Social Media and mHealth Technology for Cancer Screening: Systematic Review and Meta-analysis

Introduction

Background

The use of mobile health (mHealth) technologies and social media in the health care sphere has now become widespread [1-6] and has enabled the rapid sharing of health information, the launching of health promotional campaigns, access to peer support groups, and facilitation of appointment reminders [1,2,4,6]. The World Health Organization has defined mHealth as the use of mobile wireless devices for medical and public health practice [1]. Social media allows those with access to information and communication technology to become content creators and share content with others in virtual communities or networks in addition to accessing information and connecting communities [1,6]. The use of mHealth and social media for health presents an important opportunity to reach health consumers, as these technologies and platforms can provide more frequent interactions, deliver tailored material, and increase accessibility to health information [1], and they now constitute a major way of communicating and advertising. In addition, as access to mobile devices and the internet in low- and middle-resource nations is reported to be comparable with those in developed countries, mHealth and social media may play a role in closing the gap in health disparities between high- and low-resource nations [1,7].

With almost 19 million people expected to be diagnosed with cancer in 2020, cancer is one of the leading causes of death globally [8]. Cancer screening has been shown to reduce disease-specific mortality for a number of cancers [9-12], and as a result, many jurisdictions have implemented population-based screening programs [13,14]. However, screening participation remains suboptimal across jurisdictions and cancer types [13-16]. Emerging research has explored the use of social media and mHealth for cancer screening [17-21]. However, we currently lack an understanding of how effective mHealth and social media can be for cancer screening participation.

Objectives

This systematic review and meta-analysis aims to explore the effectiveness of social media and mHealth interventions to increase cancer screening participation and intention for screen detectable cancers.

Methods

Study Design and Registration

This systematic review was registered with the International PROSPERO (Prospective Register of Systematic Reviews; registration #CRD42019139615) and was written and reported according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) checklist [22].

Inclusion and Exclusion Criteria

Studies included in this systematic review were randomized controlled trials (RCTs) or quasi-experimental studies with a pre- and postintervention design reporting on the effectiveness of an mHealth or social media intervention on cancer screening participation or intention. We included studies pertaining to breast, cervical, colorectal, prostate, or lung cancer, as guidelines for screening exist for these cancers. We defined mHealth interventions as those that delivered health-related information via telecommunication or other wireless technologies (eg, smartphones and tablets) [4]. Social media interventions included those delivered on an already established or new purpose-built social media platform where users could create a profile and share content with other users (virtual communities) [1]. Any comparator was acceptable, including a nonintervention group; an alternate, nonsocial media, non-mHealth intervention; or studies with a pre- and postintervention design. We included studies with multifaceted interventions if at least one component involved a social media– or mHealth-based strategy. Studies were restricted to those conducted in adults aged 18 years or older and articles published in English. In case we were unable to access full-text articles for relevant abstracts, we contacted study authors to obtain the articles. If the authors did not respond, we included the abstract if we could ascertain the eligibility criteria and if the data on the primary or secondary outcome were available. Commentaries, editorials, letters, and reviews were excluded. We also excluded articles published before 2000 because the use of social media was not widespread before this time [4].

Search Strategy

The search strategy was developed by a senior information specialist (TK) and used a combination of text words and MeSH (Medical Subject Headings) terms depending on the database to capture the following concepts: cancer, screening, and social media or mHealth interventions. The search strategy was peer reviewed by a second information specialist in accordance with the Peer Review of Electronic Search Strategies checklist [23] and has been previously published [24].

Information Sources

The search was conducted using the following databases: MEDLINE, Embase, PsycINFO, Scopus, CINAHL, the Cochrane Central Register of Controlled Trials, and Communication & Mass Media Complete from inception to May 31, 2019. The search was updated on July 17, 2020.

Data Management

We used systematic review software (DistillerSR, Evidence Partners Incorporated) to manage records during the screening and study selection phases.

Study Selection

Two independent reviewers (AR and FD) used a piloted data collection form and screened the studies in three stages: title, abstract, and full text. Citations that either reviewer considered potentially eligible at the title stage were included to maximize sensitivity in the early stages of screening. Inclusion in the abstract and full-text screening stages required consensus between the reviewers. Discrepancies between the reviewers at the abstract or full-text stages were resolved by discussion.

