This systematic review was registered with the PROSPERO (International Prospective Register of Systematic Reviews; CRD42024502884) and was conducted and reported in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines (checklist provided in Multimedia Appendix 1). In the course of the study, to clearly determine the effectiveness of electronic health interventions directly applied to patients and to ensure the accuracy and referability of the combined effect sizes after the meta-analysis, health care workers were excluded from the inclusion criteria, and only patient-related electronic health interventions were included.

We included peer-reviewed articles published in English that met the eligibility criteria based on the PICOS (Population, Intervention, Comparison, Outcome, and Study Design) strategy (Textbox 1).

Only English-language publications were considered. When full-text papers were not available, the study authors were contacted to obtain the articles. If the eligibility requirements were met and data on cervical cancer screening were available, the abstract was included in the event that the authors did not reply. In addition, we excluded studies that focused on using electronic health to improve cancer screening rates across multiple types when cervical cancer screening data could not be isolated.

A total of 8 electronic databases (from inception to present) were searched on December 29, 2023: MEDLINE, Cochrane Central Registry of Controlled Trials, Embase, PsycINFO, PubMed, Scopus, Web of Science, and the Cumulative Index to Nursing and Allied Health Literature. Authors LXX and NLZ conducted the search by combining terms indicative of cervical cancer screening (eg, cervical cancer, screen*, human papillomavirus DNA tests, and Papanicolaou test) with terms indicative of electronic health (eg, mobile phone, SMS, and videos). In addition, we checked the reference lists of the retrieved articles and previous meta-analyses and reviews for additional studies. The search queries used in this review are detailed in the Multimedia Appendix 2. Studies published between 2013 and 2022 were included.

We exported the search results into a citation management system. After duplicates were eliminated, LXX (all articles), NLZ, and FWQ (each evaluated half) independently reviewed the titles and abstracts against the inclusion criteria. For full-text review, abstracts containing ambiguous information were included. LXX (18 articles), NLZ (17 articles), JCY (16 articles), FWQ (17 articles), and GLN (21 articles) reviewed full-text articles. LXX reviewed each article to ensure it was appropriately included or excluded.

To extract the data, a prespecified electronic data extraction table was created using the Cochrane Collaboration’s Risk of Bias (RoB) tool and the PRISMA guidelines. The extracted data included the study details (author, year, country of origin, study design, application model, and sample size); participant details (medical history and demographics); duration of the intervention and follow-up; details of the intervention and control groups; primary outcomes; and Cochrane rule of reference measures. Authors LXX (8 articles) and NLZ (6 articles) extracted data from full text publications. LXX verified the accuracy of all the data.

The RoB in the included studies was evaluated using the Cochrane Collaboration’s RoB tool [21]. The domains assessed were selection bias (sequence generation and allocation concealment), reporting bias (selective outcome reporting), attrition bias (incomplete outcome data), and performance or detection bias (blinding of participants, staff, and outcome assessors). For the criteria of low, unclear, and high RoB, the Cochrane Handbook for Systematic Reviews of Interventions was consulted for both within-trial and across-trial assessments. Independent reviews of RoB were conducted by authors LXX (8 publications) and NLZ (6 papers). LXX reviewed each RoB evaluation to ensure accuracy. Publication bias was assessed by visually inspecting a funnel plot for the main outcome and conducting the Egger test [22]. To estimate the pooled effect while accounting for missing studies, the Duval and Tweedie trim-and-fill analysis was used [23].

