Original Articles
Workers’ perceptions of mHealth services for physical activity and mental health: A qualitative study using a text-mining method
2023 年 5 巻 1 号 論文ID: 2022-0022-OA
詳細
2023 年 5 巻 1 号 論文ID: 2022-0022-OA
Objectives: This study explored the perceptions of workers regarding mobile health (mHealth) services for physical activity and mental health. Methods: Single, online, one-on-one, 60-minute semi-structured interviews were conducted with workers living or working in Tokyo, Japan. The transcribed text corpus of the interviews was used as data to explore their perceptions. The top 200 nouns in the utterances were extracted and modeled using a bag of words (BoW) and embedded into a two-dimensional space. Results: The interviews extracted 476 utterances with 1,294 nouns from the 12 workers (7 males and 5 females). A total of 10 themes were estimated from the top 200 nouns. The participants mostly agreed that physical activity was effective in improving their mental health. They needed individualized, attractive, and easy mHealth services. Other estimated themes were: limited effects of non-leisure physical activity on mental health, lower priority on physical activity rather than sleep and rest, reluctance to share the data within the groups, and difficulties in wearing the devices to measure physical activity due to work rules. Conclusion: Workers’ perceptions of mHealth services were consistent with previous findings: the need for individualization, attraction, and ease of use. In the working population, mHealth services for physical activity and mental health should consider working conditions and focus on leisure-time physical activity. Social sharing might not be a mandatory technique because of the private topics and variety of associations between physical activity and mental health.
Physical activity is an important health-related behavior that can be used to treat and prevent depression and anxiety, including in the non-clinical population1). In the working population, its effectiveness has also been confirmed2). It is one of the most evident methods for the primary prevention of the onset of common mental disorders in the workplace3). Considering that both physical inactivity and depression/anxiety are widespread among workers4,5,6), promoting physical activity is a primary topic of preventive approaches to mental health among workers.
Mobile health (mHealth) is a growing methodology for behavior change interventions. It is an integral part of eHealth, which refers to the cost-effective and secure use of information and communication technologies to support health and health-related fields7). It enables us to deliver behavior change techniques, such as psychological education, goal-setting, planning, feedback, data sharing, and competition, to individual mobile devices8,9). Effective use of mHealth has the potential to increase access to evidence-based intervention, better inform consumers of care, and more actively engage them in treatment10), although there is no strong scientific evidence of its effectiveness11).
Motivation, engagement, and adherence to services influence the effectiveness of mHealth interventions. Previous systematic reviews have summarized qualitative data on the perceptions, beliefs, and experiences of users of mHealth services that promote physical activity12,13,14). Overall, individualized services appear to enhance users’ acceptability, engagement, and usability12). Additionally, reliable, trustworthy, interactive, multifunctional, easy, attractive, and gamified services can motivate individuals to adhere to mHealth interventions to promote physical activity12,13,14).
However, there is no summarized qualitative data on workers’ perceptions of mHealth services for physical activity to improve mental health. mHealth services to promote physical activity for workers’ mental health may include other themes that have not been covered by previous qualitative studies. For example, work-related physical activities, such as physically demanding jobs and commuting, are not positively associated with mental health15). If workers use mHealth services optimized for increasing overall physical activity, they may not recognize these services as valuable. Unique themes for workers would be useful for making more acceptable, appropriate, and feasible mHealth interventions. This study aimed to explore the perceptions of workers in mHealth services for physical activity and mental health.
This was a qualitative study using a text-mining method. We recently developed a prototype smartphone application for workers, to promote physical activity and improve depression and anxiety. The prototype of the app could: 1) predict the level of depression and anxiety on the next day, based on information about physical activity patterns and working conditions on the previous day; 2) notify users when their depression and anxiety were expected to be high; and 3) establish groups of similar users who share/compete for the data. We conducted semi-structured interviews with non-clinical or subclinical workers who are intended future users of the app, to hear their perceptions of mHealth services for physical activity and mental health as the preparatory stage of the development of the app. Single, online, one-on-one, 60-minute, semi-structured interviews were conducted. The transcribed text corpus of the interviews was used as data to explore workers’ perceptions. The Kitasato University Medical Ethics Organization, Japan approved the study protocol (B21-119).
