Effect of Telemedicine on Quality of Care in Patients with Coexisting Hypertension and Diabetes: A Systematic Review and Meta-Analysis

Introduction

Hypertension and diabetes are two major chronic diseases collectively responsible for almost 70% of all deaths.¹ According to the World Health Organization (WHO), one in four men and one in five women had hypertension in 2015.² Worldwide, the prevalence of diabetes in 2019 is 10.4%, which is projected to be 11.9% by 2045.³ Although hypertension and diabetes can be independently diagnosed, they are each complex. These conditions have shared aspects of pathophysiology (e.g., obesity and insulin resistance), resulting in frequent coexistence in the same individual.⁴ The prevalence of comorbid hypertension and diabetes has ranged from 4.5% in the general population⁵ and 17% in older adults.⁶ This coexistence is associated with an increased risk of life-threatening cardiovascular complications.^7,8

The coexistence of hypertension and diabetes has been a great challenge to the health care system across the globe. Traditional hospital-centered and treatment-based health care systems can no longer meet the needs of people's health and the systems' sustainability. A person-centered integrated delivery system is much needed and is a central component of quality of care (QoC).⁹ With the development of technology and internet connection, telemedicine has been endorsed to improve access, efficiency, and QoC by bridging the gap between professional health care and patient self-management.¹⁰ WHO defined telemedicine as “the delivery of health care services, where distance is a critical factor, by all health care professionals using information and communication technologies for the exchange of valid information for the diagnosis, treatment, and prevention of diseases and injuries, research and evaluation, and the continuing education of health care providers, all in the interests of advancing the health of individuals and their communities.”¹¹ Due to the need for individualized and continuous support for patients with chronic conditions, telemedicine has been widely used in these populations during the past decade.^12,13

Studies have examined the impact of using telemedicine on QoC indicators in people with hypertension, including systematic reviews. However, findings have been inconsistent. Two reviews reported that telemedicine effectively reduced blood pressure (BP),^14,15 but other reviews did not report consistent findings on BP.^16,17 In addition, telemedicine may improve self-care behaviors such as medication adherence among people with hypertension.^18,19 Despite the potential benefits of telemedicine, gaps remain in our knowledge about the effect of using telemedicine in people with hypertension and comorbid diabetes, who face more challenges in disease management than those with hypertension alone. For the management of hypertension and diabetes, each involves complex self-care behaviors. The two conditions also have shared components of self-care. People with hypertension and comorbid diabetes likely benefit from aggressive treatment that targets modifiable risk factors (e.g., control of BP and glucose, physical activity, and diet).^20,21 Simultaneous management of these factors might preserve the renal function of people with hypertension and comorbid diabetes, and thus slow the progress of diseases.²² More importantly, previous studies failed to address the complexities involved in understanding QoC, which may hinder our abilities to scale-up priority health interventions. QoC is a complex construct, especially in people with multimorbidity.²³ WHO defined QoC as “the extent to which health care services provided to individuals and patient populations improve desired health outcomes.”²⁴ The classic Donabedian QoC framework makes the distinction between structure, process, and outcome.²⁵ Recent work suggests that the assessment of QoC should be person centered and take an individual's personal experience into account.²⁶ Thus, QoC with telemedicine needs to go beyond the outcome of care (e.g., BP).

Based on the above evidence, there is a need to examine the use of telemedicine in people with hypertension and comorbid diabetes. The aim of this systematic review was to summarize the telemedicine programs and identify its effect on QoC indicators in this population. Findings from this review may contribute to a more effective management of chronic diseases and thereby facilitate the improvement of patient-centered health care systems. This review will also provide important information for health care providers, policymakers, and health commissioners.