A descriptive presentation of the study, participant, and intervention features, as well as the RoB assessments, is provided. The relevant outcomes are presented in terms of absolute and relative values, and a random-effects model was used to aggregate relative risk (RR) by applying the Mantel-Haenszel technique. The study used data from the longest follow-up period if outcomes were evaluated at different time points. In cases where studies included multiple intervention arms, the meta-analysis included only the most complex intervention, defined as having the most components. The I² statistic was used to compute statistical heterogeneity, with a cutoff value of ≥75% considered significant. The Cochran Q test (P<.1) was used to conduct a formal test of homogeneity, which is considered statistically significant [24]. Subgroup analysis was performed according to subjects (women with and without cervical cancer screening experience, women with a long absence from screening, and women with HPV), intervention type (phone call, SMS, multimode, web video, and internet-based booking), and economic level (high-income and middle- and low-income). This study used 2 types of sensitivity analyses: the leave-one-out method and the removal of highly biased studies. For statistical significance, a 2-tailed P<.05 was used. Meta-analyses were conducted using Stata (version 16.0, StataCorp LLC).

After removing duplicates, the screening identified 14 articles (comprising 15 studies), [25-38], of which 11 [25-27,29-32,34-37] contained ITT results, and 12 [25-33,35,36,38] contained PP results (Figure 1), representing 23,102 unique patients (Table S1 in Multimedia Appendix 3). A total of 7 studies [27,28,31-33,35,36] had high RoB.

The included studies were conducted in America, Europe, Asia, and Africa, and they were published between 2013 and 2022. Single-intervention studies made up the majority of the studies (10/14, 71%) [25-27,29,30,32-35,38] while multimode intervention studies accounted for 29% (4/14) [28,31,36,37]. The interventions were conducted by university-based research teams, governmental or commercial screening initiatives, or health care facilities or departments. The intervention strategies used in these studies included SMS in 4 studies [27,30,32,35], multimode interventions in 4 studies [28,31,36,37], phone call interventions in 2 studies [25,26], web videos in 3 studies [29,33,38], and internet-based booking in 1 study [34]. Outcomes were measured at several time points, including 2, 3, 4, 5, 6, 7, 12, and 14 months. The interventions in most studies (8/14) were based on theoretical models, including the Transtheoretical Model [26-30,32,36,38], theory of planned behavior [27], Health Belief Model [28-30], MINDSPACE (Messenger, Incentives, Norms, Defaults, Salience, Priming, Affect, Commitments, and Ego) framework [32], and social cognitive theory [29,36,38] (Table S1 in Multimedia Appendix 3).

There was a wide variation among the study participants. For example, in some studies, participants were targeted based on geographical region or by their profession as teachers [26,33], emergency patients [27], or young women [32,34]. Some of the participants were women from the general population (regardless of their participation in cervical cancer screening), some were women who had not participated in screening for a long time or had never participated in screening [26,27,37,38], and some were women with HPV [28,35]. Supplementary information regarding the interventions in the experimental and control groups is shown in Table S1 in Multimedia Appendix 3.

The RoB evaluations for the research included in the analysis are presented in Figures 2 [25-38] and 3. In summary, 50% (7/14) [27,28,31-33,35,36] of the randomized controlled trials (RCTs) included in the study were categorized as high risk, and the remaining RCTs (7/14) [26,27,29,30,33,34,37] were considered as having some concerns. Only 2 studies (2/14, 14%) [25,38] were classified as low risk. Studies were categorized as high risk due to issues in multiple areas, such as participant and staff blinding, outcome assessment blinding, and assignment effect blinding. Upon visual examination of the funnel plot (Figure 4), no indications of publication bias were observed. The effect sizes exhibited a relatively symmetrical distribution. Most effect sizes fell within the funnel plot, and those outside it were symmetrically distributed. Egger's regression intercept test yielded a significant result (Intercept=2.5, 95% CI 0.32-4.45; P=.03), indicating that the observed results might be influenced by publication bias. The trim and fill approach developed by Duval and Tweedie was used to evaluate the presence of publication bias. After 2 iterations of the analysis, no additional trials were deemed necessary, and the combined effect size remained consistent. After accounting for potential publication bias, the effect of the intervention remained statistically significant (P<.001). Although the observed results might have been influenced by publication bias, there was no indication of a small-study effect (Egger test: t=2.5; P=.03; 95% CI 0.32-4.45).