Data Extraction

Two reviewers independently extracted data from the included studies using a piloted data collection form in Excel (Version 15.0; Microsoft). Any discrepancies were resolved by discussion. Information extracted from each study included study characteristics (authors, date of publication, location or country, funding, and study design), participant characteristics (sample size, age, sex, ethnicity, and eligibility), intervention details (type of intervention, components, comparator or control group interventions, follow-up or duration, technology platform, and delivery of intervention by whom), and outcomes of interest (screening participation or intention including timeframe).

Outcomes

Screening participation (primary outcome) was defined as the proportion of adults who participated in the screening. This included self-reported outcomes as well as those confirmed through administrative records. Screening intention (secondary outcome) was defined as per the primary study authors. Typically, this is measured as the written intention to undergo screening within a specified timeframe (eg, within the next 3 months or 6 months).

Assessment of Bias

The Cochrane Risk of Bias 2 tool [25] was used to assess the quality of RCTs, and the Cochrane Effective Practice and Organization of Care framework was used to assess bias in pre- and postintervention studies [26]. The risk of bias assessment was independently completed for each study by 2 reviewers (AR and FD). Discrepancies were resolved by discussion or by a third investigator if needed. The Robvis tool was used to create a risk of bias plot [27].

Data Synthesis and Analysis

The study, participant, and intervention characteristics and the risk of bias assessments are presented descriptively. We categorized interventions based on their nature, including (1) reminders, (2) education or awareness, (3) navigation or counseling, (4) peer support, (5) decision aids, and (6) mixed. We report on the outcomes of interest in absolute and relative terms and pooled odds ratios (ORs) for screening participation from RCTs using the Mantel-Haenszel method in a random-effects model. If the outcome was measured at several time points, we used the values from the longest follow-up for our study. In RCTs where several intervention arms had a social media or mHealth component, we included them in our analysis and divided the proportion screened of the control or comparison group equally by the number of intervention arms of interest to maintain the same proportion of those screened while not counting the sample size of the control group more than once, as recommended by Cochrane [28]. Forest plots were created to graphically display results stratified by cancer type and the nature of the intervention. Statistical heterogeneity was calculated using the I² statistic, where a cutoff of ≥75% was defined as considerable heterogeneity [28]. We conducted a sensitivity analysis in which we excluded articles that were assessed to have a high risk of bias. In addition, we conducted sensitivity analyses to explore whether the overall pooled effect estimate would differ for studies measuring the outcome of cancer screening participation through self-reporting compared with objective or administrative records and for studies conducted in low- and middle-income countries (LMICs). We checked for publication bias for the primary outcome among the RCTs using a funnel plot. Statistical significance was set at a two-tailed P<.05. Meta-analyses were performed using Review Manager (RevMan, The Cochrane Collaboration) 5.0.

Results

Search Results and Characteristics of Included Studies

A total of 18,008 records were identified in the search. After duplicates were removed, 17,788 titles, 2607 abstracts, and 687 full-text articles were screened. After all the eligibility criteria were applied, 39 articles were included [29-67] (Figure 1). Table 1 presents a summary of the included RCTs (n=30), and Table 2 presents an overview of the included pre- and postintervention studies (n=9). Briefly, the studies that were included were published between 2011 and 2020 and conducted in North America, Europe, Asia, and Africa. Most of the studies (35/39, 90%) described mHealth interventions, and 10% (4/39) of them included social media. The most common type of intervention was mixed (n=19), followed by reminders (n=13), education or awareness (n=6), and peer support (n=1). Mixed interventions were most commonly a combination of reminder and education strategies. There were 16 studies focused on cervical cancer, 14 on colorectal cancer (CRC), 7 on breast cancer, and 1 each on lung and prostate cancer screening. The interventions were implemented by public or private screening programs, university-based research teams, or health care centers or units.

