The prevalence of diabetes is rapidly rising around the world, especially in low- and middle-income countries (Ogurtsova et al, 2017; World Health Organization, 2018). In 2015 it was estimated that, globally, nearly 415 million people were living with diabetes (Ogurtsova et al, 2017). Type 2 diabetes accounts for 90% of cases (Bruno et al, 2005; Holman et al, 2015; Zheng et al, 2018).
The growing prevalence and incidence of type 2 diabetes, coupled with its costly complications, impose a significant burden on health services. It is predicted that about 12% of global health expenditure is dedicated to the health-related costs of the condition (Ogurtsova et al, 2017). In the UK, the cost of type 2 diabetes to the NHS has been estimated to be about 10% of the NHS budget (£8.8 billion), of which approximately 80% (£7.7 billion) is spent on associated complications (Diabetes UK, 2014).
The incidence of diabetes and the associated costs and complications create a need for self-management strategies and the implementation of more effective treatment for people with diabetes. Diabetes self-management education (DSME) programmes are widely advocated for all individuals with diabetes to enhance their self-care capacities (Inzucci et al, 2015; Ogurtsova et al, 2017; National Institute for Health and Care Excellence (NICE), 2019). These education programmes are associated with a reduction in the microvascular and macrovascular complications associated with diabetes (Aldasouqi and Gossain 2011; Speight, 2013). While face-to-face group education is now the gold standard for self-management education, attendance is low (Pal et al, 2013). In addition, as diabetes is a progressive condition, people need to continuously update their knowledge and skills about living with it. Importantly, with increased demands on health services, it is not surprising that it is becoming more challenging for health professionals to support people living with diabetes in this manner (Kelly et al, 2018).
The aims of diabetes self-management education programmes are set out in Table 1.
Table 1. The aims of diabetes self-management education programmes
| Topic 1 | Improving knowledge, health beliefs and lifestyle changes (healthy eating, weight management, physical activity, coping with stress, and tobacco cessation) |
| Topic 2 | Improving patient outcomes, for example: weight, HbA1c, lipid levels, smoking and psychosocial changes, such as quality of life, and levels of depression |
| Topic 3 | Improving levels of physical activity |
| Topic 4 | Reducing the need for, and potentially better targeting of, medication and other items such as blood-testing strips |
HBA1c=haemoglobin A1c
The delivery of self-management interventions via online programmes provides an important facility to people living with diabetes (Bolle et al, 2015). Additionally, they can be supported by incorporating digital consultations with health professionals, including feedback and messaging. In secondary care the use of digital consulting, at an expedient time for the patient, has been found to make a difference to how they manage their long-term condition (Griffiths et al, 2017). There is less known about its use and impact in primary care-managed long-term conditions (Atherton et al, 2018; Sturt et al, 2018).
Examples of interactive components in online applications include the person with diabetes inputting data such as knowledge, emotional responses or blood glucose levels and receiving health professional feedback to support their next steps (Glasgow and Bull, 2001; Pal et al, 2013). Therefore, people with diabetes may benefit from a greater understanding of their condition, as well as skills to self-care at home, which are likely to reduce the burden of disease (Bond et al, 2007; Bolle et al, 2015).
Midlife adults (aged between 50 and 65 years) are the main population affected by type 2 diabetes (Li et al, 2016). They can face several challenges, including having to manage their diabetes in the workplace. These include many disease management issues such as difficulties in blood glucose monitoring, insulin administration and in securing time off to attend appointments (Ruston et al, 2013). In addition to these challenges, healthcare behaviours in midlife may have a long-term effect on the nature of ageing (Ansari et al, 2014). Therefore, optimising care during midlife is essential for the prevention of microvascular and macrovascular diabetes-related complications in older age (Li et al, 2016; Mondesir et al, 2016).
A systematic review of reviews evaluated technology-enabled DSME for people living with diabetes (Greenwood et al, 2017). This included 25 systematic reviews, of which 18 found that technology-enabled DSME can considerably improve HbA1c. This systematic review also demonstrated progress in DSME. However, the studies included all age groups (people aged 1 to 80 years) in the analysis and did not explicitly include digital consulting elements. As the median age of type 2 diabetes diagnosis reduces (Wilmot and Idris, 2014) and digital literacy of the population expands (Dutton and Blank, 2011) there is a need to specifically understand the evidence supporting the growth of online self-management programmes in long-term conditions.