Methods

This systematic review was conducted based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, and the PRISMA protocol was developed to guide the review.^27,28

ELIGIBILITY CRITERIA

The inclusion and exclusion criteria were oriented by the PICO framework (P: patient, I: intervention; C: comparison; O: outcome).²⁹ Clinical trials that investigated the effect of telemedicine in people with hypertension were screened for eligibility. The inclusion criteria were as follows: (1) patients 18 years of age or older, (2) patients with a concurrent diagnosis of hypertension and diabetes, (3) the intervention used telemedicine as defined by the WHO,¹¹ featuring interactive wireless communication and operating web-based applications, and (4) the comparison group received usual or standard care. The exclusion criteria were as follows: (1) studies included inappropriate populations, including pregnant women, patients not having comorbid diabetes, or patients with a current diagnosis of other diseases (e.g., stroke and psychiatric disorders), (2) the outcomes were QoC indicators, including the commonly used outcome of care (e.g., BP and glycemic control) as well as process of care (e.g., self-efficacy or adherence) and experience of care,²³ studies that did not assess the outcomes of interest were excluded, and (3) other types of reports from which data cannot be extracted, such as conference abstract, comment, editorial, review, protocols, and qualitative studies.

SEARCH STRATEGIES

Major electronic databases were searched from inception to March 2020. The databases included PubMed, CINAHL, Web of Science, and Embase. The search terms were telemedicine, telehealth, ehealth (e-health), mhealth (m-health), digital health, remote care, electronic health, or web-based (internet-based); AND hyperten*. Combinations of these two sets of terms were used for searching. Take PubMed for example. The following logic was used: “hyperten*” in Title AND “telemedicine” in Title/Abstract. The search was limited to English language. The reference lists of eligible studies and relevant reviews were manually searched to identify additional studies.

STUDY SELECTION

The records were managed using Endnote X8 (Clarivate Analytics, Philadelphia, PA). The PRISMA flowchart was followed to select eligible studies.²⁸ Briefly, two reviewers (Author 1 and 2) independently screened the records by reviewing the title and abstract. Full texts of potential studies were retrieved for further review. Based on the inclusion and exclusion criteria, two reviewers (Author 1 and 2) read the full texts and determined the final inclusion. Any discrepancy was resolved by consulting a third reviewer (senior author).

DATA EXTRACTION

Data extraction was completed independently by two reviewers (Author 1 and 2) and disagreements were settled by a third reviewer (senior author). The extracted data included study-related characteristics (e.g., first author, year, country, design/setting, and notes on eligibility criteria) and participant-related characteristics (e.g., age, sex, body mass index [BMI], and race/ethnicity). We also extracted information about the interventions, outcomes, and associated measures, as well as the key findings from each study. The key findings were organized by indicators of QoC, including the outcome, process, and experience of care. Data from the same trial, but published in different articles, were extracted simultaneously to provide comprehensive information. Extracted data were tabulated and organized using tables.

RISK OF BIAS ASSESSMENT

The potential risk of bias related to randomized trials was assessed using the Cochrane revised tool (RoB 2.0).³⁰ This tool evaluates the risk of bias from five domains, including bias arising from the randomization process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in measurement of the outcome, and bias in selection of the reported results. Each domain is rated as low risk, high risk, and some concerns. The study is judged to be at low risk if all domains are scored a low risk of bias. The study is assigned a high risk of bias if high risk of bias is present in at least one domain or the study has some concerns for multiple domains, significantly lowering the confidence in the result. The study is scored as having some concerns if none of the domains has a high risk, but there are some concerns in at least one domain. Two independent reviewers (Author 3 and 4) performed the evaluation and any discrepancies were resolved by a third reviewer (Senior author).

DATA SYNTHESIS

Statistical analyses were performed using Stata 14.0 (StataCorp LP, College Station, Texas). Mean and standard deviation (SD) were extracted. SD was calculated from standard error if not reported, and vice versa. For continuous variables, pooled mean difference with 95% confidence interval (CI) was calculated by the inverse variance approach and presented in a forest plot. For categorical variables, pooled risk ratio (RR) with 95% CI was calculated by the Mantel-Haenszel method and presented in a forest plot. For the pooled analysis, within-group difference with the standard error of the mean (or 95% CI) was used. I² value quantifies heterogeneity between studies.³¹ A fixed-effect model was used if no significant heterogeneity (I² < 50%) was present and a random-effect model was used otherwise.^32,33 Sensitivity analysis was conducted by the leave-one-out approach to test the robustness of the findings. Due to the small number of studies included in this review (n < 10), the funnel plot for publication bias was not performed.³⁴ Statistical significance was set at p < 0.05 (two tailed). Narrative syntheses were performed if data cannot be quantified.

