Effectiveness of mobile application for menstrual management of working women in Japan: randomized controlled trial and medical economic evaluation

Abstract

Aims: Hormonal imbalances can affect a woman’s physical and mental condition, impacting her daily life and productivity. A mobile application, “Karada-no-kimochi”, predicts the menstrual cycle based on recorded data, and provides the information regarding menstruation. This study investigates the effectiveness of the application from health condition and labor productivity loss perspectives, and evaluates its cost-effectiveness for employed women in Japan.

Materials and methods: A randomized controlled trial (RCT) was performed to compare the use of the application and non-intervention in female workers (20–45 years) based on an online survey. A discrete-event model based on the RCT data, published literature, and claims data was used for the cost-effectiveness analysis.

Results: The intervention and non-intervention groups comprised 612 and 914 women, respectively. In the intervention group, the incidences of depression (0.16%) and dysmenorrhea (0.33%) were significantly lower than those of the non-intervention group (0.77% and 1.31%, respectively) in the third month. While labor productivity and absenteeism varied, presenteeism showed almost no change across groups. The quality-adjusted life year in the intervention group was 6.84 (0.07 higher than that in the non-intervention group). By analyzing medical expenses and making assumptions regarding productivity, we estimated that the aggregate of medical expenses, productivity loss, and application fee for the intervention group could be lower than that of the non-intervention group by over JPY 130,000 (USD 1,170, as of December 2017) per individual.

Limitations: The subjects included were willing or relatively willing to use the application. Because all outcomes were based on online surveys, the results depended on the objectivity and preciseness of the reports provided by users, and those with severe symptoms might not provide reports.

Conclusions: The results suggest that using the application is cost-effective and might reduce the incidence of dysmenorrhea and depression.

KEYWORDS:

• Mobile application
• randomized controlled trial
• cost-effectiveness
• menstruation-associated symptoms
• self-health management
• menstrual cycle

JEL Classification Codes:

• I10
• I12
• O33

Introduction

A woman’s physical and mental condition can be affected by imbalances in hormone secretion. This imbalance can manifest through various symptoms, such as premenstrual syndrome (PMS); and premenstrual dysphoric disorder, which has symptoms similar to depression; and can sometimes cause dysmenorrhea, which is characterized by lower abdominal pain and back pain during menstruation. It has also been reported that mental stress is related to menstrual irregularity and PMS¹; in addition, a close association is suspected between the menstrual cycle and psychiatric symptoms. The prevalence of PMS was reported for its variance across various geographies: 12% of women in France, 98% in Iran, and 47.8% as a pooled value in a systematic review². In Japan, 74% of menstruating women reported suffering from associated symptoms³. These symptoms not only affect an individual’s daily life, but also the economy. This is especially relevant in modern society, in which an increasing number of women work. The annual economic burden due to menstruation-associated symptoms in Japan was reported as JPY 683 billion (USD 6.15 billion, as of December 2017), of which 72% was due to productivity loss³. Therefore, identifying the causes of these symptoms and providing appropriate treatment are important.

Gynecological examination of women experiencing menstruation-associated symptoms was reported to alleviate these symptoms, leading to an improvement in daily life⁴. A previous study compared the use of a mobile application for easing PMS that records the basal body temperature, menstrual cycle status, and menstruation-associated symptoms input by the user, and sends them to doctors for review, with the conventional paper-based recording system⁵. Based on feedback from both groups, some advantages of using the application included becoming self-aware, feeling relieved, and better understanding of the problem when consulting doctors as their condition was recorded per se. Moreover, other advantages, including the ability to input data promptly, ease of viewing data, and feeling more connected with the doctors were experienced by the application users. The researchers of this study concluded that the use of smartphone applications may be effective in PMS patients and potential PMS patients. This is especially relevant given that the use of smartphones is quite common today. A mobile application may enable women to understand their own menstrual cycle and potentially predict their menstruation-associated symptoms based on self-recording, leading to self-health management and prevention or reduction of the associated symptoms.

In recent years, various types of mobile applications have been developed and used for managing and preventing diseases. Randomized controlled trials (RCTs)^6–8 and cost-effectiveness analyses⁹ have been conducted on some applications, leading to their certification as medical equipment^10–12 and approval for insurance reimbursement¹². Although the development and use of applications for managing menstruation are increasing, few studies have been conducted to evaluate their effects.

