Evaluation of a Population-Wide Mobile Health Physical Activity Program in 696 907 Adults in Singapore.

Background Evidence of scaled-up physical activity interventions is scarce. This study evaluates the uptake, engagement, and effectiveness of one such intervention program. Methods and Results The program was open to individuals aged ≥17 years in Singapore. The main intervention components comprised device-based daily physical activity recording paired with step count goals and financial rewards. According to the different reward opportunities, we divided the evaluation period (August 2017 to June 2018) into the baseline monitoring phase, the main challenge phase, and the maintenance phase. Uptake was assessed by the number of individuals registered, and engagement by the step recording duration after registration. The effectiveness was defined as changes in mean daily step count from baseline to the main challenge phase and the maintenance phase. A total of 696 907 participants registered, including more Singapore citizens (versus noncitizens), women, and younger (aged 17-39 years) individuals. The evaluation of engagement and effectiveness included 421 388 (60.5%) participants who provided plausible characteristic information and step count data. The median duration of engagement was 74 (IQR, 14-149) days. Compared with the baseline of 7509 (SD, 3467) steps, mean daily step count increased by 1579 (95% CI, 1564-1594) steps during the main challenge phase and 934 (95% CI, 916-952) steps during the maintenance phase. Greater engagement and activity increase were found in participants who are citizens, women, aged ≥40 years, non-obese, and using separate wearables (versus smartphones). Conclusions Mobile health physical activity interventions can successfully reach a large population and be effective in increasing physical activity, despite declining program engagement over time.

Health Promotion Board Singapore

mobile health

National Steps Challenge

National Steps Challenge Season Three

Clinical Perspective

What Is New?

This study presents a sustained scaled‐up implementation of a mobile health physical activity intervention, which reached over 14% of the adult population in Singapore between 2017 and 2018.

Based on the objectively measured daily step count, we found that the program engaged 60% of the participants for a median of 74 days and increased their daily steps by about 1500 steps during the main intervention period.

What Are the Clinical Implications?

The evaluation shows that increasing physical activity in a large real‐world population is achievable by harnessing technology, behavior change techniques, and multisectoral collaborations.

Physical inactivity, a leading risk factor for noncommunicable diseases, has been prevalent globally with little improvement. Effective physical activity interventions need to be scaled up urgently to increase population physical activity. , However, evidence of such scaled‐up implementation has been limited, and the impact remains unclear. ,

Recent advances in mobile and information technologies have provided unprecedented opportunities to use mobile health (mHealth) approaches to scale up interventions, including the promotion of physical activity. The systematic review of reviews for the US 2018 Physical Activity Guideline supports the use of mHealth to improve physical activity. Another systematic review of randomized controlled trials and quasi‐experimental studies further showed that mHealth interventions were more effective when they incorporated behavior change techniques, such as self‐monitoring, gamification, and financial incentives. However, this evidence mostly originated from relatively small studies conducted under controlled settings with short follow‐up periods. , , Recently, several scaled‐up mHealth physical activity interventions have emerged and demonstrated the potential of mHealth approaches in physical activity interventions (Data S1, Figure S1, and Table S1). However, the effectiveness of these interventions was usually not evaluated or based on small and highly selective subsamples. , , , , , , The 2 exceptions are the Carrot Rewards (N=35 014) and the Stepathlon study (N=69 219). Carrot Rewards found a small increase of 116 steps in daily step counts by the 12‐week program in Canada. The Stepathlon study, which included participants from around the world, on the contrary, observed a large self‐reported increase of 3515 steps by the 100‐day intervention. The current evidence is thus scarce and inconsistent, and the impact of scaled‐up mHealth physical activity interventions remains unclear.

The National Steps Challenge (NSC) is a nationwide physical activity program in Singapore that uses wearables, a smartphone application, and various behavior change techniques. It is implemented by the Health Promotion Board (HPB) under the Ministry of Health Singapore. With 1 event season each year since 2015, the NSC focuses on stepping activities in adults and collects extensive program data for evaluations. Multiple factors contributing to the successful implementation were discussed previously. The present study aims to provide a comprehensive evaluation of the National Steps Challenge Season Three (NSC3). Specifically, our objectives are to evaluate the NSC3 uptake, engagement, and effectiveness in increasing physical activity.