Results

SEARCH RESULTS

The process of study selection is shown in Figure 1. Searching of the four electronic databases resulted in 1,330 records. After manual and automatic removal of duplications through Endnote “find duplicates,” a total of 798 records were included for initial screening. Two reviewers screened the title and abstract and 114 studies were retrieved for full texts. Among these studies, 109 were excluded due to the reasons presented in Figure 1. Finally, five studies^35–39 from four clinical trials were included for the final review.

STUDY CHARACTERISTICS

Table 1 shows the study characteristics. Two studies^38,39 were from the same clinical trial, and characteristics of these studies were combined and presented together. The studies were conducted between 2007 and 2019 mainly in the United States and Canada. Two studies^36,37 used a single-arm design, and the other two^35,38 used a three-arm randomized design. The patients were recruited from primary care settings, health center, or VA medical center. As having hypertension and comorbid diabetes was one of the inclusion criteria, all the patients had both conditions. In addition, three studies included patients with type 2 diabetes mellitus and uncontrolled hypertension^35–37; one study³⁵ excluded patients using insulin. The patients were typically middle-aged and older adults, with a mean age of 56.9 (± 7.6) years³⁶ to 67.9 (± 9.9) years.³⁸ The percentage of women ranged from 1.0% to 61.9% across studies. BMI was not available in one study,³⁶ and the mean was between 31.3 and 33.8 kg/m² in the remaining studies. In one study,³⁸ almost all the patients were Caucasian. Two-thirds were Caucasian in the other two studies^35,37 and most were African American in another.³⁶

INTERVENTIONS AND OUTCOME MEASURES

Table 2 shows an overview of the interventions and outcomes. The sample size varied from 29 to 107 in the control group and 31 to 102 in the intervention group. The duration of the intervention ranged from 3 to 6 months. Two studies used a three-arm design and there were two intervention groups with a difference in intensity of the interventions. In one study,³⁵ the two intervention groups received the same intervention content, but with different duration (1 month vs. 3 month). In the other,³⁸ the two groups received the same duration of intervention, but with different content. The low-intensity group received a subset of interventions delivered to the high-intensity group. The content of the interventions varied between studies. One³⁵ focused exclusively on medication-taking behaviors and the other³⁷ on BP monitoring. One study³⁶ designed tailored interventions to address self-management facilitators and barriers with a focus on medication adherence. The Wakefield et al. study³⁸ was more comprehensive in content, which enabled a more prompt and dynamic interaction between patients and health care providers. Detailed interventions are presented in Table 2.

We also summarized the characteristics of the tele-system used for each study and classified them into five domains of functionality. This classification was based on a previous review.¹⁵ Overall, all studies used prompts or reminders. One study³⁵ using digital medicine did not employ self-monitoring. Other characteristics are shown in Table 3.

Various QoC indicators were assessed. The most commonly examined were indicators for outcome of care, including change in BP or A1C and BP control rate. The process of care indicators included self-efficacy, knowledge, medication adherence, and patient engagement. Two studies also assessed the experience of care such as patient perception of the system. Details about the outcomes are shown in Table 2.

EFFECT OF TELEMEDICINE ON OUTCOMES OF CARE

Table 4 shows the key findings of each study. All studies evaluated the effect of telemedicine on outcomes of care. Three studies^35,37,38 reported significant decreases in systolic blood pressure (SBP) after the intervention compared with the control group. Data from the three studies were pooled and presented in the forest plot (Fig. 2). Based on the plot, there was a significant difference in SBP change between the intervention and control group (z = 7.98, p < 0.001). The mean group difference was −10.4 mm/Hg (95% CI: −13.0 to −7.9). Sensitivity analysis was performed by removing the study using a preintervention and postintervention design.³⁷ The results remained robust (data not shown).

