Effects of personalized exercise prescriptions and social media delivered through mobile health on cancer survivors’ physical activity and quality of life

Highlights

• •The multi-component m-health intervention (personalized exercise prescriptions and social media) significantly increased cancer survivors’ daily steps than the control condition adopting usual practice post-intervention.
• •The multi-component m-health intervention led to greater increases in physical health than the control condition.
• •The social media intervention had greater increased social support than the control condition post-intervention.
• •The multi-component m-health intervention also had greater increased physical health than the social media intervention.

Abstract

Purpose

This study aimed to examine the effects of a multi-component mobile health intervention (wearable, apps, and social media) on cancer survivors’ (CS') physical activity (PA), quality of life, and PA determinants compared to exercise prescription only, social media only, and attention control conditions.

Methods

A total of 126 CS (age = 60.37 ± 7.41 years, mean ± SD) were recruited from the United States. The study duration was 6 months and participants were randomly placed into 4 groups. All participants received a Fitbit tracker and were instructed to install its companion app to monitor their daily PA. They (1) received previously established weekly personalized exercise prescriptions via email, (2) received weekly Facebook health education and interacted with one another, (3) received both Conditions 1 and 2, or (4) were part of the control condition, meaning they adopted usual care. CS PA daily steps, quality of life (i.e., physical health and mental health), and PA determinants (e.g., self-efficacy, social support) were measured at baseline, 3 months, and 6 months.

Results

The final sample size included 123 CS. The results revealed only the multi-component condition had greater improvements in PA daily steps than the control condition post-intervention (95% confidence interval (95%CI): 368–2951; p < 0.05). Similarly, those in the multi-component condition had significantly greater increased physical health than the control condition (95%CI: –0.41 to –0.01; p < 0.05) over time. In addition, the social media condition had significantly greater increased perceived social support than the control condition (95%CI: 0.01–0.93; p < 0.05). No other significant differences on outcomes were identified.

Conclusion

The study findings suggest that the implementation of a multi-component mobile health intervention had positive effects on CS PA steps and physical health. Also, offering social media intervention has the potential to improve CS perceived social support.

1. Introduction

Cancer remains a critical public health issue in the United States.¹ Studies have demonstrated that engaging in regular physical activity (PA) following a cancer diagnosis can offer numerous health benefits, including reducing the risk of all-cause and cancer-related mortality as well as cancer events among cancer survivors (CS).^2,3 Adopting a healthy and active lifestyle can also minimize the risk of cancer and improve the prognosis and quality of life for individuals with cancer.^4,5 However, like the general population, the majority of CS do not meet the recommended minimum of 150 min of moderate-to-vigorous PA (MVPA) per week.⁶ Therefore, it is crucial to develop innovative and scalable PA interventions to promote healthy behaviors and provide appropriate supportive care and guidance to CS.

Advancements in technology, including wearables, apps, and big data analysis, have enabled the delivery of mobile health (m-health) interventions to encourage PA in CS. These interventions are well-suited for broad distribution and offer individualized and timely feedback on behavior modification, and thus are a promising area of technology focused on increasing healthy behaviors.7, 8, 9, 10, 11, 12, 13 M-health utilizes modern technologies, such as smartphone apps, wearables, and social media, to improve the quality of healthcare. These commonly used technologies offer various ways to support the self-regulation of health behavior. Users can monitor different metrics, such as daily steps, activity duration, and energy usage, which can be accessed on the device or a connected mobile app. They can also receive immediate feedback or reminders, set objectives, log activities, and track their advancement. Additionally, these devices can integrate personal data into a social network to encourage self-motivation and peer support.14, 15, 16, 17, 18 Recently, researchers have applied such technologies to promote health by increasing PA and reducing sedentary behavior in CS, and the findings have been promising.19, 20, 21, 22, 23, 24, 25, 26 Despite the positive findings, the literature does have some limitations, such as small sample sizes, a lack of personalized exercise prescriptions, and a lack of using big data from mobile devices.27, 28, 29