ParticipantsParticipants in the semi-structured interviews were recruited from Tokyo, Japan using the snowball sampling method. We recruited workers engaged in the healthcare business and academic jobs, and workers that were recruited by us for a different study16). The eligibility criteria were: 1) working in public or private sector organizations; 2) living or working in Tokyo; and 3) having a personal smartphone. We excluded workers on sick leave during the interviews.
Procedure for data collectionBefore the interview, the interviewer (KW) explained the required time for the interview (i.e., 60 minutes), the background and scientific evidence for research in physical activity and mental health, and an overview of the application. Based on the explanation, informed consent from all participants was obtained via a paper-signed form that was mailed later. The participants were asked five questions: 1) “Do you feel that physical activity is effective in preventing mental health problems?”; 2) “What are your first impressions of the application?”; 3) “What do you think about the functions of the application that notify you when your depression and anxiety are expected to be high?”; 4) “What do you think about the functions of the application that make groups of similar users share/compete for data?”; and 5) “Are there any functions you would like to see added or any expectations for this research?”. After completing the interview, a voucher (QUO card) of 3,000 Japanese yen was distributed to each participant.
AnalysisAll utterances by the participants were extracted from the transcribed text corpus and divided into morphemes using morphological analysis. To explore participants’ perceptions, nouns in the utterances were extracted and summarized. For feature vectorizing, bag-of-words (BoW) modeling was conducted17). A feature vector developed by the BoW model was a large and sparse vector comprising occurrences of unique tokens of nouns. In the present analysis, the top 200 nouns frequently used in the utterances were extracted and modeled using BoW. Before vectorizing, demonstrative pronouns (e.g., this, it), proper nouns (e.g., participant names), and slang were excluded. Nouns that occurred in more than 10% of the utterances were also excluded because they were universally used and not expected to present unique topics. The feature vector was weighted using term frequency-inverse document frequency (TF-IDF) to reduce the weight of frequently occurring nouns18).
Linear and nonlinear dimension reduction techniques were employed to embed the sparse feature vector into an interpretable space. First, truncated singular value decomposition (SVD)19) was applied to decompose the feature vector of the 200 nouns into 10 dimensions, and secondary uniform manifold approximation and projection (UMAP)20) was used to embed them into a two-dimensional space. Furthermore, in the two-dimensional space, k-means++ clustering was conducted to classify the 200 nouns into 20 clusters to detect groups of nouns that co-occurred in the same contexts. The estimated themes of the utterances were qualitatively summarized from the two-dimensional embedding space and 20 clusters of nouns.
To implement these analyses, Janome 0.4.2. for morphological analysis21), Scikit-learn version 1.0.2 for feature vectorizing, truncated SVD, clustering22), and umap-learn version 0.5.3 for UMAP20) were used in Python version 3.9.12.
Interviews were conducted with 12 workers (Table 1). The majority of the participants were middle-aged (30–49 years) healthcare professionals, engineers, and academic researchers. Based on the interviews, 476 utterances with 1,294 nouns were extracted. Each utterance comprised 1–75 nouns.
| Demographic variable | N (%) |
|---|---|
| Age | |
| 30–39 years | 5 (41.7) |
| 40–49 years | 6 (50.0) |
| 50–59 years | 1 ( 8.3) |
| Gender | |
| Male | 7 (58.3) |
| Female | 5 (41.7) |
| Occupations | |
| Managers | 1 ( 8.3) |
| Experts/Engineers/Academics | 7 (58.3) |
| Clerks | 1 ( 8.3) |
| Sales | 1 ( 8.3) |
| Services | 2 (16.7) |
Figure 1 shows the top 200 nouns in utterances embedded in the two-dimensional space. K-means++ clustering classified the nouns into 20 clusters that included 6–16 words each (Table 2).