The cervical cancer screening rate was the primary outcome in 12 out of 14 articles (86%) [26-32,34-38]. In total, 10 studies [25,26,28-32,36-38] found that the intervention significantly increased cervical cancer screening rates. The intervention had an absolute impact of 34% (3125/9191) in the intervention arms. Between the intervention and comparison arms, there was an absolute risk difference of 11.9% (95% CI, 7.4-16.4), with 27% (2714/9940) screened in the comparison arm. Among the included RCTs, the overall pooled RR for participation in cancer screening was 1.464 (95% CI 1.285-1.667; Figure 5 [25-38]), meaning that receiving an electronic health intervention raised the odds of screening by 46%. Nonetheless, significant heterogeneity (I²=84%) was noted.

Similar effect estimates were obtained after stratification according to the intervention method. The order of magnitude of the effects of the various intervention modalities was as follows: phone call (RR 1.82, 95% CI 1.40-2.38; I²=0%), multimode (RR 1.62, 95% CI 1.26-2.08; I²=85%), SMS (RR 1.41, 95% CI 1.14-1.73; I²=71%), and web videos and internet-based booking (RR 1.25, 95% CI 1.03-1.51; I²=67%; Figure S3 in Multimedia Appendix 3). Subgroup analysis based on the intervention group revealed no significant difference in cervical cancer screening rates between women with HPV in the electronic health interventions group and the nonelectronic health interventions group (RR 1.17, 95% CI 0.95-1.45; I²=57%). Women with and those without cervical cancer screening experience (RR 1.56, 95% CI 1.32-1.90; I²=90%) and women with a long absence from screening (RR 1.41, 95% CI 1.08-1.85; I²=37%) showed a difference between the electronic health interventions group and the control group (Figure S4 in Multimedia Appendix 3). Subgroup analysis according to economic status showed a significant difference for the middle- and low-income group between the electronic health intervention group and the control group (RR 1.51, 95% CI 1.27-1.79; I²=68%). A significant difference was observed between the intervention and control groups in high-income countries. The RR was 1.41, with a 95% CI of 1.15-1.73, and the heterogeneity (I²) was 91%. Refer to Figure S5 in Multimedia Appendix 3 for further details. The sensitivity analysis demonstrated that the combined RR and I² value were consistent (RR 1.46, 95% CI 1.29-1.66; refer to Figure S6 in Multimedia Appendix 3). After removing studies with a high risk of bias, a meta-analysis (n=7) compared the intervention and control groups at the end of the intervention. The analysis showed an RR of 1.56 (95% CI 1.21-2.02) for cervical cancer screening (refer to Figure S7 in Multimedia Appendix 3).

A total of 11 studies [25-27,29-32,34-37] reported the results of ITT analysis of cervical cancer screening rates. The results indicated that electronic health can improve the cervical cancer screening rate of participants more effectively than that of the control group (RR 1.382, 95% CI 1.214-1.574; P<.001; I²=82.0%; Figure 6 [25-27,29-32,34-37]). The percentage of people screened in the comparison groups was 26.9% (2636/9795), while the absolute effect of screening in the intervention groups was 31.7% (3347/10558). The sensitivity analysis of the leave-one-out method, demonstrated that excluding any single study from the random-effects model did not result in a loss of significance (Figure S8 in Multimedia Appendix 3). The results of sensitivity analysis after excluding studies with high RoB show that the overall pooled RR were robust (RR 1.67, 95% CI 1.22-2.28; Figure S9 in Multimedia Appendix 3).

The findings of the PP analysis of cervical cancer screening rates were published in 12 publications [25-33,35,36,38]. According to the findings, the total pooled effect estimate for PP of cervical cancer screening was 1.565 (95% CI 1.381-1.772, I²=74%; Figure 7 [25-33,35,36,38]). The intervention and comparison groups had an absolute risk difference of 15% (95% CI 10.2-19.8). Comparatively, the intervention group’s cancer screening rate was 44% (1549/3540), while the control group’s rate was 29% (1474/5132). The sensitivity analysis of the leave-one-out method, demonstrated that excluding any single study from the random-effects model did not result in a loss of significance (Figure S10 in Multimedia Appendix 3). After removing studies with a high RoB (n=5), a sensitivity analysis revealed a pooled RR value of 1.93 (95% CI 1.64-2.27; Figure S11 in Multimedia Appendix 3).