‎View this figure

The most common reminder strategies used were text message reminders [29-39,41-43,46-55,57-65]. Educational strategies most commonly included general health information about the specific cancer and information about cancer screening, including the importance of screening. Although text messages were commonly used to deliver educational information [34,35,37,42-44,46,48,49,53-55,59,62,64,65], some studies also used electronic posters or infographics, CDs, videos, mobile apps, and podcasts [37,42,44,45,55,59,63]. Education was also provided through in-person educational or training sessions in some cases in addition to a social media or mHealth strategy or in the comparison groups [55,63]. Educational interventions using social media included social media campaigns [56] or sharing information or daily posts about screening or cancer with participants who were members of a group (virtual community) on a social media platform [66,67]. Peer support interventions on social media also leveraged groups to support participants of that virtual community through the sharing of personal stories and narratives [40]. Outcomes were measured at several time points, including the proportion attending a scheduled appointment or those participating in screening within 2 weeks [52], a month [65], 45 days [35], 60 days [29,34,41,47], 3 months [35,42,50], 3-5 months [38,39,51], or 6 months [30,31,33,36,40,45,53].

There was wide variability in the study participants. For example, the included participants were targeted based on geographical region in some studies [34,51,56] or by their profession as elementary school teachers [37], entrepreneurs [63], or university students [49,59]. Some studies were targeted to specific racial and cultural groups [44,46,54,58,67], whereas others included gay and bisexual men only [48] or women who were HIV positive [60]. The intervention intensity also differed between the studies. For example, some interventions included sending only a single text message reminder [29,31,33,38,39,41,51], whereas others included sending 22 text messages over 16 days [54] or 21 messages over a 7-day period [44]. For social media interventions, participants in one study received three daily posts over a 12-week period [67] or as many as 20 posts per day over 5 days [66].

Quality Assessment

Risk of bias assessments for the included studies are shown in Figures 2 and 3. Briefly, 27% (8/30) of the included RCTs were classified as high risk, 23% (7/30) as having some concerns, and the remainder (15/30, 50%) were classified as low risk. Common reasons for being classified as high risk included having some concerns in several domains, including bias arising from the randomization process, effect of assignment to intervention, and measurement of the outcome. All pre- and postintervention studies were classified as high risk. Figure 4 displays the funnel plot used to check for publication bias. The x-axis represents the effect estimates, whereas the y-axis represents the study size or precision. The funnel plot generated may suggest some publication bias because of the lack of studies in the bottom left corner of the plot representing studies with small effect sizes and variances.

‎View this figure

‎View this figure

‎View this figure

Primary and Secondary Outcomes

The absolute effect of being screened in the intervention arms was 22.22% (13,115/59,017). There was an absolute risk difference of 14% (95% CI 13.12-14.33) between the intervention and comparison arms, with the proportion screened in the comparison arms being 35.94% (12,524/34,872). When stratified by cancer type, the absolute proportion screened in the intervention arms was 71.68% (3935/5489) for breast cancer compared with 64.11% (7096/11,067) in the comparison arms (risk difference 8%; 95% CI 6.08-9.06). For cervical cancer, there were 35.23% (2382/6760) screened in the intervention arms compared with 28.26% (1548/5478) in the comparison arms. For CRC, the proportion screened in the intervention arms was 14.53% (6798/46,768) and 21.17% (3880/18,327) in the comparison arms, with a risk difference of 6% (95% CI 5.96-7.31).

The overall pooled OR for cancer screening participation among the included RCTs was 1.49 (95% CI 1.31-1.70; Figure 5), indicating that the odds of getting screened increased by 49% for those who received a social media or mHealth intervention. However, considerable heterogeneity was observed (I²=88%). Similar effect estimates were observed when stratified by cancer type, with the largest effect observed for cervical cancer screening studies (OR 1.71, 95% CI 1.34-2.19; Figure 5). Stratification by cancer type did not reduce the heterogeneity. When we conducted a sensitivity analysis excluding trials assessed to have a high risk of bias, the overall pooled OR and I² remained stable (OR 1.54, 95% CI 1.33-1.78; Figure 6). The overall pooled OR was not significant when including only studies measuring screening participation through self-reporting (OR 2.09, 95% CI 0.96-4.53). The overall pooled effect estimate remained stable when including only studies that captured the outcome through administrative records (OR 1.46, 95% CI 1.28-1.66). When we included only studies conducted in LMIC settings (n=3), the overall pooled OR was 3.29 (95% CI 1.02-10.60) with considerable heterogeneity (I²=93%). However, the pooled OR increased to 5.50 (95% CI 3.19-9.51) with only moderate heterogeneity (I²=38%) when only studies with a low risk of bias were included (n=2). We also conducted subgroup analyses by meta-analyzing studies based on the nature of the intervention. The results showed an overall pooled effect estimate of 1.23 (95% CI 1.08-1.41) for reminder interventions (Figure 7) and 2.07 (95% CI 1.49-5.83) for mixed interventions (Figure 8). Heterogeneity did not change when subgroup analyses were conducted.

