Assessing the Effectiveness of mHealth Interventions for Diabetes and Hypertension Management in Africa: Systematic Review and Meta-Analysis

Introduction

Managing chronic diseases often calls for a long-term care strategy [1]. Diabetes and hypertension remain two of the most common chronic conditions globally, resulting in the highest health care resource use and mortality [2-4]. Type 2 diabetes prevalence has become a substantial health issue, especially in African regions where type 2 diabetes is predicted to increase at the quickest rate (129%) in the world by 2045 [5-7]. Similarly, hypertension remains a major public health challenge among older adults in the African region, with an estimated pooled prevalence of 30.8% in Africa and between 30% and 31.1% in sub-Saharan Africa [8]. Poor blood pressure control among persons with hypertension is thought to involve intricate interactions between patients, health care providers, and socioeconomic variables [9]. Medication adherence has also been identified as one of the critical disease management issues, especially in enhancing life quality, health outcomes, and access to affordable health care worldwide [10,11].

Disease management programs using mobile health (mHealth) are promising emerging strategies to help patients self-manage their conditions (eg, measuring their blood pressure and sugar levels with remote professional support when needed [7,12]). mHealth is a medical and public health practice supported by portable electronic devices such as cell phones, patient monitoring devices, personal digital assistants, and other wireless gadgets [13]. This includes the use of phones and remote monitoring devices in health care and public health practice for communication, data collection, patient monitoring, and education, and to facilitate adherence to chronic disease management [14,15]. mHealth devices can improve service delivery and impact patient outcomes [15].

Previous studies in some low- and middle-income countries have assessed the application of mHealth as a tool to increase drug compliance in patients with a range of long-term illnesses, including diabetes, chronic obstructive pulmonary disease, and HIV infection [11,16]. Although several African countries are still in the pilot and development stages, an increasing number of mHealth apps have been put into use in clinical care settings [17]. The majority of small-scale pilot or feasibility mHealth intervention studies in Africa have been based on SMS text messaging systems to improve disease management [17,18].

Most individuals now possess mobile phones, and there are over 5.3 billion subscribers to mobile services worldwide, 67% of the world’s population [19,20]. There will be 400 million more new mobile service customers between now and 2025, the majority of whom will come from Asia Pacific and sub-Saharan Africa, increasing the total number of subscribers to 5.7 billion (70% of the global population) [21]. There have been individual studies in Africa about mHealth interventions on disease management [22], although the data on the efficacy of mHealth in the management of diabetes and hypertension in Africa are limited and have not yet been systematically evaluated. Therefore, this systematic review assessed the effectiveness of mHealth interventions on blood pressure control among patients with hypertension and glycemic control among patients with diabetes in Africa. The findings of this paper will guide improvements to the adoption of mHealth for the management of diabetes and hypertension in African countries.

Methods

This systematic review was conducted following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [23] (Multimedia Appendix 1). The protocol was registered in PROSPERO (CRD42021230642).

Search Strategy

PubMed, Cochrane Library, Google Scholar, African Journals Online, and Web of Science were searched for relevant studies published from inception to July 2022, with assistance from a clinical librarian. The full search strategies, common Medical Subject Headings (MeSH), and search terms used across databases are available in Table S1 in Multimedia Appendix 2. The reference lists of the included studies were hand-searched to identify additional relevant studies.

Study Selection

Two independent authors (PA and KA) manually assessed and screened studies for both the titles and abstracts as well as full-text articles using an Excel sheet (Microsoft Corporation). Disagreements were resolved by consensus with a third author (CA) as necessary. This was performed in three stages as follows. First, PA screened the titles of all papers to determine their relevance. KA performed a cross-check by screening 20% of the titles excluded by the first reviewer, and it was confirmed that none of the titles screened by the second reviewer met the inclusion criteria. Second, abstracts of the papers selected after the title screening stage were again screened by PA and KA following the same procedure as described in step one. Finally, the full texts of potentially relevant papers were retrieved and evaluated by PA and KA independently to ascertain their relevance and usefulness to the review. Disagreements were settled through dialogue with CA to reach an agreement. Duplicates were also identified using EndNote reference manager (version x9; Clarivate).

Inclusion Criteria

We included studies that met the following criteria: the patients had hypertension or diabetes and were 18 years or older; the patients had received treatment at a selected health care setting; the intervention included an mHealth component; the results included target values of hemoglobin A_1c (HbA_1c), systolic blood pressure, or diastolic blood pressure; the studies were randomized controlled trials (RCTs); the articles were written in English; and the studies were conducted in hospitals and primary health centers.

