it should continue working along this line, seeking gender equality in scientific 385 studies, as part of the RRI framework (Responsible, Research, Innovation) (120). 386 As regards health status, most participants were considered healthy and only 14 studies (18%) 387 included samples with both a physiological or psychological clinical condition. For the latter, the 388 most common were mood disorders. In this sense, current studies show possible relationships 389 between mood disorders and PA. In detail, PA has been presented as a good complemental 390 therapeutic strategy to reduce stress, alleviate depression symptoms and enhance psychological 391 states (121). Thus, EMA methodology could contribute to further explain the relationship 392 between psychological states and PA engagement. In addition, the main topic that accompanied 393 PA in healthy samples was "affect and emotions" (Fig. 2), highlighting the influence of PA on 394 psychological states. 395 396 Study designs used in EMA and PA, and methodological aspects. 397 An interesting result from the reviewed studies corresponds to the research designs. This 398 systematic review shows that 92% of the studies followed a longitudinal prospective design, 399 while only 7% of them followed an experimental design, and 1% a quasi-experimental design 400 (Table 2). Hence, increasing the amount of experimental designs could help in determining 401 cause-effect relationships as regards PA and other variables like mental states in future studies. 402 As regards the duration of the EMA assessments, the range was from one day (62) up to 12 403 months (38). The majority of the studies lasted between 1 and 7 days (58%). From this result, it 404 is suggested to increase the duration of the future studies using EMA in order to determine long-405 term habits. However, a limitation of extended longitudinal studies could be an increase in the 406 number of dropouts (122). This is known as attrition concern can lead to the subsequent biases of 407 auto-selection and experimental survival (123). In other words, participants who reported all 408 EMA assessments throughout a very long study could have different individual characteristics 409 from those who not complete all the study. It was then checked whether studies showing a high 410 Risk of Bias related to loss to follow-up were those with longer duration, without finding 411 significant results. Therefore, we suggest a duration between 1 and 4 weeks as an optimal 412 balance between habits information and low levels of dropout for future EMA studies. 413 In relation to the within-day intervals for EMA assessments, 35% of the studies established EMA 414 assessments throughout the whole day, whereas 32% scheduled the assessments according to the 415 availability of the participants; for example, studies with children samples avoided school hours 416 (81). Interestingly, only 26% of studies were to found directly differentiate between workingdays 417 and non-working days. The comparison of working days and non-working days enables possible 418 patterns of physical activity depending on the type of day to be studied (124). Finally, in our 419 systematic review there is a great variety in the range of the number of assessments per day. 420 Between 3 and 7 assessments per day were carried out in 37% of studies, followed by 8 to 12 per 421 day (22%) and 1 to 2 per day (16%). We recommend between 3 to 7 assessments per day. In this 422 context, it has been published that excessive prompts or requests for EMA surveys could 423 increase the number of lost responses, or it could cause participants to respond randomly (125). 424 425 Topics associated with the research in EMA and PA. 426 The main topic (omitting demographic data) that accompanied PA was psychological state (Fig  427 2), which is a global variable that includes: affection, emotions, depression, stress or anxiety. 428 This is interesting due to the high rates of mental disorders in the general population worldwide 429 (126) and the positive relation between PA practice and mental health. General activity (as well 430 as BMI) was the second topic studied along with PA. It is related to everyday behaviors like 431 active transport, watching TV or eating. These behaviors are important to understand the lifestyle 432 of the participants and their possible motives and barriers for PA practice (124). 433 Finally, some stable anthropometric variables like BMI have been assessed in 35 studies 434 (46.1%), although they have not been evaluated by EMA, demonstrating the importance of PA 435 behavior in relation to overweight and obesity (127). 436 437 Limitations 438 Our systematic review is pioneer in examining mHealth application for EMA studies in the field 439 of PA, though there are some limitations. First, it was not possible to report on the adherence 440 levels to EMA in the participants of most of the studies, because the methodological strategies 441 for reporting EMA data collection were diverse. Similar information has been reported by Liao 442 et al (14) in a systematic review on the use of EMA on diet and PA in young population, 443 highlighting the heterogeneity in the EMA data collection methods. We encourage researchers to 444 incorporate results on adherence to EMA interventions as well as the number of dropouts. 