Nouns told by interviewees in the embedding space
| Number | Words |
|---|---|
| 1 | rice, gym, workout, most, prevention, other than, weight, forward-looking, acceleration, dislike, weekdays, meaning, walk, hobby, exercise |
| 2 | together, this week, others, leisure, hard, step, result, similarity |
| 3 | alert, image, pattern, weather, awareness, early, extreme, mood, the way of saying |
| 4 | advice, rest, teacher, labor, okay, facility, sleep, meal |
| 5 | professional, merit, mental, imagination, himself, period, management, point of view, difference, useless |
| 6 | stress, mental health, individual, interpersonal, assistance, normal, utilization, male, release, relief, part, relationship |
| 7 | depression, data, health, prediction, reference, roughly, information, enjoyment, activity, life, workplace, behavior, body |
| 8 | energy, daycare, message, anxiety, human, concrete, transformation, realization, family, daytime, first, stage, registration, habit, shopping |
| 9 | group, input, comparison, occupation, self, oneself |
| 10 | COVID-19, tele-, work, duty, case, time, full house, commuting, train |
| 11 | helper, month, clerical work, nursing care, job, day-off, unit, basic, expert, industry, every day, feeling, fatigue, position, week, achievement |
| 12 | graph, company, content, tomorrow, caution, explanation, tone, baseball |
| 13 | app, theme, interaction, this time, answer, ingenuity, research, word, record, question, notification, development |
| 14 | application, questionnaire, grouping, telework, general, condition, use, midway |
| 15 | smart, phone, positive, refreshing, weekend, intensity, influence, purpose |
| 16 | switch, negative, office, attendance, place, maybe, sense, recently, itself |
| 17 | level, importance, business, impossible, state, correlation, sureness, next day, association |
| 18 | comment, today, tendency, vigor, diary, day of week, function, past |
| 19 | user, ranking, bad condition, holiday, sharing, like, situation, player |
| 20 | AI, reflection, female, improvement, season, effectiveness, really, evaluation, in the way |
A total of 10 themes were estimated, consisting of several clusters, which seemed to represent the perceptions of workers regarding mHealth services for physical activity and mental health (Table 3). Theme 1 to Theme 5 represented the participant’ perception about the association between physical activity and mental health. Theme 6 to Theme 10 represented the perception about the developed application.
| Number | Theme | Related cluster number: related words | Examples of the raw utterances |
|---|---|---|---|
| 1 | Physical activity is good for mental health | CL1: gym, workout, prevention, forward-looking, dislike, walk, hobby, exercise CL6: stress, mental health, release, relief CL15: positive, refreshing, weekend, influence | -Of course, its good effects on mental health when doing yoga. I’ve been doing it for a long time because I felt that it would make me kinder or a little more generous. -Working out definitely makes me feel positive. After finishing the muscle training, it feels really good. I’m going to be more positive than necessary. I think that I’m so glad I did and I’m forward-looking. There is a positive feeling that I want to train more, although the positive feelings won’t last long. -I’m not conscious of prevention, however, for example,I basically don’t dislike activities, rather I like, such as climbing or camping. After all, going to the mountains and using energy in a productive way makes me feel very refreshed. -Speaking from my own experience, when I’m stressed at work, I think taking a walk or walking may relieve stress. -Well, on either Saturday or Sunday, I do a little physical activity to refresh. |
| 2 | Workers cannot think about physical activity during work on weekdays | CL1: rice, gym, weekdays, meaning, walk, hobby, exercise CL11: job, basic, every day, feeling, fatigue, week CL17: business, impossible | -Apple Watch has a function to notify like standing up once an hour. It’s impossible now. For example, on the train, or having a lot of clerical work. -After all, on weekdays, I just go to work, come home, cook and eat, and then go to sleep. Is there anything I can do something between eating and going to bed? People who are overly conscious go to the gym, but many people don’t go. -Hmm, my feeling is... I am not very aware of it. As I said, I usually work until around 23:00 on weekdays, and I’m certainly active, and moving around the city because I have a lot of business trips, however, I’m not really conscious that it will change my mind. -When I have work on weekdays, I don’t think much about exercise. It was about walking one station more. Other than that, I don’t really think about it. I can’t do it, and even I don’t think about it. |
