Skip to main content
. 2021 Sep 13;23(9):e26315. doi: 10.2196/26315

Table 2.

Features of smartphone-based physical activity intervention development or evaluation.

Author (year) Intervention Control Theoretical premise Primary outcome Other outcomes Technology feature Results
Baretta et al (2019) [86] Weekly tailored PAa goals

  • Starting goal (first week): 120 METb

  • Long-term goal: 600 METs per week of PA

  • Weekly goals broken down into daily goals

  • Factors not considered in the intervention development but proposed for the next study: working hours, time of the day, day of the week, health and illness, weather, etc

 N/Ac Self-efficacy theory and dynamic decision network PA measured by HRd sensor, self-efficacy beliefs N/A

  • Android app: Muoviti (visualizing the heart-beat rate graph of the last training session, the curves of weight and waistline variations week by week, the burned calories graph, session by session, and the percentage of vigorous activity with respect to moderate activity)

  • Other: HR wristbands (MioAlpha and PulseON)

  • N/A
Direito et al (2018 and 2019) [87,88] Daily individualized and adaptive PA and SBe goals:

  • Daily activities (eg, transport to or from work, PA at work)

  • Light-intensity activity to replace SB (eg, walking to a colleague’s desk rather than call or email, stand up while on the phone)

  • Leisure-time moderate-to-vigorous PA (eg, cycling)

  • Daily goals, visual and numerical feedback on past day and historical data, tips or suggestions, infographics, videos, and links, frequently asked questions, reminders, and push notifications

  • Context: workplace (location)

 N/A Intervention mapping taxonomy to identify behavior change techniques (eg, self-monitoring, goal setting, or review of goals) from literature. Integrated behavior change model constructs and behavioral intervention technology; 33 behavior change techniques were included Test the acceptability and feasibility of just-in-time adaptive intervention on PA and SB Pilot-testing the TODAYf app

  • Android apps: Art of Living app and TODAY app. Other: built-in phone sensors for SB and activity (ie, accelerometer)

  • TODAY app: low-effort and pleasant (54.3%), provides guidance on changing activity profile (52.6%), positively framed messages (64.4%), the app sustained interest over the 8 weeks (28.8%)

  • Most favorable behavior change techniques for the users (goal setting, discrepancy between current behavior and goal, feedback on behavior, instruction on how to perform the behavior, and behavior substitution)

  • Only significant improvement was occurred on light PA (see the results for statistics)
Conroy et al (2018) [12] Five daily text messages (between 8 AM to 8 PM). Three message types (move more, sit less, general facts or trivia [unrelated to PA or SB]). Message receipt was confirmed with a reply. Factors: context (weekday and weekend) N/A Social cognitive theory and control systems engineering Stepping time N/A

  • No app or text message

  • ActivPAL3 (activity tracker)

  • (Proof-of-concept study) 50% of the sample: more pronounced behavioral responses to text messages on weekends than weekdays; 50% had similar weekend or weekday responses; 50% of responders increased stepping time in response to “move more” messages, and 50% increased stepping time in response to “sit less” messages
Middelweerd et al (2020) [89], Klein et al (2017) [90], and Middelweerd et al (2018) [91] Weekly moderate-to-vigorous PA goals: 30 minutes of moderate PA for at least 5 days a week or 20 minutes of vigorous PA for 3 days a week
Contexts (location, weather, occupation)
Connected friends (Facebook APIg), if 2 participants of the intervention are connected
Up to 3 messages a day		 N/A Social cognitive theory, self-regulation theory and health action process approach and computational agent model To increase the total time spent in moderate-to-vigorous PA N/A

  • Android app: Active2Gether

  • Fitbit One (for self-monitoring only), ActiGraph wGT3XBT and GT3X+ (activity trackers)

  • No significant intervention effects were found for the Active2Gether-full and Active2Gether-ight conditions on levels of PA compared with the Fitbit condition: larger effect size for Active2Gether-ight (β=3.1, 95% CI −6.66 to 12.78, for minutes of moderate-to-vigorous PA; β=5.2, 95% CI −1334 to 1345, for steps). Smaller effect size for Active2Gether-full (β=1.2, 95% CI −8.7 to 11.1, for minutes of moderate-to-vigorous PA; β=−389, 95% CI −1750 to 972, for steps)
Korinek et al (2018) [92] and Freigoun et al (2017) [93]. More information is available in Martin et al (2018) [22] Daily step goal:

  • Pseudorandomly assigned daily step goal (doable [based on baseline median daily step] and ambitious [ie, up to 2.5×baseline median])+ rewards (points>Amazon Gift Cards)

  • Six 16-day cyclesh (cycle 0 [baseline], cycles 1 to 5 [step goals assigned])

  • Step goals prompted every morning+there were daily, weekly and monthly surveys

  • Morning and evening EMAi assessed constructs including (eg, confidence in achieving the goal, predicted busyness for that day, previous night’s sleep quality)

  • Factors considered: perceived stress, perceived busyness, weather information, sleep quality

 N/A Social cognitive theory (particularly self-efficacy construct), goal setting and control systems engineering (system identification) Feasibility, daily steps N/A

  • Android app: JustWalk

  • Fitbit Zip (activity tracker)

  • Other: web-based mobile questionnaire

  • Linear mixed effect model: each individual walked below 5000 steps at baseline with significant variation; mean intercept value 4863.3 steps (SD 1838.42), t98=10.49; P<.001.