Over the past few decades, the Social Cognitive Theory is frequently applied to promote health and facilitate changes in PA behavior.³⁰ The Social Cognitive Theory incorporates 3 interdependent factors: (1) personal factors such as age, weight, and self-efficacy; (2) environmental factors including social support; and (3) behavior, such as levels of PA.^31,32 Researchers who utilize the Social Cognitive Theory aimed to enhance individuals’ PA by addressing and improving both personal and environmental determinants.33, 34, 35, 36, 37 For example, when an individual has a high level of self-efficacy and enjoyment for PA and is in a setting that encourages PA, that person is more likely to participate in PA. According to behavior change literature, multi-component interventions typically yield better outcomes than single component interventions.^27,38, 39, 40, 41, 42, 43 Yet few studies have examined the interactive effects of multi-component m-health technologies on PA and other outcomes in CS.^4,27,44 This highlights a major gap for advancing tailored PA interventions with multi-component m-health programs.

The purpose of this project, therefore, was to examine the effects of a combination of a personalized exercise prescription and Facebook health education intervention, as compared to personalized exercise prescriptions only, Facebook health education only, and attention control conditions, on PA and health outcomes in CS. Based on the literature review and previous studies, we formulated the following specific aims: (1) Examine the effects of the 3 m-health interventions on PA in CS as compared to the control condition over 6 months (Hypothesis 1a: CS in the 3 m-health groups would show greater increases in PA at 6 months compared to those in the control condition; and Hypothesis 1b: CS in the multi-component intervention group would show greater increases in PA than those in the exercise prescriptions only and Facebook only groups); and (2) Determine the effects of the 3 m-health interventions on CS’ health-related quality of life (HRQoL) and PA determinants (Hypothesis 2a: CS in the interventions would show greater increases in these outcomes at 6 months compared to those in the control condition; and Hypothesis 2b: CS in the multi-component intervention would show greater increases in these outcomes at 6 months compared to the other 2 m-health conditions).

This project attempted to examine innovative remote m-health interventions on PA and health outcomes in CS while offering personalized exercise prescriptions based on smart device data. If successful, it can significantly impact the development of effective and remote PA programs to promote health and protect against diseases in CS. Moreover, its findings can guide health professionals and cancer communities to initiate these feasible remote intervention programs to promote PA and health in CS.

2. Methods

2.1. Participants

To detect a mean difference in changes of PA (the primary outcome) across the 4 intervention and control conditions with the power at 80% and significance level at 5%, assuming a SD of 1500 steps/day, this project required 30 participants per condition. Recruitment occurred in the United States. Inclusion criteria for CS were: (1) aged ≥50 years old (this age group was selected because age is a risk factor for CS); (2) had 1 or more of the cancers of interest (i.e., breast, colon, bladder, prostate, endometrium, esophagus, lung, kidney and renal, pelvis, stomach, etc.) because research evidence suggests that regular participation in PA is beneficial in preventing and managing these types of cancers;⁴⁵ (3) completed active cancer treatment at least 3 months prior to enrollment, with the exception of anti-hormonal therapy; (4) had an Android or iOS operating system smartphone; (5) possessed basic English communication capability; (6) had a Facebook account, or were willing to make one; (7) male or female CS; (8) was willing to provide consent and accept randomization assignment; and (9) engaged in some type of PA as assessed by the Physical Activity Readiness Questionnaire. Exclusion criteria for participation in this study were: (1) diagnosed with Stage IV cancer; (2) completed primary cancer treatment (e.g., surgery, radiotherapy) less than 3 months ago with a new cancer diagnosis or recurrence; and (3) declined completion of the informed consent and/or the Physical Activity Readiness Questionnaire.

2.2. Design and procedures

This study was registered at Clinical Trials (ClinicalTrials.gov Identifier: NCT05069519). In this study, all CS received a Fitbit Charge 5 tracker (Google, San Francisco, CA, USA) and installed its companion app on their mobile phones. We then randomized 126 CS to 4 groups for a 6-month intervention period: (1) a personalized exercise prescriptions group (tracked daily PA via Fitbit (Google), shared PA data remotely, and received personalized exercise prescriptions from researchers); (2) a Facebook health education group where participants had access to weekly health education and interacted with one another on a private page; (3) a combination of personalized exercise prescriptions and Facebook health education group (tracked daily PA via Fitbit, received personalized exercise prescriptions, had access to weekly health education, and interacted with one another on Facebook); and (4) an attention control group where participants continued with standard care. The primary outcome was daily steps, and secondary outcomes included HRQoL and PA determinants. Each condition lasted 6 months. All CS underwent identical assessments at baseline (pre-test), 3 months (mid-test), and 6 months (end-point or post-test).