Method

Because of the sensitive nature of the data collected for this study, requests to access the data set from qualified researchers trained in human subject confidentiality protocols may be sent to the Health Promotion Board, Singapore at HPB_Mailbox@hpb.gov.sg.

Study Design and Participants

The NSC3 was a longitudinal intervention embedded in the national health promotion system of Singapore. Details of the NSC3 are available online. Major NSC3 information channels included the Healthy365 App; 538 regular and 14 mega on‐site roadshows conducted between September 21, 2017, and February 14, 2018, across Singapore; and mass‐marketing campaigns through broadcasting, outdoor, and social media. The Healthy365 App is a mobile application developed by the HPB, with a supplementary kiosk version for non–smartphone users. It has been freely available on Google Play and the App Store in Singapore since 2015. Besides monitoring health behaviors, it serves as the main touchpoint for NSC and several other health promotion programs by the HPB. While providing weight and height is optional for using the Healthy365 App, users are required to report information on their birthday, sex, postal code, and the government‐issued identification number, which can be used to differentiate Singapore citizens and noncitizens.

All Individuals in Singapore aged ≥17 years were eligible to register for NSC3. The noncitizens comprised permanent residents and other long‐term visa holders in Singapore. The participation was free, and the registration was through the Healthy365 App since September 26, 2017, self‐administered remotely or facilitated by HPB staffs at the roadshows. Singapore citizens and permanent residents were additionally eligible to collect 1 HPB‐issued fitness tracker free of charge at the roadshows or at the HPB headquarters, if they registered for the NSC3 and had not received any HPB‐issued tracker during previous NSC seasons.

Participants provided informed consent during registration and were allowed to withdraw from the NSC3 without a penalty at any point in time. Ethical approval for this study was obtained from the Institutional Review Board of the National University of Singapore.

Procedures

The entire NSC3 evaluation period lasted from August 1, 2017, to June 10, 2018, encompassing the major event period: the Sure‐Win period between October 28, 2017, and March 31, 2018 (Figure 1). The term “Sure‐Win” refers to guaranteed rewards for physical activity achievements. Public education on physical activity and various subchallenges were delivered throughout the entire NSC3 period to raise awareness of active lifestyles and to keep participants engaged. The main NSC3 intervention components were daily physical activity monitoring paired with predefined step count goals and financial rewards for achieving these goals.

Figure 1

Program implementation time frame and evaluation phases.

A, NSC3 implementation time frame. B, Participant‐specific NSC3 evaluation phases and the eligible rewards according to the participant’s registration time. Participants varied in time windows of the 3 NSC3 phases because of their different program registration time. NSC3 indicates National Steps Challenge Season 3.

The Healthy365 App allowed NSC3 participants to record daily physical activity throughout the entire NSC3 period, even before registration. Participants with the Healthy365 App on their smartphones first measured daily steps via their preferred wearables: HPB‐issued trackers, self‐purchased commercial wearables, or smartphones with built‐in accelerometers (Data S2). The measured daily step count data had to subsequently be transferred to the Healthy365 App wirelessly, at least once every 7 days to avoid data loss. The transferred data were automatically uploaded to the HPB database server when the Healthy365 App connected to the Internet. Participants without smartphones or without the Healthy365 App on their smartphones could only measure daily steps using HPB‐issued trackers and transfer their step count data using the “Sync‐for‐Friends” function of the Healthy365 App on other individuals’ smartphones. Alternatively, they could transfer the data via the HPB kiosks located at 34 outlets across Singapore. All participants were allowed to switch between supported wearables at any point in time.

Depending on participants’ registration dates and physical activity recording behaviors, each participant could experience up to 3 NSC3 phases for different durations: the baseline monitoring phase, the main challenge phase, and the maintenance phase (Figure 1). Characterized by the different incentive opportunities, the 3 phases were defined by the research team for evaluation purposes. The baseline monitoring phase included the days before the Sure‐Win period if a participant registered for the NSC3 before October 28, 2017, or the days before a participant’s registration date if the participant registered after October 28, 2017. During this phase, participants were not eligible to receive financial rewards. We defined the main challenge phase as the days of the Sure‐Win period after participants registered. Two types of financial rewards, Sure‐Win reward tiers and lucky draw chances, were available every day during this phase for achieving the predefined activity goals. The maintenance phase represented the days after participants registered for the NSC3 and after the Sure‐Win period. During this phase, participants won lucky draw chances for achieving the predefined goals.