Two studies reported the effect of telemedicine on diastolic blood pressure (DBP). Data from these two studies were pooled and presented in Figure 3. Based on the forest plot, there was a significant difference in DBP change between the intervention and control group (z = 3.97, p < 0.001). The mean group difference was −4.8 mm/Hg (95% CI: −7.1 to −2.4).

Three studies examined the impact of telemedicine on BP control. Specifically, two studies reported a significant increase in achieving the goal of BP control, with a group difference between 47.9%³⁵ and 34.6%.³⁷ One study³⁶ found that there was no significant difference in preproportions and postproportions of patients whose BP was not controlled. Data from the three studies were pooled and presented in the forest plot (Fig. 4). Based on the plot, there was no significant difference in BP control between the intervention and control group (z = 1.62, p = 0.106). The RR for not achieving the goal of BP control was 0.60 between the intervention and control group (95% CI: 0.33–1.11). Sensitivity analysis revealed robustness of the results.

Only two studies^35,38 investigated the effect of telemedicine on glycemic control, with mixed findings. Data from these two studies were not pooled and thus were summarized. In brief, Wakefield et al.³⁸ found that A1C levels in the low- and high-intensity intervention groups decreased significantly at 6 months from baseline: −0.40% (p < 0.001) and −0.44% (p < 0.001), respectively. The group difference in change in A1C was 0.51% between the high-intensity intervention group and the control group (p < 0.05). However, the impact of telemedicine on A1C was not maintained at 12 months. Frias et al.³⁵ reported that the intervention did not decrease A1C significantly when all patients were included in the analysis. Only when the analysis was restricted to those with a baseline A1C ≥ 8%, did the intervention improve glycemic control. Differences in A1C change between the control group and intervention groups ranged from −0.98% (95% CI: −1.72 to −0.24) to −0.57% (95% CI: −1.53 to 0.39).

EFFECT OF TELEMEDICINE ON PROCESS OF CARE

The process of care included medication adherence, BP monitoring, self-efficacy, and disease knowledge (Table 4). In the two studies that evaluated medication adherence, one³⁵ found that the mean medication ingestion adherence was 86% during the first 1 month and 84% for the entire 3 months in the intervention group. No comparison was made between the intervention and control group. Wakefield et al.³⁹ reported no significant group difference in medication adherence, although it was high at all time points. Logan et al.³⁷ delivered a telemedicine program designed to enhance BP monitoring and found that after the intervention, the average number of BP readings was 12.3/week; the average weekly adherence rate was above expectations (149%). One study assessed the impact of telemedicine on self-efficacy³⁹ and did not find a significant effect. However, that study reported disease knowledge in the high-intensity intervention group was significantly higher than the control group at 6 months, but the difference was not maintained at 12 months.

EFFECT OF TELEMEDICINE ON EXPERIENCE OF CARE

The experience of care in this review included patient perception and engagement (Table 4). In Frias et al.'s study,³⁵ over 90% of the participants agreed that the tele-system was easy to learn and to incorporate into their daily routine; the data viewing and sharing with their provider helped them to manage their health care and disease. Similarly, Logan et al.³⁷ found that 18 of 20 patients believed that the data helped them and their providers to best treat their BP; 17 patients endorsed to continue using the system or use it in the future; Wakefield et al.³⁹ measured perception of technology. Participants in the intervention groups had a mean score of 3.7–3.8 (out of 5) and no group difference was observed. In addition, Lewinski et al.³⁶ reported that around 83% of the patients were engaged in the system.

RISK OF BIAS ASSESSMENT

Quality appraisal of the two randomized trials was performed using the Cochrane risk of bias tool. Both studies were scored some concerns about the risk of bias. Specifically, although the Wakefield et al. study³⁸ had a low bias from the randomization process and in the selection of reported outcomes, it was assigned some concerns in the remaining domains. For the Frias study,³⁵ some concerns were raised on bias from the randomization process and bias due to deviations from intended interventions.