After obtaining institution review board approval in April 2021, we worked closely with cancer communities and used social media (e.g., Facebook cancer groups, Nextdoor, etc.) and email to send the study information and/or flyer to potentially eligible participants. All eligible participants joined a remote Zoom meeting with the research staff to collect demographic data. During this virtual meeting, the participants also discussed their use of the apps and their respective functions with the researchers. Questionnaires assessing HRQoL and PA determinants were administered remotely via Qualtrics (Qualtrics XM, Provo, UT, USA) at baseline and follow-up tests. Along with the study instructions, the research staff mailed the Fitbit trackers (Google) to participants at the beginning of this study to record weekly average PA steps during the 3 waves of data collection. With assistance from several research assistants, the project coordinator led the process of tracking and checking PA steps during each period to ensure data consistency and validity. To minimize data contamination and impact on intervention implementation, we advised that, during the 6-month intervention period, the participants could not discuss protocol-directed PA and study interventions with participants from another group, and vice versa. They could communicate with participants (within the survivor network) from other groups on non-project-related issues. All participants received monetary incentives (a total of USD210) via prorated Clincards as compensation for following the protocols of each of the different conditions (as measured by process evaluation) and taking part in the assessments.

2.3. Intervention conditions

The m-health utilized in this study were Fitbit Charger 5 tracker (Google), Facebook, and their companion apps, as well as emailed weekly personalized exercise prescriptions using previous week Fitbit PA data. All participants received a Fitbit tracker (Google) and were instructed to install its companion app only to monitor their daily PA; they did not use other Fitbit app features.

2.3.1. Personalized exercise prescription condition

Throughout the intervention period participants assigned to this condition continued with standard care and were encouraged to participate in at least 150 min of MVPA per week if their body condition allowed. Participants tracked their PA using Fitbit trackers (Google) and synchronized the Fitbit PA data to its app where they were uploaded to the Fitbit server. Researchers retrieved the data from the Fitbit's Application Programming Interface, then used a predetermined algorithm to select appropriate levels of exercise prescriptions based on participants’ average daily steps for the previous week. They then emailed the previously established weekly exercise prescriptions to participants for the next week. Exercise prescriptions included aerobic exercises, strength training exercises, flexibility exercises, and balance exercises. Among them, the flexibility and balance exercises were mainly adopted from 2 fitness books for CS.^46,47 For detailed personalized exercise prescriptions and weekly PA contents (Zeng et al.,²⁹ Pages 11–16).

2.3.2. Facebook condition

Participants continued with standard care but also received health education tips (developed in our previous studies^4,34) from a private Facebook group that only group members and researchers could access. Additionally, the researchers tracked CS’ login counts and encouraged engagement and peer interactions to facilitate social support. The researchers built 2 private Facebook groups, one for the Facebook only group and another for the combination group in which participants received weekly health education tips for improving 4 PA beliefs: (1) increasing self-efficacy; (2) improving outcome expectancy; (3) promoting social support; and (4) enhancing enjoyment. Please note that participants simply needed to use their Facebook account to join the groups and that they did not need to disclose any of their medical diagnoses to other participants in the study. No personal health information was visible in the Facebook groups. Rather, Facebook gave participants a platform to receive health education and interact with one another in terms of the contents. The project coordinator and several research assistants posted health education content 3 times; they also posted a single PA-related question every week to encourage interactions among participants (Supplementary Fig. 1). Members could only interact with peers within their own private Facebook group. The health education tips were only available to 2 of the private groups (the Facebook and multi-component groups), and non-members could not access them. Additionally, participants were encouraged to access the health education via the Facebook app on their smartphones or mobile devices (i.e., iPad) as opposed to logging into their Facebook account on a public computer, personal laptop, or home computers. Finally, the researchers reminded participants not to share their medical information in the groups. As such, participants’ Facebook accounts and personal information were not breached. The research team monitored the Facebook groups weekly for safety and privacy reasons.