Exclusion Criteria

We excluded studies in which the full text was not available after attempts to contact the author, the research participants were pregnant women or a specific patient population (eg, patients with cancer), the results did not describe primary outcomes, the primary intervention did not use mHealth devices, or the articles were unpublished manuscripts or conference abstracts.

Risk of Bias Assessment

The quality of each study was assessed using a 28-point scoring system as adopted from the Downs and Black checklist [24]. The included studies focused on the following items for assessment: items 1 through 10 evaluated whether the information provided was adequate for the reader to make an objective assessment of the study’s findings; items 11 through 13 evaluated external validity, which examined the extent to which study findings could be applied to the population from which the study participants were drawn; items 14 through 20 assessed possible bias, which focused on biases in the assessment of the intervention and the result; and items 21 through 26 assessed confounding, which focused on biases in the research participant selection. To determine if neutral research results may be the result of chance or insufficient power, item 27 evaluated the study’s power (Table 1).

Data Extraction

Two authors (PA and KA) independently extracted the following study characteristics from each included article using a tested extraction form: first author, year of publication, mean age, the country where the study was conducted, the participant (patient with diabetes or hypertension), mHealth location (primary care setting, hospital, clinic, etc), condition (diabetes/hypertension), sample size, mHealth intervention, study design, and outcome of the intervention.

Data Synthesis and Analysis

The data for the primary outcomes (HbA_1c, systolic blood pressure, and diastolic blood pressure) were analyzed separately using random or fixed effects models with a weighted mean difference (WMD) in ReviewManager (version 5.4; The Cochrane Collaboration) [32]. The I² statistic was calculated to measure the percentage of variation across trials due to heterogeneity, with values of <50% and ≥50% indicating low and high levels of heterogeneity, respectively. The WMD for blood pressure and HbA_1c between the intervention and control and Z tests were used to compare groups, and P<.05 was regarded as statistically significant. We checked publication bias subjectively by funnel plots and objectively by Begg and Egger tests using Stata version 16 (StataCorp). Begg and Egger tests with P<.05 were considered to have significant publication bias.

Results

Study Selection

We identified 2908 records from our search; authors screened 2880 titles and abstracts after duplicates were removed. In total, 7 studies (RCTs) [25-29,31] were considered eligible for inclusion. The reasons for excluding a study are provided in Figure 1.

‎

Study Characteristics

The characteristics of the studies are shown in Tables 2 and 3. Of the 2249 participants, 1824 (54.1%) were female, the mean age ranged from 51.2 to 60.6 years, and the sample sizes ranged from 60 to 1372 participants. The studies were conducted at hospitals and primary health centers in the following countries: Ghana (n=3), South Africa (n=2), Egypt (n=1), and Cameroon (n=1). Five studies [26-29,31] reported the apps’ effectiveness in controlling blood pressure among patients with hypertension (Table 4), while 2 studies [25,27] reported the effect on HbA_1c levels in patients with diabetes (Table 5). Of the 7 included studies, 2 articles had more than one intervention group with the same outcome measure [25,28].

Meta-Analysis of the Effects on Primary Outcomes

A total of 7 studies, 5 for blood pressure [26-29,31] and 2 for HbA_1c [25,27], were included in the meta-analysis.

Systolic Blood Pressure

As shown in Figure 2, one study had more than one intervention group with the same outcome measured [28]; therefore, 7 interventions are shown in the forest plot of systolic blood pressure, and the estimated WMD of systolic blood pressure between intervention and control groups was not statistically significant at −1.39 mm Hg (95% CI −4.46 to 1.68; P=.37; I²=61%). No significant publication bias was detected visually by the funnel plot (Figure 3) or statistically by Begg (P=.30) and Egger (P=.10) tests.

‎

‎

Diastolic Blood Pressure

There was no statistically significant difference in diastolic blood pressure (0.36 mm Hg, 95% CI −1.37 to 2.08; P=.69; I²=47%) between the intervention and control groups (Figure 4). No significant publication bias was detected visually by the funnel plot (Figure 5) or statistically by Begg (P=.65) and Egger (P=.81) tests.

‎

‎

Hemoglobin A_1c

Additionally, as shown in Figure 6, one study had more than one intervention group with the same outcome measured [25]; hence, 3 interventions were shown in the forest plot of HbA_1c, and the mHealth intervention had no significant lowering effects on HbA_1c levels among patients with diabetes in the pooled meta-analysis at 0.20 mmol/mol (95% CI −0.40 to 0.80; P=.51; I²=0%). No significant publication bias was detected visually by the funnel plot (Figure 7) or statistically by Begg (P=.96) and Egger (P=.10) tests.