The 3 increasing daily step count goals were 5000 to 7499, 7500 to 9999 and 10 000+ steps. During the main challenge phase and the maintenance phase, participants earned 1, 2, and 3 lucky draw chance(s) each day they achieved the increasing step count goals, respectively. The lucky draw chances were for the NSC3 grand draw held on June 18, 2018, with prizes such as airline and cruise tickets. Each day during the main challenge phase, participants additionally earned 10, 25, and 40 Health‐Points for achieving the increasing step count goals, respectively. Health‐Points were accumulated for Sure‐Win reward tiers. The first 750 accumulated Health‐Points unlocked the first reward tier equivalent to S$5. Additional 1500 Health‐Points unlocked the second reward tier of S$10. Afterwards, participants reached the next tier, up to the sixth tier (each equivalent to S$5), with each additional 750 Health‐Points. The reward tiers could be redeemed with a wide range of e‐vouchers for lifestyle retails, food and beverage outlets, and supermarkets.

Outcomes and Statistical Analysis

We defined the program uptake as the individuals who registered for the NSC3. We summarized the daily and cumulative uptake over time and explored the uptake geographically by Singapore planning areas in absolute number and the percentage among the Singapore population (citizens and noncitizens) aged ≥17 years (Data S3 and Figure S2 and S3).

We considered the characteristics of a participant to be valid for further analyses if the participant provided plausible demographic and anthropometric information: identifiable nationality, sex, age >17 years, weight between 30 and 300 kg, and height between 101 and 220 cm. Characteristics of NSC3 participants derived from this information were compared with that of the entire Singapore population aged ≥17 years using chi‐squared tests (Data S2).

The program engagement was characterized by the number of days a participant recorded valid daily steps following the registration. Daily step counts above 0 were considered to be valid. To obtain the percentage of engagement, we divided the duration of engagement by the participant‐specific number of available NSC3 days. The available NSC3 days lasted from a participant’s registration date to the end of the evaluation period on June 10, 2018. For participants who provided valid daily step count data, we compared the duration and percentage of engagement by participant characteristics using Kruskal‐Wallis tests. A participant was considered to remain in the NSC3 until the last day the participant recorded steps. On each calendar date, we computed the percentages of registered participants who were engaged and who remained in the program.

We compared participants’ mean daily step count by their characteristics using 1‐way ANOVA. Activity difference by the day of the week and public holidays was explored through a linear mixed‐effect model using mean daily step count on each calendar date as the dependent variable. NSC3 phase‐specific random intercepts and independent correlation matrix were applied in the model (Data S4 and Table S2).

The effectiveness of the program was defined by changes in mean daily step count from the baseline monitoring phase to the main challenge phase and the maintenance phase. The changes were estimated via a main linear mixed‐effect model with participant‐specific random intercepts and unstructured within‐participant correlation matrix. The dependent variable was the phase‐specific mean daily step count, instead of the raw daily step count data. This was to reduce model imbalance because participants varied considerably in the number of days with step count data for each NSC3 phase. The model included only participants with at least 1 reliable phase‐specific mean daily step count. The mean daily step count of an NSC3 phase was deemed to be reliable if the participant recorded steps for at least 4 days during the phase. The threshold of 4 days was based on a reliability analysis, reaching intraclass correlation coefficients of ≥80%. Sensitivity analyses with thresholds ranging from 1 to 30 days as well as the complete‐case analyses can be found in Tables S5 and S6. The indicator of NSC3 phases served as the independent variable of interest, which was included as a predictor in the linear mixed‐effect model applied to all participants. The model also included 8 participant characteristic variables as fixed factors: participant’s nationality, sex, age, body mass index (BMI), registration for the previous NSC season, NSC3 registration time, wearable used, and numeric duration of engagement.

To investigate whether the program effectiveness differed by participant characteristics, subgroup analyses were conducted. Participants were stratified into subgroups by each of the above‐mentioned 8 characteristic variables, including the duration of engagement categorized according to quartiles. A separate linear mixed‐effect model was fitted for each subgroup using the same list of fixed factors as the model applied to all participants, except that it excluded the variable used to generate the subgroup.