“Karada-no-kimochi”, an application provided by docomo Healthcare (Tokyo, Japan), predicts the menstrual cycle and ovulation day based on the data of previous menstrual dates and basal body temperatures recorded by a user. Moreover, it advises the user regarding the four stages of the menstrual cycle, each of which is characterized by different symptoms arising from changes in the hormonal balance. In particular, the application provides pertinent information regarding the condition and recommends meal plans and exercises that alleviate symptoms and support self-health management.

In this study, we investigated the effectiveness of this mobile application in reducing mental and physical disorders, and labor productivity loss due to menstruation-associated symptoms for working women in Japan. Further, we also examined its cost-effectiveness through analysis.

Methods

Study design

A randomized controlled parallel-group comparison study was performed to investigate the effects of the use of the smartphone application “Karada-no-kimochi” on working women in Japan. The subjects were randomly divided into the application-use group (intervention group) and the non-intervention group (control group). The number of individuals allocated to the intervention group and control group were 650 and 1,000, respectively. Subject screening was conducted in January 2017, and surveys were conducted in January, February, March, and April 2017.

A cost-effectiveness analysis of the application usage was performed using a discrete-event model based on the data from this RCT, published studies, and claims data provided by JMDC (Tokyo, Japan). The JMDC database comprises data regarding individuals (∼3 million) covered by health insurance societies, including company employees and their family members.

This RCT was conducted in accordance with the Ethical Guidelines for Medical and Health Research Involving Human Subjects by the Ministry of Education, Culture, Sports, Science and Technology and the Ministry of Health, Labor and Welfare, and was approved by the non-profit organization MINS International Review Board (Tokyo, Japan) in December 2016. The protocol was registered in the UMIN Clinical Trials Registry (No. UMIN000025513). Informed consent was obtained from all the participants online. The cost-effectiveness analysis was conducted in accordance with the official guidelines for economic evaluation of the cost-effectiveness of drugs/medical devices¹³. The reporting style followed the CONSORT statement¹⁴ and the Consolidated Health Economic Evaluation Reporting Standards¹⁵.

RCT

Participants

The subjects were recruited from an online panel (comprising 1,400,000 individuals) of a research company, Intage Inc. (Tokyo, Japan). The subjects who met the following criteria were selected based on the screening survey on the internet.

1. Aged 20–45 years;

2. Selected “company employee”, “company officer or manager”, “civil servant or association staff member”, “self-employed”, “freelancer or professional”, or “temporary employee or contract employee”, and “fulltime worker” as occupation;

3. Answered as “not being pregnant”;

4. Had never downloaded or registered for either “Karada-no-kimochi” or “Rhythm Note” (currently being integrated into “Karada-no-kimochi”);

5. Smartphone user; and

6. Answered that they would “like to use” and “relatively like to use” for the question asking whether they were interested in the smartphone menstruation cycle management application.

Those who declared that they were under treatment for depression, dysmenorrhea, or PMS at the time of screening, or had visited a doctor owing to at least one of these diseases in the past month before the baseline survey were excluded from the analysis.

Intervention

The smartphone application “Karada-no-kimochi” was provided to the subjects in the intervention group. The user can record their menstrual dates, basal body temperatures, and their mental and physical disorders. Based on the recorded data, the application predicts the menstrual cycle, i.e. it predicts the next day of bleeding, the length of the menstruation period, and the ovulation day; additionally, it provides information regarding the condition in the stage of menstrual cycle and recommends appropriate changes in lifestyle, including recommended food and exercises to alleviate symptoms. The details of the application are shown in Figure 1.

Figure 1. Smartphone application “Karada-no-kimochi” (written in Japanese), “Voice” (A and B) shows information regarding the condition and recommendations of change of lifestyle such as meal plans and exercises to alleviate the symptoms. It is expected that users can obtain the right information and address their symptoms themselves. Users record their mental and physical disorders (headache, stomachache, irritation, depressed mood, etc.) using the “Symptoms stamp” (C). Their basal body temperature is shown as a “graph” (D). Based on these records, the users can understand the time period in which the symptoms of PMS appear, and take measures beforehand.

The subjects downloaded the distributed application themselves. Use of the application was defined as the application being downloaded and the user logging in. We did not consider the features accessed and the frequency of use, although users were expected to keep using it constantly during the study period.

Control

The subjects in the control group did not take any interventions, and took the same survey at the same time as those in the intervention group.