| 3 | Non-leisure activities are not recognized as physical activity good for mental health | CL1: walk, exercise CL8: energy, family, habit, shopping CL10: COVID-19, tele, work, time, full house, commuting, train CL16: negative, office, attendance, place | -After all, commuting doesn’t count as exercise. I don’t think it feels better. When commuting, and when the station is far from home, even if I think I’m walking a little bit, it doesn’t mean much. -For commuting, I change trains twice and ride for about 50 minutes. Certainly, it’s a good time to refresh me while commuting, especially when coming home, to bring back nothing that happened at work. -Before and after telework, my opinions changed a lot. Before I started teleworking, when commuting was the norm, I did think it was good to have time to change my mind, and to have a house that is a little away from my workplace. Since I start teleworking, the commuting time has been zero. When that happened, I began to wonder how stressful this commute was. -Shopping is a daily thing, so nothing to feel. Just I think it would be good if there were fewer people, and shopping for daily necessities is like that. -Well, maybe I’m not aware of doing something with energy, such as “I’m feeling down, so let’s go for a walk.” |
| 4 | Sleep and rest are more important than physical activity | CL4: advice, rest, teacher, labor, okay, facility, sleep, meal | -Well, if you tell me directly that I’m going to be depressed today, it might not be so, but I’m dragged by it, and I feel like I’m not feeling well today. On the other hand, if you receive a caring message such as “Please be careful” or “Take a rest because you’re getting tired,” I can think that I need to take care of my body. -Employers are quite interested in sleeping time. For better or worse, whether we can sleep or not, I feel that it is information that they want. So, if I can see both rest and activity in the app, I probably want to access the app. -When they were told only about physical activity, they were like, “Even if you say that, I’ve been up all night this week.” I don’t know if there are people like that, but, they may think that, “That’s right, No wonder I am negative.” so, it is more reliable if the app included the part about rest. -Also, questions such as, when do you really exercise, such as morning, noon, or night? It’s like, “what do you do to relieve stress?” Meal, eating delicious foods, watching movies. |
| 5 | Lifecycles differ base on sex | CL5: period, difference, useless CL6: male, relationship CL11: month, unit, week CL20: female, season | -It may be unique for women, but after all, it’s hormone balance, and it is a season I’m feeling a little depressed, no matter what I do, it’s useless. -It is a cycle, I think it’s related. It’s unique to women, and I think it’s useless no matter what you do within a certain period of time. -I feel that males and females are a little different. As I said earlier, women have menstruation once a month, so I wonder if it will be some kind of unit. It may be a month or two months or something like that. Males are not like that. Extremely, I think it’s shorter in males. Even if it’s less than four weeks, for example, one or two weeks. |
| 6 | Feedback must be comprehensible and non-invasive | CL3: alert, image, pattern, weather, early, mood, the way of saying CL12: graph, content, tomorrow, caution, explanation, tone | -It’s better to be cared, than to be alerted that depression or anxiety will come out. It’s still easier to accept to be told something like, “Are you okay?” -The way of saying it is really important. Even if you say the same thing, the way you receive it will be completely different just by using slightly different wording. Various and non-invasive comments are good. -When I receive a message like this from a professional, for example, I get really defensive, but when a cute character whispers to me, it makes me realize the message. -Like an Apple Watch, if comments are given by an avatar like a geometric pattern, I don’t feel like I’m being told. In my opinion, I think the organic matter is good. Such as people, animals, and characters. -For example, when it comes to the weather forecast, I have an image of a weather girl kindly telling me things like, “Don’t forget your folding umbrella” and “go home early.” |
| 7 | Messages should be changed according to their occupation | CL9: group, comparison, occupation, self, oneself CL11: clerical work, nursing care, job, day-off, unit, expert, industry, position, week | -It may be a bit bothersome, but please separate them by occupation. I think I will look at it if the most active person of this week is displayed within a group of care workers. -It would be nice if there was a function to change the advice given to me depending on my occupation. When using various apps, the advice is generally the same. -If they don’t have similar industries and ways to spend your leisure time, they’ll have a completely different way of spending your leisure time. I think the way to find a good friend is to find someone whose job is similar. |