Aim
The aim of this systematic review was to determine the impact of online self-management interventions on health outcomes in midlife adults with type 2 diabetes.
Method
A systematic review of randomised controlled trials (RCTs) was undertaken to answer the research question. The databases searches followed the Population, Intervention, Comparator, and Outcome (PICO) framework (Santos et al, 2007). The Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidance and the Cochrane Handbook for Systematic Reviews of Interventions were followed for the preparation and development of this systematic review (Moher et al, 2009; Higgins and Green, 2011).
Study selection
The inclusion and exclusion criteria applied to articles are set out in Table 2.
Table 2. Inclusion and exclusion criteria
| Inclusion criteria | Exclusion criteria | |
|---|---|---|
| Research method | RCTs | All non-RCTs (cohort study, case study etc) |
| Participants | Type 2 diabetes (study mean age 50 to 65 years) | Other conditions (type 1 diabetes, cardiovascular disease etc) |
| Interventions | Web-based self-management interventions, providing two-way communication via feedback, and messaging methods | Studies focusing on describing only the development of web-based interventions; studies including different intervention such as community-based interventions |
| Outcomes | Primary outcome: HbA1cSecondary outcome: cardiovascular risk factors (blood pressure, and total cholesterol) diabetes distress or depression, and self-efficacy | Studies that do not evaluate the HbA1c values |
| Language | English language | Non-English language |
HBA1c=haemoglobin A1c; RCTs=randomised controlled trials
Search strategy
Systematic searches were carried out in the three largest databases via Ovid and EBSCO interfaces: Medline, Embase and CINAHL. Initially, database searches were performed in May 2018 without time limitation, and then updated in April 2019 to identify whether there were any newly published articles. The following research terms and operators were used: type 2 diabetes mellitus OR type 2 diabet* OR T2DM OR noninsulin dependent diabetes AND internet OR online OR web OR internet applicat* OR web-based OR internet-based. The additional search terms were self-management OR self-management education AND glycated haemoglobin A OR glycaemic control.
Risk of bias (quality) assessment
RCTs selected in this review were first appraised using the Cochrane Collaboration's tool for assessing risk of bias (Higgins and Green, 2011). Second, the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) system was used to appraise the quality of the included papers (Balshem et al, 2011).
Data extraction
All studies retrieved from the searching process were initially reviewed by the first author (AC) using a standardised pre-piloted template. The following characteristics were extracted: the authors, the year of publication, country, study design, the number of participants, participant age range and mean age, gender, type of patients, the aim of the study, intervention description, comparison, follow-up, outcomes, results, and intervention components, including intervention provider, digital infrastructure, patient education, emotional change, health check attendance, supportive counselling, goal setting/action planning, problem solving, and other web-based resources. The second and third authors (RF, JS) checked the extracted data. Disagreements about the studies were resolved through discussion between all reviewers. Missing data were not requested from study authors.
Data synthesis
The meta-analysis was conducted by computing the mean differences of HbA1c values using random effect model and demonstrated as a forest pilot. Heterogeneity among papers was calculated using chi-squared (χ²), and I squared (I²) (Higgins et al, 2003).
The data concerning secondary outcomes (cardiovascular risk factors such as total cholesterol and blood pressure, self-efficacy, depression and diabetes distress) were synthesised by using a narrative approach due to the substantial measurement heterogeneity between the studies.
Results
Selection of the studies
A databases search identified 2004 citations. A total of 35 full-text studies were selected for full assessment to check their eligibility criteria. After the full-text screening, eight RCTs were included in this review for data extraction. The references of the included studies were also reviewed, and no additional papers were found. See the PRISMA flow diagram (Figure 1) (Moher et al, 2009).
Characteristics of included studies
In total, eight RCTs with 1760 participants were included in this review (McKay et al, 2002; Cho et al, 2006; Ralston et al, 2009; Lorig et al, 2010; Avdal et al, 2011; McMahon et al, 2012; Tang et al, 2013; Murray et al, 2017). The mean age of participants varied between 50 and 65 years. In all included studies, participants were assigned to an online self-management intervention and compared against a control arm.