We performed all the analyses in R (version 3.6.1), and used package “nlme” (version 3.1–141) for the linear mixed‐effect models.

Results

A total of 696 907 participants registered for the NSC3, and 113 withdrew (Figure 2); 539 296 (77.4%) participants provided plausible demographic and anthropometric information; and 421 388 (60.5%) participants further recorded valid daily steps and were included in the evaluation of program engagement and effectiveness.

Figure 2

Participant flowchart.

NSC3 indicates National Steps Challenge Season 3.

Uptake of the NSC3 started from September 26, 2017, and continued to increase until April 29, 2018, with major peaks coinciding with NSC3 mega roadshows. Thirty‐three percent of the participants registered within the first month and 78% within the first 2 months (Figure 3). Geographically, participants were distributed across all residential areas in Singapore. A higher uptake in absolute numbers was observed in more densely populated areas, but we observed no clear geographical pattern in the percentage of uptake (Data S3 and Figure S2 and S3). Compared with the Singapore population, the NSC3 reached greater proportions of participants who were citizens, women, younger (aged 17–39 years), and nono‐bese and non‐underweight (BMI 18.5 to <27.5 kg/m2), whereas participants who were noncitizens and men, older, and obese were underrepresented (Table 1).

Figure 3

Daily and cumulative NSC3 uptake over time from September 26, 2017, to April 29, 2018 (N=696 907).

NSC3 indicates National Steps Challenge Season 3.

Table 1

NSC3 Participant Characteristics (N=539 296) Compared With the Singapore Population (N=4 845 000 as of June 2018)

Characteristics	NSC3 participants, n (%)	Singapore population aged ≥17 y, %
Nationality
Singapore citizen	349 259 (64.8)	59.1
Noncitizen	190 037 (35.2)	40.9
Sex
Female	313 559 (58.1)	47.5
Male	225 737 (41.9)	52.5
Age, y
17–39	270 241 (50.1)	39.4
40–59	207 871 (38.5)	38.7
60–79	58 490 (10.8)	19.4
≥80	2694 (0.5)	2.5
Body mass index, kg/m2
<18.5	35 298 (6.5)	6.4
18.5 to <23	224 905 (41.7)	38.3
23 to <27.5	188 251 (34.9)	32.3
≥27.5	90 842 (16.8)	23.0
Registration for previous NSC season
No	363 423 (67.4)	…
Yes	175 873 (32.6)
NSC3 registration time
Before Sure‐Win period	196 832 (36.5)	…
Within 30 d since Sure‐Win period began	215 571 (40.0)
>30 d after Sure‐Win period began	126 893 (23.5)
Wearables
HPB‐issued	159 730 (29.6)	…
Phone‐based	129 402 (24.0)
Commercial	26 077 (4.8)
>1 wearable type	106 179 (19.7)
Not applicable*	117 908 (21.9)

Comparisons of participant characteristics with the Singapore population using chi‐squared tests: all P <0.001. HPB indicates Health Promotion Board Singapore; NSC, National Steps Challenge; and NSC3, NSC Season 3.

No wearable information was available for participants who did not record any valid daily step.

As part of the NSC3, 40.1 million participant‐days of step count data were recorded, 91.6% of which occurred after participants registered for the NSC3. The median duration of engagement was 74 (IQR, 14–149) days, accounting for 36% (IQR, 7%–68%) of the available NSC3 days (Table 2). Temporally, the percentage of registered participants who remained in the NSC3 on each date declined steadily over time. With intermittent drops on weekends and public holidays, the percentage of registered participants who were engaged on each date followed a U‐shaped curve before the Sure‐Win period and decreased steadily afterwards (Figure 4A).

Table 2

Duration and Percentage of Engagement and Mean Daily Step Counts During NSC3 (N=421 388)

