Table 4.

Study findings

Author (year)	Uptake	Retention	Physical activity: measure used, how measured, follow-up period	Within- or between-group differences
Bond et al. (2014) [30] Thomas & Bond (2015) [47]	Not reported.	30 completed out of 35 consented and enrolled; 2 excluded for not following protocol and three did not complete.	Change in daily time spent in sedentary behaviour (primary outcome), daily number of minutes accrued in walking breaks. Objectively through independent measurement (separate accelerometer). Study duration: 21 days, 7 days in each condition.	Thomas: Daily number of minutes accrued in walking breaks was M(SE) = 37.24 (1.85) in the 3-mins condition and M(SE) = 38.73 (1.86) in the 6-mins condition. This did not differ between-groups, but both were significantly higher than the M(SE) = 32.49 (1.93) in the 12-mins condition. Number of daily minutes decreased significantly as a function of days. Bond: Proportion of daily time spent sedentary was M(95% CI) = 66.3 (61.7–71.0) in the 3-mins condition; M(95% CI) = 66.6 (61.5–71.7) in the 6-mins condition, and M(95% CI) = 69.0 (64.7–73.2) in the 12-mins condition. There was a significant difference (p < .04) between the 3-mins and 12-mins conditions in terms of change since baseline. The 3-mins condition produced significantly larger increases in proportion of time performing light intensity activity compared to the 12-mins condition (p = .04).
Ding et al. (2016) [31]	Two participants (one intervention, one control) did not start as app not compatible with their Samsung S4 phones.	One participant did not complete due to the app consuming too much battery power.	Step counts during weeks 2–4: perceived effectiveness of the app in terms of encouraging the participant to walk more. Objectively through in-built sensor/accelerometer (smartphone or smartwatch). Exit interview after four weeks, some data were collected at 10 pm each day (e.g. whether app encouraged them to do other activities).	Step counts during weeks 2–4: M(SD) = 40,350 (12,458) in intervention; M(SD) = 45,744 (15,541) in control group. No significant difference between intervention and control group in perceived effectiveness (M(SD) = 3.11 (0.93) versus 2.29 (0.95) on a 5-point Likert scale). Significant between-group difference in self-reported effectiveness of the app to encourage them to do other physical activities (M(SD) = 3.22 (0.67) versus 1.43 (0.79); p = .0002).
Finkelstein et al. (2015) [32] Ouyang et al. (2015) [48]	Not reported.	27/30 enrolled completed the study.	Number of episodes of prolonged inactivity (>  2 h) per day when the inactivity reminder was active, compared to not active. Objectively through in-built sensor/accelerometer. Eight weeks, of which four weeks with the inactivity monitor active and four weeks with the monitor inactive.	Inactivity expressed as fraction of consecutive two-hour slots between 8 am and midnight during which steps are less than 20. Inactivity was significantly lower (p < 0.02) during “message-on” periods (24.6%) as compared to the “message-off” periods (30.4%). Group A: M(SD) = 0.32 (0.23) during message-off period and M(SD) = 0.22 (0.14) during message on period. Group B: M(SD) = 0.28 (0.12) during message-off period and M(SD) = 0.28 (0.14) during message on period. Within-group difference between the two periods was significant (p < .004) for group A but not significant for group B. Higher but non-significant step counts in both groups during the message-on period compared to the message-off period (within group).
Gouveia et al. (2015) [35]	Not reported.		Distance walked per day (km). Objectively through in-built sensor/accelerometer. First 12 weeks of intervention delivery (out of 10 months).	Participants who updated their goal walked more per day (median(IQR) = 6 (3–10) km) than those who did not (median(IQR) = 2 (1–4); p < .05. There was a positive association between goal and distance walked per day (p < .05), and a negative correlation between goal set and accomplishment (p < .01).
He & Agu (2014) [36]	Not reported.	One participant dropped out due to loss of interest in the study and did not care about research credits (compensation).	Not reported.	Not reported.
Hermens et al. (2014) [49] Tabak (2014) [20]		Hermens: 2/8 patients provided activity data for analysis. Tabak: One participant dropped out after four weeks due to personal reasons, one stopped after six weeks due to non-COPD related medical reasons.	Mean activity per day for those days on which at least six hours of data were available. Other measures included whether the participants reached their physical activity goal, the number of days when accumulated activity was between 90 and 110% of the goal, and the number of days on which the balance goal was reached (> 90%). Objectively through independent measurement (separate accelerometer); and 6-min walking test to measure exercise capacity. During the three months of intervention and for one week at three months after the intervention.	Data were reported for each participant. Five participants increased their activity levels and four participants improved their activity balance between baseline (one week before the start of the intervention) and the end of the three-month intervention. Three participants had a clinically significant improvement in exercise capacity of > 25 m and one patient of 24 m. At three-month follow-up, no participant had maintained their increases in activity levels and only two participants maintained their improved activity balance. The percentage of days on which goals were achieved ranged between 23 to 59% for activity levels and between 21 and 85% for balance.
Lin et al. (2011) [27] Lin (2013), Chapter 5	Not reported.	Not reported.	Not reported.	Not reported.
Lin (2013), Chapter 6	Not reported.	Not reported.	Perceived change in physical activity. Self-report. End of intervention period.	18/21 reported that they had become a little or much more active.