‎View this figure

‎View this figure

‎View this figure

‎View this figure

Table 3 presents the results of the secondary outcomes of screening intention. Six studies (3 RCTs and 3 pre- and postintervention studies) reported on screening intention, with two studies reporting on screening intention only. There was minor variability in the measurement of screening intention among the studies. For example, screening intention was treated as a dichotomous variable in some studies [37,45,54,58] or scored using a four-point [44] or five-point [59] Likert scale in others. Half of the studies (3/6, 50%) focused on cervical cancer, followed by breast cancer (2/6, 33%) and CRC (1/6, 17%). The intention to screen increased in all studies reporting on this outcome, except for one in which it decreased. The highest increase in screening intention was observed in the study by Lee et al [58], where there was a 24% absolute increase in the intent to receive a Papanicolaou test postintervention (19/30, 63% preintervention and 26/30, 87% postintervention). The study included a 7-day text message–based intervention that included a high level of engagement with participants through quizzes, questions, and engagement in conversation [58]. Owing to the variability in how screening intention was measured or captured, we did not perform a meta-analysis on these data.

Discussion

Principal Findings

Our systematic review identified 39 studies describing the effectiveness of social media and mHealth interventions on cancer screening participation and/or intention. The overall pooled OR for cancer screening participation was significant, favoring the intervention arm (OR 1.49, 95% CI 1.31-1.70). Effect sizes were similar across all cancer types, and estimates remained stable when trials deemed to be at high risk of bias were excluded, indicating that social media, and particularly mHealth interventions, can be effective for increasing cancer screening participation.

Two systematic reviews on this topic were published in 2017 [17,18]. Uy et al [17] evaluated the effectiveness of text messaging interventions on cancer screening and identified nine studies that met the inclusion criteria. Absolute screening rates for text messaging interventions were 1%-15% higher and relative screening rates were 4%-63% higher for intervention recipients in their study [17]. The authors concluded that text messaging interventions moderately increased screening rates for breast and cervical cancer; however, additional research is needed to better quantify this relationship [17]. Tamuzi et al [18] explored mHealth interventions for cervical cancer screening only. Their review identified 17 studies, and the authors were able to perform a meta-analysis on the results by type of intervention [18]. However, their definition of mHealth was different from ours. In their study, Tamuzi et al [18] included telephone, letter, and text message reminders, whereas only text message reminders were included in our study based on our adopted definition of mHealth interventions. Text message reminders are different from these other approaches because they are sent only to mobile devices compared with telephone calls, which may be made to landlines, for which coverage has been decreasing. In addition, text messages can be sent instantly, whereas letter or postcard reminders need to be delivered by the post. Moreover, text messages have the opportunity to reach those with no fixed addresses. For example, a recent systematic review on technology use among homeless adults showed that a majority (94%) owned a cell phone [68]. Overall, Tamuzi et al [18] found that call reminders were the only intervention to show a statistically significant pooled effect estimate. Only one study included in their review reported on the effect of text message reminders, and a meta-analysis of this type of intervention was, therefore, not possible [18].

The results of this study enhance our understanding of the effectiveness of social media and mHealth interventions for cancer screening. Although both previous reviews were published in 2017, nearly 44% (17/39) of the studies in this area have been published since that time. Our review provides a comprehensive and more contemporary understanding of this topic. In addition, although previous reviews focused primarily on breast and cervical cancer, our study provides valuable insights into the effectiveness of these interventions in CRC screening as well. We included 13 studies focused on CRC in our meta-analysis and found a significant pooled effect estimate, suggesting that the use of these types of interventions can be extended to CRC as well. In comparison with the study by Uy et al [17], we found that absolute screening rates between the intervention and comparison groups were higher in our study. This may suggest that multicomponent interventions that couple social media or mHealth with additional strategies may be more effective at increasing screening rates compared with mHealth or social media strategies alone.