Outcomes

We obtained all outcomes from questionnaires. Subjects were asked to answer the questions on the web before beginning the use of the application (baseline survey). They repeated this process in the first, second, and third month after using the application.

Primary outcomes were labor productivity, incidence of diseases, and self-evaluated severity of depression. Labor productivity was evaluated based on the Work Productivity and Activity Impairment Questionnaire (WPAI)¹⁶. Absenteeism was evaluated as work time (hours) missed owing to health problems in the last 7 days. Presenteeism was evaluated as the time when productivity was impaired owing to health problems during work hours in the last 7 days, answered with a score (0–10). Severity of depression was evaluated based on the Patient Health Questionnaire (PHQ-9) that comprised nine questions. Given that PHQ-9 is applied to assess the symptom severity of depression in the guidelines provided by institutions for health technology assessment, such as the National Institute for Health and Care Excellence, and that it can be self-administered and has a Japanese version, we considered it to be appropriate for this study. The replies were scored from 0–3^17,18, and the severity was classified as one of the five levels based on the total score (Supplementary Table S1).

Physical conditions were also reported as secondary outcomes. The frequency of each condition was selected from four levels, and the average score was calculated for each condition. The status of occupation and query on pregnancy were also included in every questionnaire. The questions are listed in Supplementary Table S2.

Sample size

To evaluate the sample size, the incidence of dysmenorrhea and PMS in the control group were estimated as being 4% each, and their reduction effect by using the application was set to 25% each. Based on these values, the number of individuals per group required to have a power of 80% to detect a difference at a level of significance of α = 0.05 between groups was calculated to be 4,175. However, we assigned only 650 individuals to the intervention group. This number was chosen such that the application download was not considered as promotional advertising, based on the rules of the research company in this study. Because there was no limit for the number of individuals in the control group, we assigned 1,000 individuals, a larger number than intervention group. This number was selected to maximize the chance of detecting a significant difference in outcomes other than the incidence of dysmenorrhea and PMS. This sample size may be insufficient to detect a significant difference in the incidence of dysmenorrhea and PMS; however, it is considered to be in a reasonable range for the evaluation of other outcomes such as labor productivity, severity of depression, and physical conditions.

Randomization

Classification of subjects was conducted by ANTERIO Inc. (Tokyo, Japan) using a completely randomized method. Random values generated by a computer were given to subjects, and, based on these values, subjects were assigned to either the intervention or control group.

Analysis

Each outcome was compared between the intervention and control groups. Based on the principle of intention-to-treat, the total number of subjects in the beginning of RCT was used as the denominator. Statistical significance was evaluated by using the test of quotient difference or Welch’s t-test, and the difference was considered to be significant if the p-value was <0.05. Analyses based on the per protocol set (PPS) using the number of respondents as the denominator were also performed as a sensitivity analysis.

Cost-effectiveness analysis

Condition

The target population comprised premenopausal female workers (aged 20–45 years) in Japanese companies. The age distribution was taken as that of the JMDC database. The study was conducted from the perspective of the health insurance society and companies; consequently, productivity loss was included in the analysis.

Intervention was defined as the use of the smartphone application. Control was defined as non-intervention because no corresponding application was provided as a service by the health insurance society or company. The time horizon was set as the period until the end of age 45, which corresponded to the target age of the application. Discount rates of cost and effectiveness were set to 2% per year as per the guidelines¹³.

Analysis

This study is the only RCT conducted for this application. Therefore, we used the reduction in incidence of dysmenorrhea and depression in the third month after the initiation of application use, in which statistically significant differences were observed, as the effectiveness measure in this study. The reduction in the incidence of these diseases calculated from the result of RCT was further adjusted by values obtained from the JMDC database (from January 2005 until August 2015). Data from the JMDC database and values from previously published studies were used to obtain parameters aside from those obtained from the RCT in the current study. In the JMDC database, dysmenorrhea patients were defined as individuals who had a diagnosis code of N94 in ICD-10¹⁹ at least once, and depression patients were those who had a diagnosis code of either F32 or F33 at least once. We assumed that there were no adverse effects from using the application.