2.3.3. Multi-component intervention

Participants assigned to this condition received both personalized exercise prescriptions and the Facebook health education program. That is, researchers formulated weekly personalized exercise prescriptions based on Fitbit PA data via email as well as created a private Facebook page for the health education of participants in this group.

2.3.4. Attention control

Participants assigned to the control condition continued with standard care (not changing their current routine) and did not receive any other intervention during the intervention period.

For process evaluation, monthly phone check-ups were conducted for all the groups to confirm whether participants were following the research protocols. The project coordinator and a research assistant checked CS’ Fitbit engagement by obtaining the Fitbit step data of all participants every Wednesday and Saturday through an Application Programming Interface from the Fitbit server (Google). This data was then coded using Python 3.4 (Python Software Foundation, Beaverton, OR, USA). In cases where participants failed to synchronize their data with the server or did not wear the Fitbit tracker (Google), reminders were sent via email, followed by phone calls or text messages. Through these efforts, it was ensured that all participants wore the Fitbit tracker (Google) and synchronized their data on a weekly basis. In addition, the project coordinator and 3 research assistants monitored the names of participants who read each post, as well as the likes and comments made by participants in both the Facebook and combination groups. More information is available upon request. Overall, intervention fidelity was continuously monitored for all intervention components in this project.

2.4. Measures

Participants self-reported their height and weight at home. Then, body mass index (BMI) was calculated using the height and weight provided by the participant.⁴³

2.4.1. PA

CS' 1-week daily steps were assessed using Fitbit trackers (Google). Participants were instructed to wear the Fitbit on their non-dominant wrist at all time throughout the study. Their daily steps at 3 time points were used as the primary outcome. The Fitbit-generated data have been widely used in assessing PA among cancer clinical populations.⁴⁴ In this study, the researchers retrieved the Fitbit data from its Application Programming Interface feature. The change scores between baseline and 6 months were used as outcomes in this study.

2.4.2. HRQoL

The Patient Reported Outcome Measurement Information System⁴⁸ was used for the assessment of HRQoL—physical and mental health. We defined physical health according to 3 factors, namely physical function, pain interference, and capability to participate in social roles and activities; to determine mental health, we looked at anxiety and depressive symptoms. This scale has demonstrated acceptable validity and reliability in the clinical populations.^4,33,34 Details of the subscale items and scorings can be found at Gao et al.²⁷ It is important to note that, the lower scores meant better physical and mental health in the present study.

2.4.3. PA determinants

CS’ PA determinants were assessed by standardized self-report (i.e., self-efficacy, outcome expectations, social support, enjoyment) through previously established questionnaires.49, 50, 51, 52, 53 All questionnaires used 5-point Likert scale (e.g., 1 = almost never, 5 = almost always for the Social Support Scale; 1 = strongly disagree, 5 = strongly agree for the Enjoyment Scale). They have demonstrated acceptable validity and internal consistency in previous studies³⁴ as well as this one (Cronbach's α ranged from 0.72 to 0.86). These outcomes were measured at baseline, mid-test, and post-test. The change scores of these 4 variables between baseline and 6 months were used as outcomes in the present study.

2.5. Data analysis

In this study, data were imported into SPSS 27.0 (IBM, Corp., Armonk, NY, USA) for analysis. A descriptive analysis was conducted to describe the characteristics of CS. The unit of analysis was the individual participant. Aim 1: The 4-way analysis of covariance with repeated measures was performed to examine changes in CS’ daily steps from baseline to 6 months, with age and BMI as the covariates. Aim 2: Two separate 4-way multivariate analyses of covariance with repeated measures were used to examine changes in the outcome variables of CS’ HRQoL and PA determinants from baseline to 6 months, respectively. In the present study, the within-subject factor was time (3 times of measurements), the between-subject factor was Group membership, and the covariates were age and BMI. The multiple comparisons among the 4 groups were adjusted by the Bonferroni approach. The pairwise group comparisons were also performed between 3 intervention groups vs. the control group, as well as among the 3 intervention groups. The significance level was set at 0.05 for all analyses, and effect sizes were reported for each comparison. Also, eta-squared (η²) with small, medium, and large effect sizes being designated as 0.01, 0.06, and 0.14 were used as an indices of effect size, respectively.⁵⁴