Discussion

This systematic review and meta-analysis aimed to investigate the effect of telemedicine on QoC in patients with hypertension and comorbid diabetes. Building upon previous studies, we developed this review following the PRISMA guidelines and included five studies from four clinical trials. We found that telemedicine improved outcomes of care and particularly decreased BP. Telemedicine may also improve experience of care (e.g., patient perception and engagement); however, its impact on process of care has been inconsistent. These findings provided further evidence for the use of telemedicine in patients with multimorbidity such as hypertension and diabetes.

Telemedicine could improve BP in patients with hypertension and comorbid diabetes. Overall, compared with usual care, telemedicine decreased BP by 10.4/4.8 mm/Hg. This finding is consistent with the one found in people with only hypertension. Specifically, Ma et al.¹⁴ reviewed 13 studies and found a 6.0/3.4 mm/Hg drop in BP by telemedicine. In this review, the effect of telemedicine on SBP was larger than that found in Ma et al.'s study. One possible explanation was that participants included in this review typically had uncontrolled hypertension, whereas Ma et al.'s study¹⁴ had a comparable proportion of patients with controlled and uncontrolled hypertension. For those with relatively well-controlled hypertension, telemedicine may have limited ability to result in significant improvement in BP. Telemedicine is effective in decreasing BP, possibly due to the features it has. In this review, the studies embedded functionalities into the intervention system such as self-monitoring, using prompts and reminders, automatic feedback. These functions enabled communications between patients and health care providers. The health care team could access patients' data and provide reminders, feedback, or relevant educations in a timely manner. These features are in line with current guidelines for hypertension management.⁴⁰ In this review, the effect of telemedicine on BP control was not statistically significant. The RR for not achieving the goal for BP control was 0.60 (95% CI: 0.33–1.11). In contrast, Ma et al.¹⁴ reported an RR of 0.69 (95% CI: 0.57–0.84). This inconsistency could be largely explained by the relatively small sample number of studies included in this review. Despite the null findings on BP control, the reduction in BP is clinically important. In this review, patients also had comorbid diabetes, which put them at an increased risk for severe macro- and micro-vascular complications. It has been reported that each 2 mm/Hg decrease in SBP or 1 mm/Hg decrease in DBP is related to a 10% and 7% reduction in mortality from stroke and ischemic heart disease, respectively.⁴¹ Thus, patients with coexisting hypertension and diabetes may benefit more from telemedicine in terms of reducing BP.

Telemedicine may improve diabetes control. Specifically, Wakefield et al.³⁸ reported an ∼0.40% decrease in A1C by the 6-month telemedicine intervention. In addition, the group difference in change in A1C was 0.51% between the high-intensity intervention group and the control group. However, this effect was not maintained at 12 months. Interestingly, in Frias et al.'s study,³⁵ telemedicine did not decrease A1C when all patients were included in the analysis. In patients with a baseline A1C ≥8%, the intervention significantly decreased A1C by 0.57–0.98%. These findings coincide with a recent review.⁴² It was found that telemedicine significantly reduced A1C by around 0.5% in patients with diabetes. A higher reduction rate was found in those with higher baseline A1C (>8%). Based on the above evidence, telemedicine may have added values for patients with hypertension and comorbid diabetes by decreasing A1C as well. Patients with poorly controlled diabetes might benefit more from using telemedicine. More clinical trials are needed to confirm these findings.

Evidence is scarce about the effect of telemedicine on the process of care in patients with hypertension and comorbid diabetes. In the two studies that evaluated medication adherence, one³⁵ found that the mean medication ingestion adherence was between 84% and 86%. Another³⁹ reported high levels of medication adherence with no significant group difference. In a previous review, Xiong et al. found evidence that telemedicine programs improved medication adherence among people with hypertension.¹⁹ Logan et al.³⁷ used telemedicine to enhance BP monitoring and found that after the intervention, the average number of BP readings was 12.3/week, the average weekly adherence rate was above expectations. Only one study assessed the effect of telemedicine on self-efficacy³⁹ and did not find a significant impact. This finding could be attributed to the design of the intervention, which failed to address self-efficacy. Recent evidence indicates that telemedicine should also target improving self-efficacy for patients with chronic diseases (e.g., diabetes) during the design stage.⁴³ Both medication adherence and BP/glucose monitoring are essential components of self-care, which is the key to disease control. Meanwhile, self-efficacy-focused programs have been associated with increased diabetes control and hypertension management.^44,45 Given the limited number of studies investigating the impact of telemedicine on the process of care indicators, future studies should consider including the assessment of these indicators.