Cost (calculated in Japanese yen) was defined as the sum of the application fee, medical expense, and productivity loss. The application fee was set at JPY 400. The medical expenses were calculated collectively by age group every 5 years using the JMDC database. The wage for calculation of productivity loss was set to JPY 304,000 based on the average wage across all industries, all ages, and for both genders obtained from the 2016 Basic Statistical Survey on Wage Structure²⁰. Regarding presenteeism, rates of 3.4% for dysmenorrhea (5.9% for disorders due to irregular menstruation and PMS²¹ was applied for a quarter month), 11.4% for depression²², and 2.6% for others²² were applied by referring to previous literature. Absenteeism was calculated for each disease as follows: the number of absences was calculated as the sum of the number of days of inpatient and outpatient care for a disease per year based on the JMDC database, and then divided by the number of work days per year (251.6 days). Resignation rate due to a disease, which was calculated from the JMDC database, was included in the productivity loss.

The quality-adjusted life year (QALY) was used as an outcome measure. Based on previous studies, quality-of-life (QoL) scores of 0.89 for dysmenorrhea (0.74, based on a score of 0.73²³ converted using Frank’s model²⁴ or 0.74²⁵ was applied for a quarter month), 0.5 for depression (either 0.53 and 0.55²⁶ or 0.46²⁷ for depression), or 0.94 for others (calculated based on Shiroiwa et al.’s²⁸ study) were applied. For both diseases, increase in mortality owing to the disease was considered to be zero.

A discrete-event-model that included the incidence of dysmenorrhea and depression was used for prediction in the case of working females who attain the age of 45 years.

Sensitivity analysis was carried for scenarios characterized by a (1) 20% increase in the application fee, (2) 20% reduction in the decrease in incidence of dysmenorrhea owing to application use, and (3) 20% reduction in the decrease in incidence of depression owing to application use.

Results

RCT

Participants and attributes of each group

The total number of individuals who met the criteria was 1,729, and the number of individuals who did not meet the criteria was 16,988. The survey was conducted on individuals through an online panel of a research company. Information regarding the criteria, such as history of application use and willingness to use it, was not available to the research company. Therefore, the percentage of individuals who did not meet the criteria in this study is large compared to conventional RCTs. Among the individuals who met the criteria, 1,650 individuals were selected. From among these 1,650 individuals, 650 and 1,000 individuals were randomly assigned to the intervention and control groups, respectively. From amongst these individuals, 612 and 914 individuals from the intervention and the non-intervention groups, respectively, were considered for analysis (Table 1). All of them responded to the baseline survey. The response rate declined throughout the study; however, the percentages remained similar between groups (1st month 83.2 vs 75.7, 2nd month 77.0 vs 70.9, 3rd month 76.3 vs 67.8, for control vs intervention).

Table 1. Attributes of each group at the time of screening.

	Control group	Intervention group
Number of individuals	914	612
Mean age (in years)	33.0	33.0
Age group (%)
20–29 years old	34.5%	35.6%
30–39 years old	45.5%	44.4%
40–45 years old	20.0%	19.9%
Occupation group (%)
Company employee	65.0%	64.5%
Company officer or manager	1.0%	1.1%
Civil servant or association staff member	7.5%	7.8%
Self-employed	2.4%	2.6%
Freelancer or professional	3.8%	4.1%
Temporary employee or contract employee	20.2%	19.8%

Outcomes

Absenteeism changed in both groups over the experimental period, and no significant decrease in the intervention group was observed (Table 2). Presenteeism did not change over the same period in both groups (Table 2).

Table 2. Absenteeism and presenteeism.

	Control group	Intervention group
	(n = 914)	(n = 612)
Absenteeism (hours, Mean ± SE)
Baseline	2.2 ± 0.36	1.6 ± 0.28
1st month	1.4 ± 0.23	1.6 ± 0.28
2nd month	1.6 ± 0.24	0.8 ± 0.16
3rd month	2.1 ± 0.27	1.4 ± 0.26
Presenteeism (hours, Mean ± SE)
Baseline	3.8 ± 0.09	3.7 ± 0.11
1st month	3.8 ± 0.09	3.8 ± 0.11
2nd month	3.7 ± 0.09	3.6 ± 0.10
3rd month	3.8 ± 0.09	3.7 ± 0.11

Abbreviation. SE, standard error.

The percentages of patients who said that they visited the doctor because of either dysmenorrhea, PMS, or depression are shown in Table 3. The incidence of depression in the second and third month, and dysmenorrhea in the third month were significantly low in the intervention group. The incidence of PMS was significantly lower in the intervention group in the second month; however, this was less prominent in the third month. The incidences of dysmenorrhea in the third month, and depression in the second and third month, were confirmed to be significantly lower in the intervention group as compared to the control group by sensitivity analysis with PPS (Supplementary Table S3).