The main outcome and objective of all included studies were to reduce HbA1c levels. Four trials evaluated blood pressure at follow-up (Ralston et al, 2009; McMahon et al, 2012; Tang et al, 2013; Murray et al, 2017). Five trials assessed total cholesterol level (McKay et al, 2002; Cho et al, 2006; Ralston et al, 2009; McMahon et al, 2012; Murray et al, 2017).
Only two studies assessed self-efficacy and used different tools; Lorig et al (2010) used a self-efficacy scale, and Murray et al (2017) measured it using the Diabetes Management Self-Efficacy Scale (DMSES).
Four trials assessed diabetes distress, three of which used the Problem Area in Diabetes (PAID) scale (McMahon et al, 2012; Tang et al, 2013; Murray et al, 2017). One study assessed diabetes distress using a health distress scale (Lorig et al, 2010). Although four trials assessed depression outcome, they used different measurement tools. Lorig et al (2010) and Tang et al (2013) used the Patient Health Questionnaire (PHQ-9); Murray et al (2017) used the Hospital Anxiety and Depression Scale (HADS); and McKay et al (2002) used the Centre for Epidemiologic Studies Depression Scale (CES-D).
Table 3 describes intervention components in all studies where interventions were explicitly identified. The interventions were based on patient education, including diet management, physical activity, emotional change including self-efficacy, distress management, health check attendance, supporting counselling, goal setting/action planning, problem solving, and other web-based sources.
Table 3. Intervention components
| Study author and year | Intervention title | Intervention provider | Digital infrastructure | Patient education | Emotional change | Health check attendance | Supportive counselling | Goal setting/action planning | Problem solving | Other web-based resources |
|---|---|---|---|---|---|---|---|---|---|---|
| Avdal et al (2011) | Web-based diabetes education | Diabetes nurses | Website | ✓ | ˟ | ✓ | ˟ | ˟ | ˟ | ˟ |
| Cho et al (2006) | Online-based glucose monitoring system | Three endocrinologists (a professor, two clinical instructors), a nurse, a dietitian | Website | ✓ | ˟ | ˟ | ˟ | ˟ | ˟ | ˟ |
| Lorig et al (2010) | Online-based diabetes self-management program | Two peers who were trained as self-management small group leader | Website | ✓ | ✓ | ˟ | ✓ | ✓ | ✓ | ✓ |
| McKay et al (2002) | Online-based diabetes support program | A professional who had expertise in providing dietary advice to diabetes patients | Web portal (a D-net system) | ✓ | ✓ | ˟ | ✓ | ✓ | ˟ | ˟ |
| McMahon et al (2012) | Web-based care management | Care manager | Website | ✓ | ˟ | ˟ | ˟ | ˟ | ˟ | ✓ |
| Murray et al (2017) | Web-based self-management support | Two nurses | Website called ‘Healthy Living for People with Diabetes’ (HeLP diabetes) | ✓ | ✓ | ˟ | ˟ | ✓ | ✓ | ˟ |
| Ralston et al (2009) | Web-based collaborative care | Care manager | Website | ✓ | ˟ | ✓ | ✓ | ✓ | ✓ | ˟ |
| Tang et al (201✓) | Online disease management of diabetes | Nurse care manager | Website called ‘Engaging and Motivating Patients Online with Enhancing Resources for Diabetes’ (EMPOWER-0) | ✓ | ˟ | ˟ | ˟ | ✓ | ˟ | ✓ |
Assessment of risk of bias in the included studies
Details of the risk of bias assessments of the included studies were separately performed for each domain, as shown in Figure 2.
Effects of online interventions with digital consulting
Primary outcome: HbA1c
All eight studies assessed HbA1c. In the random effect, when data in HbA1c levels from the studies were combined in a meta-analysis, the pooled results demonstrated a statistically significant improvement in HbA1c values (P<0.0001), in which HbA1c value was better than in the control group. It means that a model online self-management intervention was more likely to reduce HbA1c values; the pooled mean difference was -0.35%, 95% CI (-0.52 to -0.18) (P<0.0001), indicating a small effect size. The results demonstrated significantly high levels of heterogeneity (χ²=26.94, df=7, P=0.0003, I²= 74%) across the studies.