| 8 | Comparing with others is not very useful | CL2: together, others, leisure, hard, step, result, similarity CL9: group, input, comparison, self, oneself CL19: user, ranking, bad condition, sharing | -My own way of thinking is, “Others are others, I am myself.” I probably won’t use this function. After all, people spend their leisure time differently. -Is it a grouping function that allows you to browse similar people? Looking at this, it’s a bit difficult to change my mental health for the better. It is hard to understand by looking at this, why good people are good, and what kind of actions they do. -If you ask me if I want to compare myself with similar men, for example, it would be nice to compete for the number of steps, but I don’t want to share my mental health with others. I’m not that strong of a person. -I thought that, I might not like the comparison with a target person as a role model. |
| 9 | Sometimes mHealth services are not available due to working | CL11: job, day-off, basic CL15: smart, phone, weekend | -Basically, I don’t wear a smartphone during work. I don’t carry it around. -I have a Fitbit, but I only wear it on my days off, and when I commute. I think that care workers, medical workers, firefighters, and the police probably don’t carry it around. -I didn’t carry it before. I consciously tried to carry it around when I participated in this study, but I don’t normally carry it around. Then it becomes an issue of the usefulness of the app. |
| 10 | Self-monitoring using diaries is useful | CL18: comment, today, tendency, diary | -It may be burdensome, but I comment on why I got over today’s bad condition. -I think this app is easy to use if he can write a diary and understand that this kind of thing has caused him to feel this way. |
CL, cluster.
Owing to the translation from Japanese to English, some words in the raw utterances are not inconsistent with the words in the clusters.
A summary of the 10 themes that represent the perception of workers with the noun extracted from their utterances was as follows:
Theme 1: Physical activity is good for mental healthMost participants perceived physical activity as being effective in improving mental health and relieving job stress. Some participants exercised before or after work, and on days off. Exercises included yoga, strength training, climbing, camping, jogging, and walking.
Theme 2: Workers cannot think about physical activity during work on weekdaysThe participants understood that physical activity was good for health, however, they were too busy to spend time on it. Sometimes, they were unable to adopt suggestions or recommendations from mHealth services.
Theme 3: Non-leisure activities are not recognized as physical activities good for mental healthParticipants considered that the effects of physical activity on mental health differed by the domain (work-related, household, transportation, and leisure). Although their perceptions around commuting were inconsistent, those concerning transport, work, and household physical activity were not recognized as good activities for mental health.
Theme 4: Sleep and rest are more important than physical activityWhen participants discussed mental health, especially when they were stressed, physical activity was not always considered the most important topic. Rather, sleep and rest were prioritized to prevent the deterioration of mental health.
Theme 5: Lifecycles differ based on sexAn important category for tailored services was sex. Both male and female participants mentioned that women have unique cycles of menstruation and that they influenced the association between physical activity and mental health. They thought that sex differences should be considered when designing the mHealth service.
Theme 6: Feedback must be comprehensible and non-invasiveA notification for high levels of depression and anxiety may be invasive and hard to accept. From whom and how to get feedback were also important themes for the participants. In the interviews, they volunteered ideas, such as giving feedback using metaphors of the weather (i.e., rainy means high levels of depression) and giving feedback via a cute character.
Theme 7: Messages should be changed according to their occupationThe participants considered their occupations as important information for individually tailored services. They believed that workers with similar occupations had similar lifestyles, physical demands, and physical activity.
Theme 8: Comparing with others is not very usefulThe participants considered their mental health status to be private information. A common strategy for promoting physical activity is sharing and competing for the levels of physical activity among users. However, they were reluctant to share data to improve mental health via physical activity.
Theme 9: Sometimes mHealth services are not available due to workingAccording to feasibility, some workers could not use a smartphone or wearable device due to workplace rules. This may lead to low validity and reliability of physical activity measurement.
Theme 10: Self-monitoring using diaries is usefulSeveral workers mentioned that keeping a diary was useful for self-monitoring mental health, although it conflicted with the ease of data input.