Table 3. Incidence of each disease.

	Control group	Intervention group	p-value
	(n = 914)	(n = 612)
Incidence in each month (n (%))
Dysmenorrhea
1st month	8 (0.88)	4 (0.65)	0.311
2nd month	10 (1.09)	5 (0.82)	0.290
3rd month	12 (1.31)	2 (0.33)	0.013*
PMS
1st month	10 (1.09)	6 (0.98)	0.414
2nd month	12 (1.31)	3 (0.49)	0.040*
3rd month	7 (0.77)	6 (0.98)	0.331
Depression
1st month	4 (0.44)	2 (0.33)	0.364
2nd month	7 (0.77)	1 (0.16)	0.035*
3rd month	7 (0.77)	1 (0.16)	0.035*
Total incidence during the entire study period (n (%))
Dysmenorrhea	21 (2.30)	11 (1.80)	0.316
PMS	20 (2.19)	14 (2.29)	0.396
Depression	12 (1.31)	2 (0.33)	0.028*

Abbreviation. PMS, premenstrual syndrome.

* p < 0.05.

The average value of the PHQ-9 score was significantly lower in the second month for the intervention group, and continued to be lower, but not significantly, in the third month (Table 4). Based on the classification of severity according to the PHQ-9 score, no obvious overall trend was found (Supplementary Table S1).

Table 4. PHQ-9 score^17,18.

	Control group(n = 914)	Intervention group (n = 612)	p-value
	Score (Mean ± SE)
Baseline	14.2 ± 0.16	14.1 ± 0.19	0.367
1st month	14.1 ± 0.16	13.9 ± 0.19	0.272
2nd month	14.3 ± 0.18	13.6 ± 0.19	0.005*
3rd month	14.2 ± 0.18	13.8 ± 0.19	0.057

Abbreviation. SE, standard error.

* p < 0.05.

Considering the average value regarding the physical condition of the individual, no significant difference was observed at baseline, but some significant differences were seen between the intervention and control groups after the start of the study (Supplementary Table S4).

Cost-effectiveness analysis

Calculation of parameters

The proportion of the target population (between 20 years and less than 45 years, grouped into 5-year intervals) in the JMDC database relative to the entire population is shown in Supplementary Figure S1. The proportion of each group is not different from the data (2015) published by the Statistics Bureau of the Ministry of Internal Affairs and Communications²⁹.

The incidence of dysmenorrhea calculated from the JMDC database appears to have increased over the years (Supplementary Figure S2). From 2013 onward, the rate has almost always been over 1% (Supplementary Figure S2A); therefore, the rate of 1.31% in the third month for the control group in the RCT is considered to be reasonable. We assumed that the incidence did not change across different age groups; therefore, the incidence of the third month in the RCT was used as the monthly incidence of dysmenorrhea, regardless of age. The incidence of depression obtained from the JMDC database was 1.5% in 2012, and the rate tended to increase over the years (Supplementary Figure S2B). The incidence of the control group in the third month of 0.77%, in the RCT is approximately half of the value compared with that of the JMDC database. We assumed that this difference was owing to an RCT bias. Therefore, we used the rate from the JMDC database, i.e. 1.5%, as the incidence of depression in the control group, and 0.3% (1.5% × 0.16%/0.77%) in the intervention group across all ages.

The medical expense per insured person per month calculated from the JMDC database was JPY 5,000 for the group in their twenties, JPY 7,000 for those in their early thirties, and JPY 8,000 in their late thirties and early forties (Supplementary Figure S3A). The relative cost of medical expense was 2–2.5-times higher in individuals with dysmenorrhea for all age groups, ∼ 4.5-times higher for those with depression in their early twenties, 3–3.5-times higher for those with depression in their late twenties till early forties compared with those without these diseases (Supplementary Figure S3B).

The resignation rate calculated from the JMDC database was moderately higher for dysmenorrhea patients across all age groups, more than twice in depression patients in their early twenties, and approximately twice for those in their late twenties and thirties when compared to that of individuals without these diseases (Supplementary Figure S4). Absenteeism calculated based on the JMDC database was 2.7% for dysmenorrhea, and 3.9% for depression (Supplementary Table S5).

