Effectiveness of interventions based on patient empowerment in the control of type 2 diabetes in sub‐Saharan Africa: A review of randomized controlled trials

Abstract Background It is estimated that 1.6 million deaths worldwide were directly caused by diabetes in 2016, and the burden of diabetes has been increasing rapidly in low‐ and middle‐income countries. This study reviews existing interventions based on patient empowerment and their effectiveness in controlling diabetes in sub‐Saharan Africa. Method PubMed, MEDLINE, EMBASE, CINAHL, Web of Science, PsycINFO and Global Health were searched through August 2018, for randomized controlled trials of educational interventions on adherence to the medication plan and lifestyle changes among adults aged 18 years and over with type 2 diabetes. Random‐effects meta‐analysis was used. Results Eleven publications from nine studies involving 2743 participants met the inclusion criteria. The duration of interventions with group education and individual education ranged from 3 to 12 months. For six studies comprising 1549 participants with meta‐analysable data on glycaemic control (HbA1c), there were statistically significant differences between intervention and control groups: mean difference was −0.57 [95% confidence interval (CI) −0.75, −0.40] (P < .00001, I 2 = 27%). Seven studies with meta‐analysable data on blood pressure showed statistically significant differences between groups in favour of interventions. Subgroup analyses on glycaemic control showed that long‐term interventions were more effective than short‐term interventions and lifestyle interventions were more effective than diabetes self‐management education. Conclusion This review supports the findings that interventions based on patient empowerment may improve glycaemia (HbA1c) and blood pressure in patients with diabetes. The long‐term and lifestyle interventions appear to be the most effective interventions for glycaemic control.


| INTRODUC TI ON
Patient empowerment has evolved since the Alma-Ata conference in 1978 into one of the health promotion strategies in the Ottawa Charter of 1986, and nowadays as one of the general principles of the World Health Organization (WHO)'s Global action plan for the prevention and control of noncommunicable diseases (NCDs) 2013-2020. 1 Chronic NCDs such as diabetes are among the leading causes of morbidity and mortality in sub-Saharan Africa (SSA). 2,3 Diabetes is a long-term management disease, and its management is quite expensive for patients and their families who carry its financial burden in SSA, given the shortfall or nonexistence of the health insurance system. [4][5][6] It is characterized in much of SSA by a preponderance of patients' nonadherence to therapeutic plans, and there is an urgent need to implement cost-effective patient-based interventions that empower patients to control their own disease.
Self-management of diseases 7 or patient empowerment 8 is broadly defined as the most important decision taken by the patient affecting the diabetic patient's health and well-being. Based on such definition, the content, complexity and effectiveness of patient empowerment interventions vary significantly from one study to another. It varies in terms of study's aims, target behaviours (eg self-monitoring of blood glucose, diet or exercise), intensity, duration, place of delivery (eg clinic-or community-based), mode of delivery (eg group or individual), type and training of the facilitator (eg physician, nutritionist, nurse or peer) and theoretical underpinnings. 8,9 Indeed, the development of interventions based on the patient empowerment approach has been influenced by several theories of health behaviour change. [10][11][12] Antonovsky 13 proposed the salutogenic theory to summarize and operationalize patient empowerment in three dimensions: intelligibility, manageability and meaningfulness from the patient's perspective.
These three dimensions constitute the sense of coherence (SOC), and a stronger SOC is predictive of salutogenesis or a production of health.
He also recommended the presence of internal and external resources as prerequisites to develop a stronger SOC for patient empowerment. 14,15 Therefore, interventions that integrate the three dimensions of SOC and the resources for patients are more likely to be effective for disease self-management by the patients. 14 Increasingly, scientific evidence supports the hypothesis that patient empowerment interventions improve patients' abilities, allowing them to better control their biochemical and physical parameters as well as their lifestyle. 9,16,17 Several systematic reviews have been conducted in high-income countries, sometimes showing inconsistent effects of patient empowerment in the control of chronic NCDs. 18,19 In a study on an African American population bearing a disproportionate burden of diabetes and its complications, Ricci-Cabello et al (2013) showed that PE interventions could be at least partially effective in improving both processes of care and health outcomes. To our knowledge, no review has been conducted so far in SSA. Positive effects recorded in some interventions 20,21 have been highly variable from one intervention to another, [22][23][24] and even not statistically significant in others. [25][26][27] This variability of interventions based on patient empowerment makes it difficult to assess their effectiveness, thereby limiting their usefulness in the decision-making process for the improvement of the quality of health care without a measurement of their effect sizes. This review considers the following research question: What are the existing interventions based on patient empowerment and their effectiveness in controlling diabetes in SSA?

| MATERIAL S AND ME THODS
The review was registered in PROSPERO (registration number: CRD42018095070).