This meta-analysis evaluated the effectiveness of electronic health interventions in increasing cervical cancer screening rates. We found that electronic health interventions can significantly improve cervical cancer screening rates among women. The ITT and PP analyses also showed similar results.

Our systematic review included 14 papers (15 studies) that examined the impact of electronic health interventions on enhancing participation in cervical cancer screening. The overall effect size for cervical cancer screening rates significantly favored the electronic health interventions group (RR 1.464, 95% CI 1.285-1.667). In addition, the subgroup analysis revealed no discernible impact of electronic health interventions on increasing cervical cancer screening rates among women who tested positive for HPV. For individuals who qualified for a PP analysis, the effect size on cervical cancer screening rates was the largest (RR 1.565, 95% CI 1.381-1.772). The sensitivity analyses demonstrated the stability of the result estimates, indicating that electronic health interventions are effective in enhancing cervical cancer screening rates.

Three studies similar to the present study were published before 2021. Uy et al [39] and Ruco et al [12] evaluated the effectiveness of electronic health-related interventions for improving cancer screening rates. These studies included different types of cancer. Uy et al [39] focused on SMS interventions, and the results showed that SMS can improve the screening rates for cervical cancer and breast cancer. The main intervention studied by Ruco et al [12] was social media, and the results showed that mobile health technology, especially social media, can effectively improve cancer screening rates. However, the study by Ruco et al [12] included controlled clinical trials, whereas the present analysis included only RCTs, which may provide a more accurate pooled effect size. Tamuzi et al [40] looked at mobile health interventions for cervical cancer screening and found 17 papers in total, which were then categorized by type of intervention into a meta-analysis. In this study, the intervention is electronic health, which includes SMS, phone calls, multimode interventions, web videos, and internet-based booking. In their study, the intervention was evaluated in the context of mobile health, which included phone calls, letters, and text alerts. According to Tamuzi et al [40], the only intervention with statistically significant estimates of aggregate impacts was phone alerts. However, a meta-analysis of these treatments could not be carried out since only 1 article in their review discussed the effectiveness of SMS alerts. According to the current analysis, 4 out of the 14 publications included discussed how text messaging increased cervical cancer screening rates. The results of this study also suggest that text messaging is effective for improving cervical cancer screening rates (RR 1.41, 95% CI 1.14-1.73). In addition, his study was conducted in 2017, and most of the articles included in this study were published after 2017 (10/14). Our review offers a thorough and updated overview of this subject. In addition, this study analyzes the effect size of electronic health interventions on cervical cancer screening rates using PP and ITT analyses to clarify their effectiveness and extensibility.

The popularity of electronic health is increasing in parallel with technological advances [41]. In a survey of US adults, over 80% of individuals aged 18-49 years old and 73% of those aged 50-64 years old use social media sites [42]. Electronic health interventions include various methods, such as SMS, phone calls, and web videos [43]. The results of this study show that, regardless of the type of intervention, electronic health can improve cervical cancer screening rates. The results of this study show that the aggregate effect size was statistically significant for both ITT and PP analyses, which strongly supports the effectiveness of electronic health in improving cervical cancer screening rates.

When the studies were grouped according to the intervention method, the results showed that each of the intervention modalities (including phone calls, SMS, multimode, web videos, and internet-based booking) could increase the cervical cancer screening rate, but the combined effect size was highest when phone calls were used as the means of intervention (RR 1.82, 95% CI 1.40-2.38; I²=0%) [27,33]. Phone calls are a cost-effective intervention [44], because direct contact with the recipient allows timely responses to questions, which further validates the importance of direct communication [45]. However, phone calls intervention is associated with higher manpower costs. The development of voice interaction can save manpower costs while retaining some of the advantages of telephone intervention [46]. Further studies are necessary to understand the impact of voice interaction on cervical cancer screening rates.