Cost-effectiveness analysis

The QALY in the intervention group was 6.84, which is 0.07 higher than that in the control group (Table 5). The total cost of medical expenses (JPY 669,000 for the intervention group and JPY 694,000 for the control group), loss of productivity (JPY 144,000 for the control group), and application fee (JPY 35,000 for the intervention group) was less for the intervention group than for the control group by more than JPY 130,000 (USD 1,170) in total. Sensitivity analyses for all scenarios also showed dominant results for the intervention group (Table 5). The dominant results did not change when the decrease in incidence of either dysmenorrhea or depression reduced based on the upper limit of the 95% confidence interval of the incidence in the intervention group.

Table 5. Result of cost-effectiveness analysis.

	Base scenario		Sensitivity analysis
			Scenario 1		Scenario 2		Scenario 3
	Cont	Int	Cont	Int	Cont	Int	Cont	Int
Incidence (monthly)
Dysmenorrhea	1.31%	0.33%	1.31%	0.33%	1.31%	0.53%	1.31%	0.33%
Depression	1.50%	0.30%	1.50%	0.30%	1.50%	0.30%	1.50%	0.54%
Application fee (monthly)	¥0	¥400	¥0	¥480	¥0	¥400	¥0	¥400
Medical expense, thousand yen	¥694	¥669	¥694	¥669	¥694	¥670	¥694	¥672
Labor productivity,^a thousand yen	¥25,738	¥25,882	¥25,738	¥25,882	¥25,738	¥25,877	¥25,738	¥25,858
Application fee, thousand yen	¥0	¥35	¥0	¥42	¥0	¥35	¥0	¥35
Total cost (medical expense + loss of productivity^b + application fee), thousand yen	¥838	¥704	¥838	¥711	¥833	¥705	¥814	¥707
QALY	6.77	6.84	6.77	6.84	6.77	6.84	6.77	6.83

^a Labor productivity: present value of weighted average of labor productivity of dysmenorrhea, depression, and others, which was calculated regarding presenteeism and absenteeism.

^b Loss of productivity: difference of labor productivity between intervention and control groups.

Abbreviation. Cont, control; Int, intervention; QALY, quality-adjusted life years.

¥ , Japanese yen (¥100 = $0.9, as of December 2017).

Discussion

In this RCT, the incidence of depression and dysmenorrhea was significantly lower in women in the intervention group compared with that in women in the control group in the third month. A probable reason to explain the reduction in the incidence of depression is that the application helped users who previously had symptoms of depression, including those caused by PMS, by improving their lifestyle and alleviating symptoms. Another reason could be that a user who had symptoms of depression could have learnt about PMS from the application, and visited a gynecologist instead of a psychiatrist. The latter might cause the intervention group to exhibit an increase in the incidence of PMS in the third month from the second month, resulting in higher incidence than the control group. The reduction in incidence of dysmenorrhea could be attributed to the effective use of the advice provided by the application. The application provides information on how to alleviate the associated symptoms, recommended meals, massages, bathing methods, pushing points, and so on. Regarding the recommended meals, it was suggested that some nutrients might be able to alleviate symptoms of dysmenorrhea in Japan³⁰. Additionally, it is reported that attributes of lifestyle such as sleeping hours and wake-up time may be associated with physical discomfort during menstruation³¹. We believe that the appropriate use of the information provided by the application can lead to alleviation of symptoms during menstruation, thereby causing a reduction in dysmenorrhea.

Labor productivity and absenteeism varied among the surveys, and presenteeism showed almost no change in both groups. Although differences exist across countries, monthly fluctuation is reported for labor productivity³². In Japan, labor productivity exhibits a downward trend in April from February and March³³. The variation observed for absenteeism, especially the worsening in the third month surveyed in April in both groups in our study, may have been caused because of such seasonal fluctuation. Although not reflected in presenteeism, where there was no statistical difference between the groups, some symptoms, such as headache and shoulder pain, were significantly lower in the intervention group. These differences potentially had an advantageous influence on labor productivity. The average value of the PHQ-9 score was lower than that measured in the baseline survey in the intervention group, but was not significantly different from the control group other than in the second month. The reason for the lack of change in the score, even though the incidence of depression changed, may be attributed to improvements due to treatments. Furthermore, we do not exclude the possibility that the RCT period of 3 months was relatively short for these parameters in order to observe a significant difference. Data recorded by users in the longer term can make prediction of the menstrual cycle and ovulation day more accurate, which may have a more positive effect on self-health management, especially for women with irregular menstruation. Regarding the evaluation of the application, in a systematic review of RCTs that targeted healthy people to investigate the effectiveness of applications to promote a healthy lifestyle, only 10 out of 40 applications showed statistical differences. In most of the studies, the observation period was <6 months, and half of them targeted <100 individuals³⁴. Because statistical difference was observed in the case of some variables including incidence of depression and dysmenorrhea, despite the short period of this RCT, it would be worthwhile to conduct a longer RCT (e.g. 6 months to >1-year period) using the same application, and check if statistical significance is seen in more variables.