Characteristics	Duration of engagement in days, median (IQR)	Percentage of engagement, median (IQR)	Mean daily step counts, mean (SD)
Total	74 (14–149)	36 (7–68)	8184 (4073)
Nationality
Singapore citizen	81 (16–153)	41 (8–70)	8346 (4134)
Noncitizen	60 (10–140)	28 (5–62)	7870 (3933)
Sex
Female	79 (15–152)	38 (8–69)	7931 (3925)
Male	67 (12–144)	33 (6–66)	8543 (4247)
Age, y
17–39	54 (10–129)	27 (5–59)	7839 (3686)
40–59	94 (19–165)	46 (10–74)	8403 (4238)
60–79	108 (20–173)	52 (10–79)	8912 (4847)
≥80	108 (29–163)	51 (15–73)	9077 (4869)
Body mass index, kg/m2
<18.5	68 (12–141)	33 (6–64)	7870 (4017)
18.5 to <23	79 (15–152)	39 (8–68)	8309 (4036)
23 to <27.5	75 (14–151)	37 (7–69)	8291 (4131)
≥27.5	59 (11–139)	29 (5–64)	7762 (4028)
Registration for previous NSC season
No	46 (7–127)	23 (4–61)	7602 (4098)
Yes	120 (49–179)	53 (23–77)	9251 (3803)
NSC3 registration time
Before Sure‐Win period	127 (38–186)	52 (16–76)	8871 (4062)
Within 30 days since Sure‐Win period began	60 (11–133)	28 (5–61)	7940 (4023)
>30 days after Sure‐Win period began	24 (4–75)	18 (3–52)	7208 (3939)
Wearables
HPB‐issued	64 (13, 136)	31 (6, 62)	8698 (4589)
Phone‐based	43 (2, 135)	23 (1, 65)	6831 (3486)
Commercial	96 (29, 165)	46 (15, 74)	9769 (3402)
>1 wearable type	114 (41, 170)	53 (20, 75)	8673 (3609)

Participant characteristics for the 421 388 participants can be found in Table S8. Data of engagement and daily step counts by the NSC3 phases in mean (SD) can be found in Tables S3 and S4. Comparisons by participant characteristics using Kruskal‐Wallis tests and ANOVA tests: all P <0.001. Percentage of engagement was relative to participants’ available NSC3 days (median of available NSC3 days: 223 days; IQR, 205–240 days). HPB indicates Health Promotion Board Singapore; NSC, National Steps Challenge; and NSC3, NSC Season 3.

Figure 4

Percentage of daily step recording and the mean daily step counts over time (N=421 388).

A, Percentage of registered participants on each calendar date who engaged in NSC3 as well as who remained in NSC3. Registered participants for a specific calendar date reflected all the participants who registered for NSC3 by that date. A registered participant remained in NSC3 until the last day the participant recorded daily step counts. B, Mean daily step counts on each calendar date, stratified according to the 3 specified NSC3 phases (baseline monitoring phase, main challenge phase, and maintenance phase). Black dots indicate weekend days; red dots indicate nonweekend public holidays. N=421 388 reflects the total number of participants who contributed data to Figure 4, but the mean daily step counts on specific dates were computed on the basis of participants who provided step count data on the date. The number of participants in each NSC3 phase with step count data on each calendar date can be found in Table S9. NSC3 indicates National Steps Challenge Season 3.

Both duration of engagement and mean daily step count level were significantly higher in participants who were citizens, older (≥40 years), non‐obese (BMI, <27.5 kg/m2), and who joined previous NSC seasons, registered for the NSC3 earlier, and did not solely use smartphone‐based wearables (Table 2). Male participants had shorter engagement, but a higher mean daily step count level. Across the NSC3 phases, daily step counts were lower on Sundays by 616 steps and public holidays by 789 steps (Figure 4B).

After adjustment for potential confounders, mean daily step count was 1579 (95% CI, 1564–1594) steps greater during the main challenge phase and 934 (95% CI, 916–952) steps greater during the maintenance phase than the level during the baseline monitoring phase (Table 3). These increases were robust in sensitivity analyses (Tables S5 through S7). Subgroup analysis revealed greater increases among citizens and noncitizens, women, the older (≥40 years), non‐obese (BMI <27.5 kg/m2) participants, and those who were engaged longer and who did not solely use smartphone‐based wearables.