This study explored workers’ perceptions of mHealth services for physical activity and mental health. The clusters of nouns from the participants showed impressions and the need for mHealth services in this area. Overall, participants agreed that physical activity was effective in improving mental health. This was consistent with previous findings12,13,14) that users needed research for individualized or tailored mHealth services. The following unique themes were also estimated: limited effects of non-leisure physical activity on mental health, lower priority on physical activity rather than sleep and rest, reluctance to share the data within the groups, and difficulties in wearing the devices to measure physical activity due to work rules. The unique themes for workers would be useful for making more acceptable, appropriate, and feasible mHealth interventions.
As recommended by previous findings, mHealth services for physical activity and mental health should also focus on individualization, attraction, and ease of use. Compared with previous systematic reviews12,13,14), workers also needed mHealth services to be tailored and individualized. Their perception that messages should be changed according to their occupation might be the primary factor to be tailored, and it was supported by the previous finding that levels of physical activity vary by occupation23). Based on the theme of lifecycles differ based on sex, sex differences also need to be studied in mHealth services for physical activity and mental health. Physiological regulation is different for women, especially during puberty, menstruation, pregnancy, and menopause24). Feedback must be comprehensible and non-invasive and self-monitoring using diaries is useful, might also be similar themes to those mentioned in previous studies, and might be related to acceptability, appropriateness, and feasibility of the services. Although not included in the 10 themes, the ease of using mHealth services was also mentioned by the participants in the present study (e.g., CL9, input: That’s right, I can use the app if I get advice automatically, or get by with just a little input at the beginning).
Unique themes in the working population were also extracted: workers cannot think about physical activity during work on weekdays; non-leisure activities are not recognized as physical activity good for mental health; comparing with others is not very useful; and sometimes mHealth services are not available due to working. Busyness or lack of time are consistently considered barriers to adherence to physical activity25), and it is apparent that busyness is due to participants’ jobs. In addition, recommendations for promoting occupational and transport activities may mislead them15). In the working population, mHealth services for physical activity and mental health should consider working conditions and focus on leisure-time physical activity. Tailoring the mHealth services to the workplace environment, such as work rules, is also important to support behavioral changes among workers26). Interestingly, the participants were reluctant to share their data on physical activity and mental health with other workers. Social sharing is an important behavioral change technique to increase physical activity in mHealth research8). However, private topics of mental health and various associations between physical activity and mental health might influence acceptability of social sharing. Therefore, social sharing might not be a mandatory technique in mHealth interventions for physical activity and mental health among workers.
Strengths and limitationsThe unique themes for mHealth services for physical activity and mental health among workers were the strengths of the present study, as they had not been considered in the previous literature. These themes might be useful for making mHealth services more acceptable, appropriate, and feasible. For example, messaging and advising — considering working conditions, such as occupation, busyness, and work rules — or focusing on leisure time and off-days can enhance the quality of mHealth services. However, this study has several limitations. Due to the snowball sampling method, the interview could not cover all the themes of the workers’ perception. The majority of the participants were professionals and, therefore, might be more knowledgeable about physical activity and mental health than the general working population. Workers that work in industries such as agriculture, forestry and fisheries, construction, and transportation, or who have lower literacy and motivation for health, should be interviewed. As the procedure of feature vectorizing, embedding from the interviews, and summarizing the estimated themes was decided by a single investigator (KW), the results of text mining were biased.
Workers’ perceptions of mHealth services for physical activity and mental health were consistent with previous findings: the need for individualization, attraction, and ease of use. Unique themes were also identified. In the working population, mHealth services for physical activity and mental health should consider working conditions and focus on leisure-time physical activity. Social sharing might not be a mandatory behavioral change technique in mHealth interventions for physical activity and mental health among workers.
We would like to thank Editage (www.editage.jp) for the English language editing.
This study was supported by Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (JP20K19671) and AMED (JP21de0107006). The funder had no role in the study design, collection, analysis and interpretation of data, writing of the report, or decision to submit the article for publication.
KW and AT declare that there are no conflicts of interest regarding the present study.
Qualitative data are not available for sharing because they include personal information.
Ethical approval was obtained from Kitasato University Medical Ethics Organization, Japan (B21-119).
KW contributed to research design, data collection, analysis, and drafting of the manuscript. AT contributed significantly to the revision of the draft, and all of them approved the current manuscript.