The results of our study must also be understood within the larger context of interventions for cancer screening. Brouwers et al [69] conducted a systematic review of interventions for increasing cancer screening rates and looked at client reminders, client incentives, mass media, small media, group education, one-on-one education, reducing structural barriers, reducing out-of-pocket costs to clients, provider assessment and feedback, and provider incentives. Similarly, the authors found wide heterogeneity across studies and interventions and chose not to meta-analyze their data. For example, their results showed that small media interventions, including videos or printed materials such as letters, brochures, newspapers, magazines, and billboards, resulted in a point percentage increase for cancer screening participation ranging from −32.8% to 26% among studies on breast cancer, cervical cancer, and CRC [69]. Our review showed that the absolute difference between the intervention and comparison arms was 14%. The magnitude of effect varied considerably among and between intervention categories in the review by Brouwers et al [69], suggesting that additional evidence is needed for interventions related to client reminders, mass media, group education, one-on-one education, reduction of structural barriers and out-of-pocket costs, and provider incentive interventions. Given the need for additional, high-quality evidence, it is difficult to ascertain whether social media and mHealth interventions fare similar, better, or worse than non-mHealth or non–social media interventions. In addition, costs should also be considered when making any comparisons between the effectiveness of these interventions to inform the translation of these findings into practice.

Although the pooled effect estimate in our meta-analysis was consistent in the subgroup and sensitivity analyses, significant heterogeneity remained. This may be because of the variability in populations, interventions, or outcome measurement across studies. For example, the populations randomized in the studies in our review included all adults up to 79 years [44], or highly specialized populations such as emergency department patients [62] or HIV-positive individuals [60]. Moreover, many of the studies included insured samples, which may not be reflective of population-level interventions, and therefore, must be considered in the generalizability of these results. In addition, the follow-up and the intensity of each intervention varied across studies. For example, some studies may have sent a single text message reminder [37], whereas other interventions included sending multiple text messages in combination with telephone reminders [33]. Interestingly, when we looked at studies conducted in LMIC settings and excluded those with a high risk of bias, the overall pooled OR was even larger with only moderate heterogeneity. These results suggest that the effectiveness of these interventions for cancer screening participation may be more pronounced in these settings. This may be because there may be a limited number of other campaigns in these resource-low settings, whereas access to mobile phones and the internet has been reported to be comparable with that of developed nations [1].

Only a limited number of studies (n=4) tested social media interventions. As such, our results are more indicative of the effectiveness of mHealth interventions. A narrative systematic review focusing on describing the characteristics of social media interventions used for cancer prevention and management found that cancer screening participation or intention was not measured in any of the 18 studies included in the review [70]. The most common outcome measured in these studies was knowledge [70]. Although research related to social media and cancer screening participation has started to emerge [71], the inclusion of this work was limited in our review, as there are few RCTs and before and after comparisons also capturing the outcome of screening participation or intention. This suggests areas for future research to generate more evidence on the use of social media interventions for cancer screening participation. In addition, very few studies have been conducted on prostate and lung cancer screening, which is similar to what was observed in a previous study [17].

Our review and meta-analysis included a variety of mHealth and social media interventions and multicomponent interventions. Our review is comprehensive and contemporary and uses a rigorous systematic approach to screen and review the literature. As such, it includes a large number of studies for the most established screening programs for breast cancer, cervical cancer, and CRC. Owing to the large number of studies included in our review, we were able to calculate pooled effect estimates by cancer type to inform practice and future research. However, this study has limitations. Although we made every effort to obtain full-text articles, there were some records identified from our search that we could not locate. We also did not calculate a Cohen κ coefficient to report the interrater reliability between the 2 reviewers. Our review is also limited in regard to social media interventions, as only four studies were identified, with only one RCT included in the meta-analysis. This may be a reflection of current practice or due to the fact that it may be more difficult to link direct patient outcomes with the use of social media.

Conclusions

In conclusion, our results suggest that mHealth interventions may have a significant effect on cancer screening participation, particularly for breast cancer, cervical cancer, and CRC screening. Screening programs should consider the use of mHealth interventions to increase screening participation. Further research focusing on social media interventions for cancer screening participation is needed, as there was insufficient evidence available at the time of this review.