3. Results

The CONSORT diagram illustrates detailed information regarding participant flow in this study (Fig. 1). Two CS dropped out from the study and another did not strictly follow research protocols prior to the post-intervention data collection and so was removed from further data analysis. Our final sample consisted of 123 CS (age = 60.37 ± 7.41 years, mean ± SD). Full demographic and anthropometric information for the sample at the baseline are displayed in Table 1. Among them, the vast majority were women (n = 120). Nine were born outside of the USA (Canada, China, Colombia, Germany, Japan, the Netherlands, South Africa, UK, and Vietnam). In terms of educational background, 2 had a high school education; 13 participants received some college/technical school education; 59 were college graduates; and 49 had a graduate school education. With regard to their annual income, 6 earned less than USD40,000; 2 participants’ salary ranged from USD 40,001 to USD50,000; 17 earned USD50,001–USD74,999; 19 made USD75,000–USD99,999; 64 earned greater than USD100,000 annually; 15 participants did not respond to this question. A total of 84 CS had private health insurance (e.g., Blue Cross, Kaiser), 9 had Medicare (government insurance for people aged 65 and over), 15 had both private health insurance and Medicare, and the rest had other insurance. In terms of cancer types, 110 had breast cancer, including 11 who also had another type of cancer (ovarian, endometrial, cervical, etc.), the rest had other types of cancers, such as endometrial (n = 3), ovarian (n = 1), fallopian tube cancer (n = 3), thyroid (n = 2), prostate (n = 1), anal (n = 1), and neck/throat (n = 1) cancers. In the last 12 months, 107 CS visited their primary care doctor (or a family doctor) or medical oncologist or hematologist, and 16 did not visit any doctor. Among those who visited doctor, 2 visited a specialist once, 79 visited twice, 4 visited 3 times, 1 visited 4 times, 10 visited 5 times, 11 visited other health care providers, such as a nurse practitioner, homeopathic doctor, acupuncturist, and naturopathic doctor.

Fig. 1.

CONSORT study participant flow diagram. *The participant did not follow the intervention and measurement protocols.

Table 1.

Characteristics of cancer survivors (mean ± SD or n).

Variable	Prescription group (n = 32)	Facebook group (n = 31)	Multi-component group (n = 31)	Control group (n = 29)	Whole sample (n = 123)
Age (year)	61.0 ± 8.02	61.03 ± 6.82	59.13 ± 6.86	60.24 ± 8.06	60.37 ± 7.41
Height (cm)	167.12 ± 7.38	164.55 ± 8.25	166.16 ± 6.14	164.81 ± 6.59	165.69 ± 7.14
Weight (kg)	75.77 ± 17.36	78.80 ± 23.05	74.23 ± 14.88	76.05 ± 14.72	76.21 ± 17.72
BMI (kg/m2)	27.06 ± 5.48	29.00 ± 8.11	26.86 ± 5.06	28.06 ± 5.63	27.74 ± 6.17
Gender	32 females	29 females and 2 males	31 females	28 females and 1 male	120 females and 3 males
Race/ethnicity
Caucasian	30	30	30	26	113
African American		1		1	2
Asian	1	2		1	4
Native American			1		1
Pacific Islander				1	1
Prefer not to state	1	1			2
Hispanic	1 (yes) 31 (no)	1 (yes) 30 (no)	1 (yes) 30 (no)	1 (yes) 28 (no)	4 (yes) 119 (no)

Abbreviation: BMI = body mass index.

Table 2 displays the descriptive results for CS’ daily steps, HRQoL, and PA determinants across the 4 groups at 3 time points. On average, CS displayed a moderate level of daily steps, averaging greater than 7000 steps at baseline. Their HRQoL, namely, physical and mental health, demonstrated relatively moderate levels since the mean values of these 2 outcomes (scores ranging from 1.50 to 1.79) were lower than the median scores (3.00; reverse scored) across time. Notably, CS displayed low-to-moderate levels of PA determinants at baseline (scores ranging from 1.99 to 3.84). A series of analyses of variance were performed on the study outcomes by groups at baseline, and no significant differences were identified (p > 0.05), meaning that CS had equivalent daily steps, HRQoL, and PA determinants prior to the interventions across these 4 groups.

Table 2.