Based on the cost-effectiveness analysis, we obtained a dominant result for the use of the application. In this analysis, we set the application fee at JPY 400, and assumed no incidence of side-effects by the application use based on its actual use. We considered that these two assumptions substantially contributed to the cost-effectiveness. The results indicated that the utilization of application that supports self-management, leading to alleviation and prevention of symptoms, can stop or even reduce the amount of medication. Moreover, such an application appears to generate no side-effects, making it safe, and is also available at a low price. Therefore, promoting the use of applications for healthcare might be justified.

Limitation and generality

In this RCT, subjects included those who were willing or relatively willing to use the application; in addition, we did not ask users about the frequency and features of the application they used. Compared with actual users, for example, those who purchased the application on their own, the motivation to use the application by the subjects might be low, resulting in a lower effect observed in this study. Similarly, if the application was provided by employer, the motivation to use it might be lower than this RCT. Hence, the motivation to use is expected to vary among the target populations, which may cause the differences in the effects of the application use. The response rate for each survey was relatively high, but it decreased with time and was lower in the intervention group than that of the control group. This can be attributed to the assumption that the individuals in the intervention group might have felt reluctant to answer the survey when they stopped using the application. Individuals with severe symptoms probably did not reply, which could contribute to the overall lower responsiveness in both groups, leading to a difference in the incidence rate when compared to that from the JMDC data. All outcomes were based on online surveys; therefore, the results depend on objectivity and preciseness of the reports provided by users. Regarding these potential biases and representativeness of subjects recruited from the online panel of the research company, we adjusted the incidence rate of diseases by using JMDC data. Although we did not consider the change in occupation and working environment for analyses, these factors might possibly affect the results. Although it is reasonable to expect that a relatively short study period (3 months) might result in smaller changes during observations when compared with a long-term study, the change in status of use of the application should be considered when assuming the effect in the case of a long-term study.

We used the JMDC database as one of the data sources for our cost-effective analysis because its coverage is considered to be appropriate. However, differential results could be generated if the population distribution (age and occupation) is different. The preciseness of diagnosis depends on the preciseness of the claims data. The value of presenteeism and QoL used were taken from previous studies, and, thus, the results could be biased.

Conclusions

This RCT study suggested that the use of “Karada-no-kimochi” may be effective in reducing the onset of dysmenorrhea and depression. The cost-effectiveness analysis indicated a dominant result from the use of the application. These results demonstrated the economic value using smartphone applications for self-health management. We believe that the use of this application is recommended because of the reduced costs, increased safety associated with its use, and the alleviation of physical and mental symptoms.

Transparency

Declaration of funding

This study was supported by docomo Healthcare, Inc.

Declaration of financial/other relationships

MS reports personal fees for general honoraria, consulting and lectures, and non-financial support for analysis, writing, and editing from docomo Healthcare, Inc. during the course of the study. In addition, personal fees were obtained for lectures from ROHTO Pharmaceutical Co. Ltd, Takeda Pharmaceutical Co. Ltd, Otsuka Pharmaceutical Co. Ltd, and Jex Co. Ltd outside the submitted work. HK is an employee of docomo Healthcare, Inc. A peer reviewer on this manuscript has disclosed that they are an employee of SPD Development Company Ltd, a wholly owned subsidiary of SPD Swiss Precision Diagnostics GmbH, the manufacturer of fertility and pregnancy tests. The remaining peer reviewers have no conflicts of interest to disclose.

Acknowledgments

Kosuke Iwasaki, MBA, and Tomomi Takeshima, PhD, employees of Milliman Inc., provided data analysis, and writing and editorial support, respectively. We would also like to thank Editage (www.editage.jp) for English language editing.