Secondary outcomes: cardiovascular risk factors
Blood pressure Four studies found that the effects of interventions on blood pressure were mixed. McMahon et al (2012) assessed the mean diastolic and systolic blood pressure and reported a statistically significant reduction in the control arm—diastolic blood pressure reduced from 83+/-15.8 mmHg to 77.3+/-11.3 mmHg (mean difference=-5.8 mmHg, P=0.012), whereas systolic did not change in either group. The study by Murray et al (2017) reported that the mean systolic blood pressure demonstrated a statistically significant decrease in the intervention arm, ranging from 134.7+/-1.5 mmHg to 130.5+/-1.5 mmHg (mean difference=-3.8 mmHg, P=0.01). Ralston et al (2009) and Tang et al (2013) found no significant difference between intervention and control arms.
Total cholesterol Five studies found that the effects of interventions on total cholesterol were inconclusive. The mean difference in the RCT by McMahon et al (2012) was found to be statistically significant in favour of the control arm, ranging from 174.0 mg/dl to 163.1 mg/dl (mean difference=-10.7 mg/dl, P=0.019). The other studies reported that there was no significant difference between the control and intervention groups after the follow-up.
Diabetes distress Four studies included changes in diabetes distress. Tang et al (2013) reported that diabetes distress significantly decreased in the web-based group (P<0.001). However, McMahon et al (2012) reported a significant reduction in the control group only (P=0.003). Lorig et al (2010) and Murray et al (2017) did not find any differences between the groups (P=0.064, P=0.209 respectively).
Depression Four studies measured depression-related outcomes. One study (McKay et al 2002) found a significant difference, with depression scores decreasing more in the intervention arm after 3 months measured by the CES-D (P<0.08), in which the depression score decreased from 18.11+/-11.79 at baseline to 16.42+/-11.79. The other studies stated that there were no differences between groups over the follow-up of the trials (Lorig et al, 2010; Tang et al, 2013; Murray et al, 2017).
Self-efficacy Two studies looked at self-efficacy. Lorig et al (2010) demonstrated that self-efficacy improved significantly in favour of the intervention group after 6 months and this result persisted to 18 months after the start of the trial (P=0.001). Murray et al (2017) did not find any significant difference between the groups (P=0.474).
Discussion
This systematic review assessed the impact of online self-management interventions with digital consulting on health outcomes in midlife adults (aged 50 to 65 years) with type 2 diabetes. The findings of the meta-analysis about HbA1c found a considerable, statistically and clinically significant improvement in the intervention arm, with the mean difference of -0.35%, 95% CI (-0.52 to -0.18) (P<0.0001), because reducing the HbA1c level by 0.2% is likely to reduce diabetes-related mortality by 10% (Stratton et al, 2000; Selvin et al, 2004; Hameed et al, 2012; Sherwani et al, 2016). This finding of HbA1c is consistent with the results of a systematic review of reviews by Greenwood et al (2017), which evaluated technology-based DSME for people living with diabetes. Greenwood et al (2017) highlighted the importance of improving HbA1c with reductions changing from 0.1% to 0.8% and included all age groups (1 to 80 years old) in their systematic review. However, the majority of systematic reviews in this review did not include HbA1c as an outcome. Therefore, this study emphasises that midlife adults can benefit from online self-management education with digital consulting to improve HbA1c values.
The evidence of online self-management interventions on blood pressure (diastolic and systolic blood pressure) and total cholesterol was inconclusive. According to the European Society of Hypertension/European Society of Cardiology (ESH/ESC), the American Diabetes Association (ADA), and NICE guidelines, blood pressure goals for people living diabetes are <140/90 mmHg to decrease the risk factors for cardiovascular morbidity and mortality (Mancia et al, 2013; Catena et al, 2015; NICE, 2011). After 12 months, Murray et al (2017) reported that systolic blood pressure was below 130 mmHg in the intervention arm, whereas McMahon et al (2012) reported diastolic blood pressure below 80 mmHg in the control arm. Thus, there is not enough evidence to show benefits on blood pressure.
One study (McMahon et al, 2012) reported evidence of improvement on total cholesterol in the control group; however, this result was not clinically significant (ranging from 174.0 mg/dl to163.1 mg/dl, P<0.01) because total cholesterol levels are considered to be desirable up to 200 mg/dl (Nayor and Vasan, 2016; Veeramalla and Madas, 2017). More research therefore is needed to determine whether online self-management is indeed important to reduce blood pressure and total cholesterol.