Descriptive statistics of cancer survivors’ outcomes across groups over time (mean ± SD).

Time	Prescription group	Facebook group	Multi-component group	Control group	Whole sample
Steps
Pre-test	7310 ± 3468	7805 ± 4743	7223 ± 3497	7052 ± 3089	7352 ± 3723
Mid-test	7836 ± 4278	7707 ± 3688	8376 ± 3952	7401 ± 3120	7837 ± 3767
Post-test	7730 ± 3751	8225 ± 4561	8737 ± 4159	6840 ± 3276	7898 ± 3985
Physical health
Pre-test	1.79 ± 0.63	1.69 ± 0.58	1.70 ± 0.41	1.65 ± 0.52	1.71 ± 0.54
Mid-test	1.84 ± 0.63	1.75 ± 0.57	1.63 ± 0.49	1.81 ± 0.60	1.76 ± 0.57
Post-test	1.75 ± 0.59	1.79 ± 0.54	1.61 ± 0.47	1.77 ± 0.61	1.73 ± 0.55
Mental health
Pre-test	1.59 ± 0.67	1.50 ± 0.58	1.58 ± 0.60	1.50 ± 0.54	1.55 ± 0.60
Mid-test	1.53 ± 0.52	1.56 ± 0.51	1.58 ± 0.79	1.48 ± 0.52	1.54 ± 0.59
Post-test	1.57 ± 0.67	1.50 ± 0.51	1.40 ± 0.53	1.45 ± 0.48	1.48 ± 0.55
Self-efficacy
Pre-test	2.94 ± 0.89	2.84 ± 0.91	2.94 ± 0.86	2.42 ± 0.78	2.79 ± 0.88
Mid-test	2.73 ± 1.04	2.42 ± 0.79	2.90 ± 1.02	2.19 ± 0.73	2.97 ± 0.54
Post-test	2.90 ± 1.03	2.63 ± 0.92	2.97 ± 0.99	2.15 ± 0.86	2.67 ± 0.99
Outcome expectations
Pre-test	3.56 ± 0.51	3.64 ± 0.53	3.48 ± 0.41	3.46 ± 0.40	3.54 ± 0.47
Mid-test	3.55 ± 0.52	3.57 ± 0.53	3.40 ± 0.54	3.26 ± 0.44	3.45 ± 0.52
Post-test	3.62 ± 0.52	3.59 ± 0.43	3.40 ± 0.54	3.25 ± 0.42	3.47 ± 0.50
Social support
Pre-test	2.18 ± 1.16	2.33 ± 1.15	2.09 ± 0.97	2.28 ± 0.96	2.22 ± 1.06
Mid-test	1.99 ± 1.00	2.22 ± 0.96	1.93 ± 0.90	1.99 ± 0.94	2.03 ± 0.95
Post-test	2.08 ± 0.98	2.46 ± 0.95	2.09 ± 0.84	1.93 ± 0.98	2.14 ± 0.95
Enjoyment
Pre-test	3.84 ± 0.67	3.78 ± 0.70	3.69 ± 0.46	3.75 ± 0.68	3.77 ± 0.63
Mid-test	3.63 ± 0.64	3.77 ± 0.77	3.67 ± 0.61	3.60 ± 0.71	3.67 ± 0.68
Post-test	3.77 ± 0.51	3.80 ± 0.62	3.69 ± 0.55	3.59 ± 0.69	3.71 ± 0.59

As shown in Fig. 2, the group effect for PA daily steps of the analysis of covariance approached the significant level (F (1,117) = 2.22, p = 0.09, η² = 0.09). Pairwise comparison was then conducted to evaluate the differences in step changes across groups. The multi-component intervention group had significantly greater increased steps compared to the control group from baseline to post-test (p < 0.05, 95% confidence interval (95%CI): 368–2951). Although the other 2 intervention groups had increased steps and the control group displayed decreased steps from baseline to 6 months, no significantly inferential differences were identified among these groups over time.

Fig. 2.

Differences of cancer survivors’ daily steps across groups over time.