Table 3

Mean Daily Step Count During the Baseline Monitoring Phase and Changes in Mean Daily Step Count From Baseline to the Main Challenge Phase and the Maintenance Phase (N=384 691)

Characteristics	Mean daily step count during the baseline monitoring phase, mean (SD)	Changes of mean daily step counts from baseline monitoring phase, model estimate (95% CI)
		Main challenge phase	Maintenance phase
Total	7509 (3467)	1579 (1564–1594)	934 (916 952)
Nationality
Singapore citizen	7591 (3612)	1864 (1844–1883)	1148 (1125–1171)
Noncitizen	7410 (3284)	1225 (1203–1247)	722 (693–750)
Sex
Female	7114 (3337)	1707 (1688–1725)	1055 (1033–1077)
Male	8106 (3574)	1488 (1463–1512)	881 (852–911)
Age, y
17–39	7267 (3063)	1280 (1260–1300)	827 (802–852)
40–59	7638 (3650)	1852 (1828–1875)	1123 (1095–1151)
60–79	8161 (4318)	2303 (2252–2353)	1377 (1319–1436)
≥80	8027 (4672)	2767 (2511–3023)	1633 (1336–1929)
Body mass index, kg/m2
<18.5	7042 (3467)	1651 (1589–1712)	989 (916–1063)
18.5 to <23	7517 (3477)	1673 (1651–1696)	1027 (1000–1055)
23 to <27.5	7664 (3480)	1621 (1596–1645)	984 (954–1014)
≥27.5	7325 (3389)	1472 (1436–1508)	892 (848–937)
Registration for previous NSC season
No	6875 (3307)	1635 (1614–1656)	1184 (1158–1209)
Yes	8223 (3505)	1625 (1604–1645)	756 (732–781)
NSC3 registration time
Before Sure‐Win period	7889 (3593)	1781 (1761–1800)	932 (908–957)
Within 30 d since Sure‐Win period began	6999 (3177)	1375 (1347–1403)	903 (870–937)
>30 d after Sure‐Win period began	6523 (2988)	1307 (1267–1347)	990 (945–1035)
Wearables
HPB‐issued	8519 (4397)	1999 (1965–2034)	1085 (1044–1126)
Phone‐based	7054 (2932)	738 (720–756)	322 (301–344)
Commercial	9456 (3261)	1028 (989–1067)	592 (545–640)
>1 wearable	6993 (3109)	2353 (2326–2379)	1749 (1715–1782)
Quartiles of engagement duration in days
<15	6301 (2995)	166 (121–212)	413 (310–517)
15–74	6975 (3144)	893 (864–922)	735 (695–775)
75–149	7480 (3394)	2004 (1974–2034)	1377 (1341–1413)
150–258	8471 (3636)	2052 (2029–2075)	1229 (1202–1256)

Results were based on step records that comprised the reliable phase‐specific mean daily step count. A reliable phase‐specific mean daily step count required at least 4 records of daily step counts during the corresponding NSC3 phase. Changes in mean daily step counts were estimated by linear mixed‐effect models, adjusting for nationality, sex, age, body mass index, registration for previous NSC season, NSC3 registration time, wearable used, and numeric duration of engagement where applicable. For each participant subgroup, a separate model was fitted using data from the corresponding subgroup. All model estimates were P<0.001. Participant characteristics for the 384 691 participants can be found in Table S8. HPB indicates Health Promotion Board Singapore; NSC, National Steps Challenge; and NSC3, NSC Season 3.

Discussion

The NSC represents a sustained and scaled‐up mHealth physical activity intervention embedded in the national health promotion system of Singapore. The large‐scale program data, including objectively measured daily physical activity, provided a unique opportunity for in‐depth evaluations. This evaluation of the NSC3 included all the 696 907 NSC3 participants. Sixty percent of them actively participated and were engaged for a median of 74 days. Compared to baseline, the mean daily step count increased by about 1500 steps during the main challenge phase and remained about 900 steps higher during the maintenance phase. To our knowledge, this study is the largest evaluation of an mHealth physical activity intervention. , , , , , , ,

In 7 months, the NSC3 attracted more than 14% of the adult population in Singapore, over three‐quarters of whom registered within the first 2 months. The participants presented diverse characteristics and were spatially distributed across the entire country. A substantial proportion of the participants belonged to subpopulations with elevated risks of physical inactivity, such as those aged >60 years or overweight. The considerable and rapid reach of the NSC3 is likely attributable to various factors, including the focus on simple stepping activities, attractive incentives (eg, free fitness trackers, and Sure‐Win rewards), versatile activity recording modes, and widespread on‐site roadshows and public education. According to our literature review, previous evaluations of mHealth physical activity interventions were substantially smaller and reached <70 000 participants. , , , , , , , Meanwhile, several other large‐scale mHealth studies have emerged, along with many commercial smartphone applications. , , However, these studies and applications usually pursued different objectives and did not implement or evaluate physical activity interventions. The Apple Heart Study, for instance, assessed the ability of a smartwatch application to identify atrial fibrillation using pulse rate data from >400 000 participants. Another study illustrated worldwide physical activity inequality using smartphone‐based physical activity data from >700 000 people. Together, our evaluation of the NSC3 and these studies and programs support the potential and feasibility of using mHealth approaches for large‐scale real‐world investigations.