445 Although mHealth technologies can help to provide objective EMA recordings, they also have 446 some difficulties. For instance, the high cost of development and maintenance for mHealth Apps, 447 the lack of standardization, data management, technical problems, slow connections, and so on 448 are possible problems that should be borne in mind (128). In addition, mEMA devices could be 449 expensive if they are only used for research purposes. 450 451 Future research directions. 452 mHealth technology could be of great help to apply EMA strategies in developing countries 453 (129), since it reduces the costs compared to a traditional intervention (130). A few years ago, 454 specific and expensive sensors like isolated accelerometers were used to investigate PA 455 behaviors. 456 We encourage researchers to require participants to use their own devices instead of providing 457 specific research instruments to take advantage of the increasing ability of smartphones for 458 synchronously monitoring different objective parameters (16,17,18). This should be the current 459 trend in research on healthy lifestyles like PA behavior. Smartphone Apps and sensors (mHealth 460 technology) allow the concept of mEMA to be accepted, especially in relation to PA. It is ideal 461 to use built-in sensors or other sensors easily connected by Bluetooth for monitoring the PeerJ reviewing PDF | (2019:09:41586:2:0:CHECK 21 Feb 2020) 462 duration, frequency and intensity of PA. Thus, built-in sensors like accelerometer, GPS, altimeter 463 or gyroscope can provide continuous information about PA running in background while the user 464 performs their daily activities. Or you can also add simultaneous parameters from other external 465 sensors connected via Bluetooth, such as cardiorespiratory information using thoracic bands or 466 other wearables. The tendency should be the automatic recognition of the user activity without 467 interrupting their behavior, based on machine learning algorithms. The ecological information 468 could be completed with voice messages, image captures or brief text selections on the touch 469 screen made in real time. Current mobile devices already have the ability to process all this 470 information, but it will be necessary to persuade the user to carry their smartphone during the 471 whole day. In our review we have classified the EMA methodologies used for assessing PA behaviors and 475 found that 71% of studies specifically used the term "EMA". Just over half (51.3%) of studies 476 used mHealth technology, mainly smartphones, for collecting EMA data. An accelerometer was 477 the main built-in sensor used for collecting PA behavior by means of mHealth (69%). The 478 change of trend in the use of tools for EMA in PA coincides with the technological advances of 479 the last decade due to the emergence of smartphones and mHealth technology. 480 There is enough evidence to use the term mEMA (mobile-based EMA) when mHealth 481 technology is being used together with the EMA methodology for monitoring lifestyle behaviors 482 such as PA, in real time and in natural environments. We define mEMA as the use of mobile 483 computing and communication technologies for the ecological momentary assessment of health 484 and lifestyle behaviors. It is clear that the use of mHealth is increasing, but there is still a long 485 way to make the most of this technology in order to apply EMA to PA behavior. Thus, mEMA 486 methodology can help in the monitoring of healthy lifestyles under both subjective and objective 487 perspectives. The tendency for future research should be the automatic recognition of the user's 488 physical activity without interrupting their behavior. From our review, we suggest the use of 489 mEMA methodology with experimental designs, a duration between 1 and 4 weeks as an optimal 490 balance between habits information and low levels of dropout, a number of assessments per day 491 between 3 to 7, differentiating between working days and non-working days. The ecological 492 information could be completed with synchronized information from other sensors or wearables, 493 all managed through smartphone Apps. This methodology could be extended when EMA 494 combined with mHealth are used to evaluate other lifestyle behaviors.   Table 1. Inclusion criteria for this review.

Criteria
Description 1) EMA has to accomplish the following points: a) Instruments that collect data in real time. Participants reported the activities and/or behaviors, moods, etc. at the moment they are experiencing them or up to 24 hours, after the activity was carried out. b) In a natural environment. c) Repeated measures (2 or more measures). d) Self-reports and/or automatic recordings. e) Using both electronic devices and/or paper and pencil format.
2) PA is considered: PA has to be spontaneous or planned activity, carried out individually or collectively and that incorporates a physical effort component of any intensity.

3) Type of article:
All articles that provide original data on the use of EMA in PA (criteria 1 and 2) and are published in a scientific journal without taking into account the type and number of sample as well as the experimental design.
1 Manuscript to be reviewed