‎

‎

Sensitivity Analyses

Sensitivity analyses were conducted using the leave-one-out method. For systolic blood pressure, the Kingue et al [29] study had an impact on the WMD, with the pooled WMD being statistically significant after the exclusion of the Kingue et al [29] study (−2.22, 95% CI −3.94 to −0.60; P=.01; Figure 8). For diastolic blood pressure and HbA_1c, the exclusion of each of the studies rendered the WMD nonsignificant (Figures 9 and 10).

‎

‎

‎

Discussion

Discussion of the Key Findings

This systematic review and meta-analysis identified 7 RCT studies that assessed the effectiveness of mHealth interventions on blood pressure and glycemic control among patients with hypertension and diabetes in Africa. In this review, the effectiveness of mHealth interventions on blood pressure and glycemic control among patients with hypertension and diabetes in Africa did not show conclusive evidence.

Systolic Blood Pressure Control

For systolic blood pressure, we observed a reduction after the mHealth intervention compared to usual care by an average of 1.39 mm Hg; however, it was not statistically significant. After conducting a leave-one-out analysis, a study by Kingue et al [29] had an impact on the WMD, and the exclusion of this study resulted in a pooled WMD reduction of 2.22 mm Hg in systolic blood pressure. This finding is consistent with previous RCT studies [20,34-36] that examined mHealth interventions on systolic blood pressure control and showed that mHealth interventions reduced systolic blood pressure by 10.4 mm Hg [20], 5.5 mm Hg [34], 3 mm Hg [35], and 3.9 mm Hg [36], respectively. In contrast, a study performed by Rubinstein et al [13] reported that the mHealth intervention did not reduce systolic blood pressure compared with usual care. This discrepancy could be explained by the relatively small sample number of studies included in this review. Another reason could be due to the different study populations, interventions, ages, and medications used.

Diastolic Blood Pressure Control

For diastolic blood pressure, our study observed no lowering effect of the mHealth intervention, which reduced by an average of 0.36 mm Hg, which is inconsistent with studies performed by Lu et al [36] and Zhang et al [20] who reported a reduction of 2.2 and 4.8 mm Hg, respectively, after the mHealth intervention compared to usual care. The disparity is that the previous studies [20,36] were conducted among patients with stroke and heart failure with a more complicated pathogenesis of hypertension, which might have resulted in the observed significant decrease in diastolic blood pressure control in this study. In patients with stroke, lower blood pressure might be achieved with strict treatment targets that also lead to a controlled condition. In previous studies, they noted a significant net reduction in body weight and intake of high-fat and high-sugar foods after the intervention [20,36]. Despite no significant findings on diastolic blood pressure control after mHealth interventions, the study by Rubinstein et al [13] reported that each 1 mm Hg decrease in diastolic blood pressure is associated with a 7% decrease in mortality from stroke and ischemic heart disease. Thus, the mHealth intervention may still be a measure worth considering for reducing blood pressure.

Glycemic Control

For glycemic control, the meta-analysis results showed no improvement after the mHealth interventions. Our study contradicts previous studies by Mao et al [35], Moattari et al [37], Kitsiou et al [4], and Huang et al [38] who found significant improvements in glycemic control following mHealth interventions among patients with diabetes. These studies have reported that patients and health care professionals who communicated by SMS text messages, telephone calls, and even electronic reminders or web servers reported greater improvement in HbA_1c outcomes compared with usual care [4,35,37,38]. Thus, patients with poorly controlled diabetes might benefit more from using mHealth, therefore more clinical trials are needed to confirm these findings. The Adjei and Marschollek [25] study in Ghana, despite not reporting HbA_1c, saw a substantial reduction in fasting plasma glucose of −1.6 mmol/L. Although evidence is scarce about the effect of mHealth interventions on the management of patients with diabetes, the difference could be attributed to the better care the system generates from the health providers. Another reason is that long-term interventions likely result in more significant changes in glycemic control than short-term mHealth interventions.

Strengths and Limitations

To the best of our knowledge, this review was the first that assessed the effectiveness of mHealth interventions in diabetes and hypertension management in Africa. Quality appraisal suggests that the quality of the included studies was good. Additionally, the included studies show no publication bias. However, there are limitations to acknowledge. Despite a thorough search, the number of included studies was relatively small, signifying that using mHealth interventions in Africa on patients with hypertension and diabetes remains an emerging area. This review may not be able to capture some significant effects due to the small samples in the included studies. Given the above limitations, future studies with larger samples are needed to validate our findings.

Conclusion

Our study showed no conclusive evidence on the effect of mHealth interventions on systolic blood pressure, diastolic blood pressure, or glycemic control. However, the sample sizes of the included studies were small; therefore, there is a need for larger RCT studies to confirm these findings.