The 4-way multivariate analysis of covariance did not yield significant group effect for HRQoL (Wilk's lambda = 0.93, F(2,116) = 1.34, p = 0.24, η² = 0.03). A significant group effect was observed for physical health (p < 0.05) but not mental health (p = 0.23). The pairwise comparison suggested that the multi-component intervention group had greater increased physical health compared to the control group from baseline to post-test (p < 0.05, 95%CI: –0.41 to –0.01). In terms of PA determinants, our data did not yield significant group effect (Wilk's lambda = 0.94, F(6,232) = 1.19, p = 0.31, η² = 0.03). Yet the pairwise comparison indicated the social media group had greater increased social support compared to the control group from baseline to post-test (p < 0.05; 95%CI: 0.01–0.93). No other significant differences were identified.

4. Discussion

PA among CS is one of the most effective lifestyle factors in disease prevention and management.^43,55,56 M-health interventions with personalized exercise prescription may improve PA and health outcomes in this population.³⁷ Yet, to our knowledge, few studies have integrated multiple m-health components to facilitate PA and health promotion in CS.^4,33,34,44 In response, the present study attempted to fill this knowledge gap by investigating the efficacy of multi-component intervention (personalized exercise prescription and social media) on PA steps, HRQoL, and PA determinants among CS. Overall, CS displayed a moderate level of daily steps and HRQoL at baseline, as well as low to moderate levels of PA determinants. Study observations yielded mixed results.

For the first aim, we hypothesized that CS in the 3 m-health groups would have greater increased PA steps at 6 months compared to those from the control group. The descriptive analysis yielded that CS in all intervention groups showed a greater increase in steps than their control counterparts over the course of 6 months. However, the inferential tests indicate that only the multi-component intervention group had significantly greater increased steps than the control group. The findings partially support our first hypothesis by corroborating the postulation that multi-component PA interventions are effective at promoting PA in clinical populations, including CS.^4,9,33,57 Specifically, empirical evidence has suggested that health wearables and exercise apps, in partnership with social media programs, could help to improve CS PA participation.^4,33 Although the other 2 single component interventions did not display significantly greater increased steps than the control group, they did further illustrate the potential that m-health programs show for promoting PA in CS. We also hypothesized that CS in the multi-component intervention group would have greater increases in steps than those in the single component intervention groups at 6 months. However, although the multi-component intervention group had a greater increase in steps compared to the social media group at the mid-test, our data failed to support this hypothesis at post-test. It has been postulated that multi-component m-health interventions have advantages over single component interventions.^4,38, 39, 40, 41, 42, 43, 44 For example, in a recent network meta-analysis, McDonough et al.⁹ suggested multi-component accelerometer/pedometer intervention was the most effective strategy for reducing BMI among clinical populations, followed by commercial health wearable-only intervention. It is plausible that all CS received a commercial health wearable (Fitbit) and its companion app in this study and, thus, their healthy behaviors might improve partnered with other single component m-health programs. The findings could have practical implication for researchers and clinicians alike. Offering remote multi-component m-health programs, such as apps and personalized exercise prescription, along with health wearables could be a feasible and effective way to facilitate PA participation in CS.

Our data suggested that, post-intervention, the multi-component intervention group had greater increased physical health than the control group and the social media group had greater increased social support than the control group. This partially supports our second hypothesis, which states that CS in the interventions would show greater increases in HRQoL and PA determinants at 6 months compared to those in the control condition. This finding is in line with previous studies suggesting m-health PA programs could improve physical health³³ or fatigue⁵⁸ among cancer populations. Studies also reported that m-health interventions enhanced individuals’ self-efficacy and social support.59, 60, 61 For example, Gao et al.⁶⁰ used exergaming as the intervention channel and found that children's self-efficacy and social support significantly improved post-intervention. No significant difference in mental health was observed over the course of 6 months across groups. One explanation could be that participants’ mental health was already moderately high at the baseline assessment and, consequently, the ceiling effect might have masked the inferential difference across groups. It is also possible that many CS have had cancer for years and have learned effective strategies for dealing with stress and depression through various stress reduction programs, including yoga and meditation. The finding deserves further investigation; perhaps in-depth interviews would provide insight into a more detailed explanation.