Our evaluation illustrates that >60% of NSC3 participants were engaged, with a median of 74 days of the 223 program days. Engagement of other scaled‐up mHealth physical activity interventions was quantified sporadically and heterogeneously, generally appearing lower than in the NSC3. , , , , , , , For instance, the 10 000 Steps Australia program found an average of 31 days of engagement during the 190 program days among the 17 000 participants who recorded step counts between July 2013 and April 2014. Despite the comparatively long engagement observed in the NSC3, a substantial decline was observed over time, with less than a third of participants recording activity by the end of the Sure‐Win period. However, this finding is consistent with most mHealth interventions. Some smaller mHealth randomized controlled trials, on the other hand, demonstrated greater engagement than the NSC3. , , An important example is the TRIPPA (Trial of Economic Incentives to Promote Physical Activity) trial, an mHealth study to increase physical activity in Singapore adults. About 70% of the 396 individuals, who received wearables and financial incentives, continued recording physical activity by the end of the 6‐month intervention period. However, the percentage dropped rapidly to only 10% by the end of the 6‐month follow‐up. The higher financial incentives during the TRIPPA intervention, averaging around S$470 (versus maximum S$30 in NSC3), may partly explain the greater engagement.

We identified several participant characteristics and patterns associated with greater engagement. Such knowledge has been scarce for mHealth approaches, especially for scaled‐up programs. Before our evaluation, an analysis of 100 000 participants from 8 studies between 2014 and 2019 was the largest investigation on indicators of higher mHealth engagement using individual‐level data. The study found clinical referral, monetary benefit, older age, and disease condition as key drivers. Our evaluation revealed additional positive correlates of engagement: female sex, non‐obesity, previous program registration, early registration, and usage of separate wearables (versus smartphones) for activity recording. The NSC3 also identified lower engagement on weekends and public holidays. Most factors positively related to NSC3 engagement were also associated with higher activity levels, except for male sex, which was associated with greater step count levels despite lower engagement. While these findings have potential to inform future strategies to increase engagement, mHealth interventions such as the NSC may not need a full‐time engagement to achieve the desired behavior change. ,

A recent systematic review of randomized controlled mHealth trials found that wearables and financial incentives together increased the mean daily step count in adults by 607 steps during the incentive period and 513 steps during the postincentive period. The TRIPPA trial, which also targeted adults in Singapore, showed that wearables plus cash incentives improved daily step counts by 1000 and 500 steps at the end of the incentive and follow‐up period, respectively. Another recent systematic review of interventions using step count monitors (with or without financial incentives) in community‐dwelling adults reported increased physical activity by 1126 steps/day at ≤4 months, 1050 steps/day at 6 months, and 464 steps/day at 1 year. Our evaluation suggests that the NSC3 yielded larger increases in daily step counts. The larger increase may, in part, be attributable to other behavior change techniques in this multicomponent intervention, such as incremental step count goals, activity feedback, gamification with Health‐Point tokens, and physical activity education. It thereby indicates that the effectiveness of mHealth physical activity interventions can be retained in large real‐world settings if they are carefully scaled‐up. In the NSC3, greater increases in step counts were particularly apparent among participants of older age, who used separate wearables (versus smartphones), and who had longer engagement. These findings are consistent with the above‐mentioned systematic review of randomized controlled trials. The present evaluation further found that citizens and women increased their physical activity more substantially. In addition, subgroups with higher baseline activity levels generally observed greater increases, except for men, participants with previous program registration, and participants using self‐purchased fitness trackers.

Longitudinal studies have repeatedly demonstrated that higher daily step counts are associated with reductions in all‐cause mortality, cardiovascular diseases, and type 2 diabetes. , , , , For instance, in prospective cohort studies, every 1000‐steps‐per‐day increment was associated with 6% lower mortality among Australians and 14% lower mortality among Britons. Similarly, randomized controlled trials illustrated that increases in daily step counts comparable to or less than those observed in the NSC3 resulted in significantly improved health outcomes. , , The increases in daily step counts observed in the NSC3 could therefore lead to substantial individual and population benefits.