In this review, telemedicine demonstrated the potential to improve the experience of care in patients with hypertension and comorbid diabetes. For instance, in Frias et al.'s³⁵ and Logan et al.'s³⁷ studies, the participants reported that the telemedicine system can be easily used. The data viewing/sharing functions and interaction with the providers helped them to manage their diseases. In the Wakefield et al. study,³⁹ the participant has a moderate-to-high perception of technology with a mean score of 3.7–3.8 (out of 5). Based on a review conducted in patients with hypertension, telemedicine interventions were generally accepted by the participants and easy to use.¹⁵ Current interventions typically focus on the outcome and process of care (e.g., BP and medication adherence). Emerging evidence supports the inclusion of experience of care when assessing QoC.^26,46,47 Particularly, for people with multiple chronic conditions, their preferences and voices in the form of self-reported experiences are critical toward achieving high-performing health systems that can meet their health care needs.²³ Thus, more research is needed to confirm the effect of telemedicine on the experience of care.

To the best of our knowledge, this review was among the first that investigated the use of telemedicine in patients with hypertension and comorbid diabetes. Each condition requires intensive daily self-care and treatment regimens. We also took into consideration the complexity of QoC by including different indicators such as outcome, process, and experience of care. However, there are limitations to acknowledge. Despite an exhaustive search, the sample size was relatively small, suggesting that using telemedicine in patients with co-existing conditions remains an emerging area. This review may not be able to capture some significant effects due to the small sample size. Relatedly, all clinical trials were included in this review regardless of the design. Quality appraisal suggests that the quality of the included studies was moderate. Thus, the causal inference may be undermined. In addition, we cannot rule out the possible presence of the Hawthorne effect. With the Hawthorne effect, behavioral changes could be a result of being in the study instead of the interventions.⁴⁸ Nonetheless, it has been argued that the Hawthorne effect might not be a major concern as it is usually transient.⁴⁹ Given the above limitations, future studies with a more rigorous design are needed to validate our findings.

Conclusion

In conclusion, telemedicine demonstrated its potential to improve QoC, particularly outcomes of care, for patients with hypertension and comorbid diabetes. More clinical trials are much needed to examine the effect of telemedicine on other indicators of QoC, including process and experience of care. Such trials should use a more rigorous design and consider the complexity of multimorbidity. This line of work could provide further evidence for the management of chronic diseases using telemedicine and thereby facilitate the improvement of patient-centered health care systems.

This systematic review has important implications for research and clinical practice. Current interventions tend to focus on a single disease and do not capture the complex processes required for the management of multimorbidity. This review demonstrates the usability of telemedicine in patients with concurrent hypertension and diabetes. Thus, future research may consider using telemedicine among people with multimorbidity. Corresponding intervention should address commonly reported issues such as inadequate infrastructure, lack of equipment, and technology gap.⁵⁰ There are extra challenges to use telemedicine in people with multiple chronic conditions. For instance, telemedicine programs focusing on medical management and information provision may not be adequate for patients with chronic diseases and should also provide behavioral and emotional support.⁵¹ The system should also incorporate more comprehensive functionalities, making it more effective.¹⁵ Meanwhile, future research should add an assessment of the experience of care. For clinical practice, health care providers may consider using available telemedicine systems to facilitate communication and interaction with their patients and thereby help them with disease management. Multistakeholders (e.g., system development company, the insurance company, health care providers, and patients) should work together to explore the best way to deliver health care to the patients.

Authors' Contributions

W.Z. and B.C. contributed to data collection, data interpretation, and drafting of the article. W.Z. and X.H. contributed to data analysis, data interpretation, and drafting of the article. C.S. contributed to study conception and design, data interpretation, and critical revision of the article. All authors have given final approval of the version to be published and agreed to be accountable for all aspects of the work.

Disclosure Statement

No competing financial interests exist.

Funding Information

No funding was received.