There was some evidence to show online interventions with digital consulting have positive effects on diabetes distress and depression. However, there seemed to be difficulty in converting the positive effects on diabetes distress and depression because only one of the four studies demonstrated a statistically significant improvement on these outcomes in questionnaire scores. Diabetes distress and depression were explicitly associated with glycaemic control in adults with type 2 diabetes (Fisher et al, 2015). Fisher et al (2016) indicated that, among type 2 adults, elevated diabetes distress is highly stable over time and that the point prevalence of elevated diabetes distress is approximately 46%, suggesting a widespread clinical problem in this population.
A study by Sturt et al (2015) stated that motivational interviewing is found to significantly reduce both diabetes distress and HbA1c. This has been explained in relation to improvements in diabetes management self-efficacy through knowledge and skill acquisition in communication and reflection. According to a systematic review by Joo (2016), participants' self-efficacy with the help of online resources might improve owing to education opportunities without time limitations and online counselling services that provided prompt feedback from healthcare providers. However, this study did not find evidence to support online interventions with digital consulting on self-efficacy, diabetes distress and depression.
Methodological consideration, strengths and limitations of the systematic review
This review included RCTs, strengthening the internal validity of the review. No serious indirectness was considered because the results found were directly associated with the research question in this review (Balshem et al, 2011). Also no serious imprecision was considered owing to the fact that four out of the eight studies reported power calculation, and so the number of the participants was powered. As a result, in this review, the effect estimate in terms of the quality of evidence was moderately confident according to the GRADE approach (Balshem et al, 2011).
This study has several limitations: first, the three electronic databases (Medline, Embase, CINAHL) were searched to find relevant papers; other databases were not used due to resource limitations, and so some papers may have been missed. Second, this review included only the papers published in English, which may have excluded other potentially relevant studies published in other languages. Third, although the studies included in this review are all RCTs and the quality of evidence is relatively high, only one study reported blinding of the outcome assessment process and participants. Therefore, it is possible that performance and detection bias were introduced. Fourth, in some studies, control groups used some form of intervention rather than purely usual care, such as internet access and training without a coach. Furthermore, the context of ‘usual care’ across studies may also vary. Thus, this might also bias the general effect of online interventions with digital consulting.
Recommendations for future practice and research
The use of online interventions with digital consulting has the potential to improve the HbA1c of midlife adults with type 2 diabetes and these interventions might meet the needs described by midlife adults, including emotional management and behaviour change, providing access to healthcare professionals. This review included only quantitative studies, and thus no patient feedback was found in the data. More research is needed to understand the experiences and ne eds of midlife adults living with diabetes using the online interventions.
The addition of digital consultations to online interventions is not well understood and more research is needed into the delivery and effectiveness of consultations. The authors suggest that future studies would benefit from more evidence-based components and more detailed evaluation and reporting.
Future studies are also needed to determine the cost of online interventions with digital consulting on diabetes care as diabetes is a costly chronic disease to manage. Therefore, the development of a cost-effective intervention is necessary. However, online interventions are associated with high costs, including internet services, technical services, and system development.
Conversely, these interventions might also be effective tools to reduce the overall burden of health care, including the cost of long-term conditions in respect of health professionals' time and hospital visits. The studies did not report their interventions' cost-effectiveness, therefore further research is needed to examine accessible internet infrastructure to determine the cost of online interventions on diabetes care.
Conclusion
This review has demonstrated that online self-management interventions with digital consulting have a small beneficial effect on HbA1c in midlife adults with type 2 diabetes by a minimally important difference. Since such online interventions represent tailored, patient-centred, efficient, and good-quality care, health professionals need to consider adopting them as self-care interventions. It is also recommended that the cost of online self-management should be explored in future trials.
KEY POINTS
- Online self-management interventions have a small beneficial effect on haemoglobin (HbA1c)
- Online self-management interventions are recommended due to representing patient-centred, tailored, efficient, and quality care
- Further research is needed to determine whether online self-management interventions with digital consulting are effective in improving cardiovascular and wellbeing outcomes
- The addition of digital consulting to online interventions is poorly understood and more research is needed into their effectiveness and delivery
CPD reflective questions
- Think about the behavioural change techniques you use with your patient group—which have been the most successful?
- If you use paper-based information to help patients achieve behavioural change, consider whether online self-management interventions with digital consulting would be helpful
- Consider how you would go about setting up such a service or improving your current service