In addition, our second hypothesis predicted that the multi-component intervention would show greater increases in HRQoL and PA determinants at 6 months compared to the other 2 m-health conditions. As seen, slight changes occurred in the HRQoL and PA determinants from baseline to post-test among CS in all 3 intervention groups. Interestingly, the results showed that only the multi-component intervention group had greater increased physical health than those from the social media group, which also partially supports the hypothesis. We believe that CS from both personalized exercise prescription only and multi-component m-health groups benefited from the PA engagement promoted by the exercise plans as their physical health improved. Meanwhile, CS from the social media only group received health education tips and social support from peers but no personalized PA encouragement, which may have led to a difference in physical health between the multi-component intervention group and social media group. Furthermore, since most m-health interventions use bundled packages with multi-components (i.e., wearables, app, support calls, and social media) delivered simultaneously,^4,33,44 it is difficult to disentangle the intervention components to determine at which levels and in what combination(s) individuals’ PA and health outcomes are influenced. For this reason, the present study used traditional experimental design (2 × 2 factorial design) to test each component individually along with their interaction effects. It is recommended that, in future studies, the Multiphase Optimization Strategy framework may be adopted to evaluate intervention components’ individual and combined effects and to compare the individual effects of multiple intervention components simultaneously.⁴⁴

The present study has the following strengths: (1) all participants received the base intervention with a Fitbit and its companion app, and as a result they all benefited from the PA monitoring and promotion in this study; (2) it was the first study to apply multiple novel m-health components (i.e., fitness wearable, exercise apps, social media app) and big data applications into home-based practice in CS during the pandemic, as well as to explore the individual and combined effects of different intervention components; (3) it offered weekly personalized exercise prescriptions to CS according to objective Fitbit data from the previous week; and (4) intervention fidelity was ensured through process evaluation, which included weekly Fitbit data check-ups and monthly phone follow-ups. Nevertheless, study findings should be interpreted with caution given the following limitations. First, the vast majority of CS were breast CS. There is a need to recruit CS of other cancer types in the future. Second, the majority of participants were well-educated Caucasian women who were of relatively high socioeconomic status, which may limit external validity of the findings. A few CS knew each other from a cancer network and many had average daily steps of over 7000 at baseline. Hence, more diverse, less-active samples are needed in the future. Third, personalized exercise prescriptions were offered manually via email every week after research staff retrieved Fitbit data and allocated the weekly personalized exercise prescriptions based on weekly steps. Future studies should consider utilizing sophisticated apps, guided by algorithm and artificial intelligence, to automate these processes. To this end, the researchers recently developed a prevision exercise app, iFitRx (iRecOO Mobile Technology, Lewes, DE, USA) to deliver automated personalized exercise prescriptions based on big data analysis and predetermined algorithms. Finally, since the programs were remotely delivered and all data except for Fitbit data were self-reported, we missed the opportunity to assess objective outcomes such as BMI, MVPA, and functional fitness. In the future we may include objective instruments for assessing physical and physiological outcomes (e.g., MVPA via accelerometers, body composition).

5. Conclusion

According to the study observations, the implementation of a multi-component m-health intervention had positive effects on the PA steps and physical health of CS at the 6-month follow-up. The social media m-health intervention also showed promise for enhancing perceived social support among CS. These findings shed new light on the roles remote m-health could play in terms of encouraging PA and health in CS. They are particularly meaningful results given that CS are at risk for low PA and HRQoL, particularly during and beyond the COVID pandemic. M-health with multiple emerging technologies has the potential to improve PA and hence reduce cancer-related death and cancer events (e.g., recurrence) in CS. According to our findings, professionals and practitioners may choose to utilize remote precision PA programs because these m-health interventions have proven to be feasible and cost-effective ways of promoting health among CS.^33,34

Acknowledgments

This study was funded by College of Education and Human Development Acceleration Research Award at the University of Minnesota Twin Cities, USA. At the time of this study, the first author was with the School of Kinesiology at the University of Minnesota-Twin Cities, USA.

Authors’ contributions

ZG developed the project, wrote the manuscript, and prepared the figures; SR coordinated with the intervention implementation, was involved in data collections, helped with the manuscript preparation, and critically revised it; WZ, KA, and MH coordinated with the intervention implementation and involved in data collections; RZ, AB, JW, and JS helped with the manuscript preparation and critically revised it. All authors have read and approved the final version of the manuscript, and agree with the order of presentation of the authors.

Competing interests

The authors declare that they have no competing interests.