  • Daily steps increased by 2650 steps per day on average from day 0 to day 16 (cycle 0 to cycle 1); t98=6.54, P<.001.

  • Quadratic mixed effect model: each individual walked roughly 5000 steps at baseline with significant variations; mean intercept value 5301.5 steps (SD 1862.04); t98=11.29, P<.001.

  • Daily steps increased by 1500 steps per day on average from cycle 0 to cycle 1 (1505 steps; t=5.52, P<.001); however, daily steps decreased by 247.3 steps per day on average from day 0 to day 16 (cycle 0 to cycle 1); t98=-5.01, P<.001

  • High adherence was observed (only 10 days of having missing step data; only 40 days of nonwear; <500 step counts). Common problem: sync lag with Fitbit
Rabbi et al (2015) [94] Daily personalized context-sensitive suggestions (PA and stationery). Manual and automatic logging to track activity and user location. Start of each day: 10 in-app activity suggestions (90% users’ most frequent activities [exploit]; 10% from users’ infrequent activities [explore]). MyBehavior app included both PA and dietary interventions Nonpersonalized generic recommendations Learning theory, Fogg behavior model, social cognitive theory, and exploit-explore strategyj Adherence, acceptability, behavior change N/A

  • Android app: MyBehavior; other: phone accelerometer and GPS

  • Intervention participants more intended to follow personalized suggestions than control (effect size=0.99, 95% CI 0 to 1.001; P<.001). Most intervention participants (78%) had a positive trend in walking behavior (also increased daily walking by 10 minutes during the intervention), whereas most control participants (75%) showed a negative trend. The users found MyBehavior app suggestion very actionable and wanted to follow them
Rabbi et al (2018) [95] Context-sensitive suggestions (PA and stationery). Manual and automatic logging to track activity and user location. In-app suggestions (80% users’ most frequent activities [exploit]; 20% from users’ infrequent activities [explore]); total time for each selected activity must not exceed 60 minutes. End of day reward score Static suggestions Learning theory, Fogg behavior model, social cognitive theory (self-efficacy) and exploit-explore strategyj Use, acceptability, early efficacy Qualitative feedback

  • Android app: MyBehaviorCBP; other: phone accelerometer and GPS

  • Intervention condition increased daily walking by 4.9 minutes (β=4.9; P=.02) significantly. Exercise time was increased nonsignificantly by 9.5 minutes (β=9.5; P=.31). MyBehaviorCBP was opened 3.2 times a day (on average). MyBehaviorCBP suggestions were perceived as low-burden (β=.42; P<.001). Back pain was reduced in the intervention condition, but not significantly (β=−.19; P=.24). Participants suggested consideration of weather, weekend or weekday, and level of pain for future interventions
Zhou et al (2018) [96] Daily step goals (real-time, automated adaptive). Push notifications via app. Daily notifications at 8 AM. If the goal was accomplished before 8 PM, a congratulation notification was sent. Steady step goals (10,000 per day) Goal setting and behavioral analytics algorithmk Change in daily step Step goal attainment, weight, height, barriers to being active quiz, IPAQl-short form

  • iOS app: CalFit; other: built-in health chip in the iPhone

  • Mean daily step count was decreased by 390 steps (SD 490) per day in the intervention versus 1350 steps (SD 420) per day in the control from baseline to 10 weeks (net difference: 960 steps, P=.03)

aPA: physical activity.

bMET: metabolic equivalents.

cN/A: not applicable.

dHR: heart rate.

eSB: sedentary behavior.

fTODAY: Tailored Daily Activity.

gAPI: application programming interface.

hStep goals did not increase between cycles.

iEMA: ecological momentary assessment.

jGrounded in artificial intelligence and a subcategory of a broader decision-making framework called multiarmed bandit, which stems from probability theory.

kBehavioral analytics algorithm uses machine learning to build a predictive model–based on historical and goal steps for a particular person and then uses this estimation to generate challenging yet realistic and adaptive step goals based on a predictive model that would maximize the physical activity in the future.

lIPAQ: International Physical Activity Questionnaire.