When the studies were grouped according to economic level, the subgroup analysis results showed that the aggregate effect size of electronic health on cervical cancer screening rates was higher in low- and middle-income areas than in high-income areas, suggesting that electronic health is more effective in low- and middle-income populations. One possible reason for this result is that resources are scarce in low- and middle-income areas. Electronic health interventions can help the residents of low- and middle-income areas improve their awareness and attention to cervical cancer through internet-based health education, health knowledge dissemination, internet-based reminder services, and other forms, which may increase the motivation for screening [20,47]. In high-income areas, electronic health applications are more common, and electronic health incentives are therefore weaker for the high-income population than for the low-income population [48].

Subgroup analysis based on distinct intervention populations revealed no statistically significant differences in cervical cancer screening rates among women with HPV between the electronic health intervention group and the conventional intervention group. Most of the studies included in the analysis were based on the theoretical health belief model [28,30], and the main strategy was to provide patients with health education related to cervical cancer screening in the form of electronic health [26-28,35-37]. Overall, the screening rate in women with HPV (45%) was higher than the average rate (30.52%), possibly because women with HPV are more aware of their risk of cervical cancer as carriers of high-risk HPV [49,50]. Factors such as physician recommendations, secondary tests, and education also play a role in promoting their participation in cervical cancer screening, resulting in a relatively high screening rate. Because the main limitation to follow-up screening in women with HPV is not cervical cancer awareness, electronic health is not effective in improving screening rates in this population [51]. However, advances in science and technology are increasing the diversity of electronic health methods available, and the intervention mechanism is not limited to reminder education [52]. Further studies are needed to explore the effect of electronic health on cervical cancer screening rates in women with HPV.

The Egger’s test results of this study suggest potential publication bias; however, after using the trim and fill method, no additional studies were deemed necessary, and the combined effect size and P values remained unchanged, thereby affirming the robustness of these findings. The RoB results indicated that 50% of the studies exhibited high bias risk, prompting a sensitivity analysis excluding these high-risk studies, which enhanced the precision and reliability of the outcomes. After excluding studies with high bias risk, the pooled results were as follows: cervical cancer screening: 1.56 (95% CI 1.21-2.02), cervical cancer screening (ITT): 1.67 (95% CI 1.22-2.28), cervical cancer screening (PP): 1.93 (95% CI 1.64-2.27). Consequently, although studies with high bias risk may reduce some intervention effects, the overall effect size exhibited minimal variation, thereby confirming the robustness of the study results.

Despite the consistency of the pooled effect estimate in the subgroup and sensitivity analyses, significant heterogeneity remained in this meta-analysis. The variation in demographics, treatments, or outcome measures among the studies may be the cause of this. For instance, all adults aged 65 years or younger or highly specialized populations like ER patients or individuals with HPV were among the populations randomized in the studies included in this review. Furthermore, we only included papers published in English, which could have an impact on the meta-analysis’s findings. Despite our best efforts, we were unable to locate some of the records that the search identified. Finally, because it was unable to identify the outcomes unique to cervical cancer, RCTs that enrolled patients with a variety of malignancies and reported their combined findings were excluded. Furthermore, non-RCT designs (such as adaptive trials) were disregarded.

In conclusion, the results of this study suggest that electronic health interventions may significantly impact increasing cervical cancer screening rates. However, the study exhibited a high degree of heterogeneity. Phone call–based electronic health interventions demonstrated the most substantial improvements in increasing cervical cancer screening rates, with low heterogeneity. Multimodal interventions constitute a growing segment of electronic health treatments. The RoB review highlighted variations in blinding processes, and only a few studies disclosed plans to upscale, assess the cost of the intervention, or maintain the outcomes. Future cost-effectiveness analyses should be conducted to further elucidate the scalability of electronic health interventions.