Pellegrini et al. (2015) [34]	Not reported.	8/9 completed the intervention and were followed up.	Proportion of the day spent sedentary and proportion of the day spent in light physical activity. Objectively through independent measurement (separate accelerometer). One month, which was during the intervention period.	Data are reported within-participants. 7/8 participants reduced the proportion of day spent sedentary and increased time spent in light physical activity: in these 7 participants, sedentary time decreased by M(SD) = 8.1 (4.5)%, p = .003) between baseline and one month, and light physical activity increased by M(SD) = 7.9 (5.5)%, p = .009. Inclusion of the eighth participant reduced the decrease in sedentary time to the level of a trend (p = .08) and maintained a significant increase in light activity (p = .047). Breaks in sedentary time decreased over the month by M(SD) = 15.8 (8.8) (p = .003); whereas break duration increased by 1.0 (0.5) mins (p = .002).
Rabbi et al. (JIMR, 2015) [28]	Not reported.	N = 17 completed the 3-week period.	Behaviour change from participants’ logs of daily activity. Self-report. Collected during the three weeks of intervention delivery.	78% of participants in intervention group showed positive trends (upward trend to longer walks from first to third week) whereas 75% of control group participants showed negative trends (p = .05). Intervention group walked 10 min per day more during the 3 weeks (within group) whereas control group participants showed no change (between group-difference in change in walking duration p = .055).
Rabbi et al. (UBICOMP, 2015) [25]	Not reported.	Not reported.	Behaviour change from participants’ logs of daily activity. Self-report. Data collected during the 2–4 weeks of control condition and subsequent 7–9 weeks of intervention condition; total study duration was 14 weeks maximum.	Significant improvements (within-participant) were found over the final three intervention weeks (compared to 2–4 weeks of control condition) in minutes of walking per day (intervention M(SD) = 24.9 (7.4); control M(SD) =14.5 (5.9); d = 1.41; p < .005) and calories burned through non-walking exercise per day (intervention M(SD) = 126.7 (35.3); control M(SD) = 83.5 (33.1); d = 1.23; p < .05). Patterns over time showed that minutes of walking per day did not change much during the intervention period, and non-walking exercises were maintained during this period.
Rajanna et al. (2014) [37]	Not reported.	Not reported.	Not reported.	Not reported.
Van Dantzig et al. (2013) [33]	Not reported.	Not reported.	Computer activity (proxy for sedentary behaviour) and physical activity during the 30 mins before a text message were compared with computer activity and physical activity during 30 mins after the text message. Objectively through independent measurement (separate accelerometer); computer activity through specially installed software. Collected during the 6-week intervention period (study 2).	Study 2: Average computer activity before the text message: 28.3 (SD 0.32) mins in the intervention and 27.7 (0.43) mins in the control group. Average computer activity after the text message: 18.3 (4.0) mins in the intervention group (10.0 mins reduction within-group) and 21.8 (2.9) mins in the control group (5.9 mins reduction). The decline in computer activity was higher in the intervention than control (p < .001). Physical activity (proportion of active minutes) before the text message: 0.68 (SD 0.49) in the intervention group and 0.42 (0.26) in the control group. After the text message: 0.71 (0.41) in the intervention and 0.47 (0.24) in the control group. The between-group difference was not significant. Subgroup analyses showed that reduction in computer activity did not differ as a function of the number of messages read, so the text message only was sufficient to trigger a break and persuasive content may not be needed.
Van Dantzig et al. (2018) [29]	Not reported.	Ten participants were excluded because they did not meet one ormore of the study criteria (e.g., they had been ill or away from their work for several days, or their lifestyle had varied drasticallydue to external or unforeseen circumstances).	Average daily step count per participant. Objectively through independent measurement (separate accelerometer). Collected during a 2-week calibration period, 1-week intervention period and 1-week fade-out period.	The authors do not report precise figures for intervention and control group. Mean daily step count appears to be approx. 8200 in the intervention group and 7600 in the control group during the intervention period, and 7500 in intervention and 7600 in control during the fade-out period. Between-group differences were not statistically significant. Authors divided each group into three clusters: cluster 1: steps<=6500, cluster 2: 6500 < steps<=9500, and cluster 3: steps> 9500. The increase in mean step count from calibration to coaching period for intervention and control groups: Cluster 1: 16 and 19%; Cluster 2: 17 and 6%, and Cluster 3: − 4% and − 7%, respectively. Step counts in each cluster and group were as follows. Calibration period: M(SD) in cluster 1 intervention (n = 10) = 5273 (594); cluster 1 control (n = 11) = 4866 (1125); cluster 2 intervention (n = 15) = 7708 (786); cluster 2 control (n = 17) = 7616 (854); cluster 3 intervention (n = 3) = 11,200 (1675); cluster 3 control (n = 4) = 11,383 (728). Intervention period: M(SD) in cluster 1 intervention = 6115 (985) (larger than during the calibration period, p = 0.037); cluster 1 control = 5771 (1496) (larger than during the calibration period (p = 0.042) and fade-out period, p = 0.01)); cluster 2 intervention = 9002 (1877) (larger than during the calibration period, p = 0.007, and fade-out period, p = 0.035); cluster 2 control = 8088 (1965) (non-significant differences with calibration and fade-out period); cluster 3 intervention = 10,789 (2118) (all differences non-significant); cluster 3 control = 10,597 (474) (all differences non-significant). Fade-out period: M(SD) in cluster 1 intervention = 5481 (732); cluster 1 control = 5903 (1095); cluster 2 intervention = 8055 (1413); cluster 2 control = 7864 (2036); cluster 3 intervention = 10,569 (1581); cluster 3 control = 11,356 (2522).