Multiple factors and resources have contributed to the NSC3’s large‐scale implementation. First, the program adopted extensive multisectoral partnerships led by Singapore’s governmental HPB, including collaborations with academic institutes, technology and manufacturing industries, retail establishments, and other governmental and community organizations. This is consistent with existing evidence that partnerships beyond the health sector and political support are key factors for successful scaling up of physical activity interventions. Embedded in the country’s broad systematic health promotion framework, the NSC3 was able to synergize with various other programs and benefited from sharing major health promotion resources and infrastructures with them. In addition, the NSC3 tailored the design and implementation of its multibehavioral intervention components by integrating the local practice‐based health promotion experience with scientific research evidence. While these features and the practice‐based nature are integral for the scaled‐up implementation of the NSC3, it is worth acknowledging that many of the implementation aspects were not research driven, which created challenges for the scientific evaluation. For instance, considering the complexity of this mHealth program, we are unable to determine specific effects of individual intervention components and their detailed interactions because no dedicated efforts were made to collect relevant data.

Despite the novelty and strengths of this evaluation, several important limitations should be acknowledged. First, the NSC3 was a population‐wide health intervention program and was not conceptualized as a research study. As such, our evaluation relies on routinely collected data and does not have the same internal validity and quality of outcome measures as a purposefully designed randomized controlled trial. Besides, the availability of physical activity measurements depends on participants’ program engagement. The estimates of effectiveness could therefore be subject to bias and confounding. We conducted multiple sensitivity and secondary analyses to address this concern and found that our observations were robust. Second, despite its large reach, certain subpopulations including the obese and elderly were underrepresented. Although the large sample size enabled us to conduct subgroup analysis, it is conceivable that individuals in specific subgroups may fail to represent the corresponding population groups. This may partially explain the higher activity level and activity increase, for example, in older participants. Third, the duration and intensity of the NSC3 intervention varied among participants because of their different registration dates and involvement in the subchallenges. However, this reflects the real‐life nature of the NSC3. Fourth, the maintenance phase in the current evaluation was relatively short, and the program’s effectiveness may be affected by other confounders unavailable. Therefore, further investigations with longer follow‐up and more comprehensive assessments of confounders are warranted. Finally, information on the cost effectiveness of health promotion programs is important in terms of program sustainability and potential implementation in other countries. At this stage, we are not able to include a cost‐effectiveness analysis of the NSC3 because of the lack of relevant data. However, in Singapore, the NSC has benefited from multisectoral collaborations and effects of economies of scale. The program has continued since 2015 and its sixth yearly season started in October 2021. While future cost‐effectiveness analyses of NSC and similar mHealth programs are essential for policy makers, the NSC’s continuous implementation could be an indicator of the program’s sustainability, at least in high‐income settings.

Conclusions

The NSC represents an mHealth physical activity intervention on a large scale. Our evaluation shows that the NSC3 reached a very large and diverse population from across the country within a few months. While NSC3 engagement was not ideal and declined steadily over time, it appears to compare favorably with other scaled‐up mHealth programs. Moreover, substantial and clinically meaningful increases in daily step counts were achieved and maintained during the NSC3. This evaluation also extends current knowledge on factors associated with uptake, engagement, and effectiveness of mHealth interventions, and suggests directions for future program improvement. In sum, the evaluation presents an encouraging example for researchers, practitioners, and policy makers around the world to scale up effective physical activity interventions.

Sources of Funding

This research is supported by the Singapore Ministry of Health’s National Medical Research Council under the Seed Funding Programme by Singapore Population Health Improvement Centre (NMRC/CG/C026/2017_NUHS).

Disclosures

During the conduct of the study, Drs Müller‑Riemenschneider and Tan received research funding from Ministry of Health Singapore, and Drs Müller‑Riemenschneider, Tan, and Müller received research funding from the Health Promotion Board Singapore. The remaining authors have no disclosures to report.

Data S1–S4

Tables S1–S9

Figures S1–S3

References38, 39, 40, 41, 42, 43, 44

Click here for additional data file.