| Participants
Only studies conducted in sub-Saharan Africa among adult patients aged 18 and over with type 2 diabetes mellitus 28 were selected.
There were no restrictions on patient sociodemographic characteristics, the background of the person providing the patient educational empowerment, the sample size or the target groups. Studies carried out on mixed populations of patients with type 1 and type 2 diabetes were excluded from this review because the results were not reported separately for type 1 and type 2 diabetes; our focus being on type 2 diabetes, it was not possible to extract relevant data.

| Interventions
All selected studies were randomized controlled trials of an educational intervention: diabetes self-management education, pharmacist-led intervention, lifestyle education programmes, and cognitive behavioural coaching and peer-led intervention. These interventions aimed to lead the patient to be able to self-manage type 2 diabetes (T2D) in terms of adherence to the medication plan, lifestyle changing and follow-up. The interventions varied in duration, intensity, frequency, strategy, topics and educational content. The selfmanagement of T2D was analysed using the three dimensions of the salutogenic theory: intelligibility (knowledge about T2D and related factors, disease process, complications and treatment options), manageability (taking medication, self-monitoring of blood glucose, insulin titration, measurement of food intake, frequent exercise and follow-up) and meaningfulness (psychosocial support). 17 Only studies describing interventions and the process of empowering T2D patients were included (Appendix S1).

| Control
The control or comparison group was the treatment as usual/standard care.

Primary outcomes
The two primary outcomes were the glycosylated haemoglobin (HbA1c) or fasting blood sugar (FBS) and self-efficacy in disease control. They are the primary outcomes used in the literature as direct outcomes when evaluating the effectiveness of intervention based on patient empowerment targeting diabetic patients. 19,29 Secondary outcomes The secondary outcomes were blood pressure (BP), lipid profile parameters (total cholesterol), physical parameters (body mass index) and lifestyle (diet, physical activity, smoking and alcoholism). When available, the use of services (hospital admission) and medication adherence were evaluated as secondary outcomes. Secondary outcomes of interest were outcomes often used in the literature as indirect outcomes to evaluate the effectiveness of intervention based on patient empowerment targeting diabetic patients. 29

| Data extraction
Two authors (AM and CO) independently reviewed all studies based on inclusion criteria, starting with the title and abstracts, and through the full publications to generate a final selection. In cases of disagreement between the two authors on the eligibility of a study, a discussion with the senior author (BKD) was necessary to find a point of agreement. An adapted PRISMA flow chart of study selection was used (Figure 1). 31 The two review authors independently extracted the data from studies that met the inclusion criteria using a summary table (Table 1). Any disagreements were resolved by discussion, and if required by the senior author. The original authors of each publication were contacted for any relevant missing information on the trial.

| Analysis
Two authors (AM and CO) independently evaluated the risk of bias in each selected study according to the recommendation of the

International Cochrane Collaboration for Systematic Reviews of
Interventions. 32 This involved a description and a judgement for random sequence generation, allocation concealment, blinding, incomplete outcome data, and selective outcome reporting and other potential sources of bias. The criteria for judgement were 'low risk', 'high risk' or 'unclear risk' and presented as percentages across included studies ( Figure 2); they were also assessed individually ( Figure 3). 32 Any disagreements between authors were resolved by consensus, or with consultation of the senior author (BKD).
A narrative description of population and study characteristics of selected studies were carried. Review Manager version 5.3 was used to perform statistical analysis. 32 A random-effects approach was used for all analyses because it was unlikely that the underlying data represented the true effect due to differences in the populations and interventions in the different studies. For continuous outcomes when the same measurement scale was used (eg HbA1c, BP, BMI), the mean difference was calculated. Results were described narratively for continuous outcomes with different measurement scales such as self-efficacy, when treatment effects such as standardized mean differences (SMDs) were not quantifiable due to insufficient data to allow formal meta-analyses.
Heterogeneity was identified by visual inspection of the forest plots and by using the chi-square test (significance level of 0.1) and I 2 statistic (0%-40%: might not be important; 30%-60%: may represent moderate heterogeneity; 50%-90%: may represent substantial heterogeneity; and 75%-100%: considerable heterogeneity). 33,34 When heterogeneity was found (I 2 ≥ 50% or P < .1), we examined individual study and subgroup characteristics to determine its potential sources. We performed subgroup analysis as a hypothesis-generating exercise. There were enough data to perform subgroup analyses on the duration and the type of intervention.

| Sensitivity analysis
We performed a sensitivity analysis excluding the studies that reported high losses to follow-up of all participants at the end of the intervention, the studies with nonsignificant results and the community-based intervention. The robustness of the results was tested by repeating the analysis using fixed-effects model and random-effects model.

| Results of the search
The search yielded 1495 publications and 767 after deduplication.
A screening based on title and abstract excluded 690 publications, 77 full publications that were assessed for eligibility, and 11 publications from nine studies were reviewed ( Figure 1). Table 1 summarizes the main characteristics of the 11 publications.

| Characteristics of included studies
In some cases, there were more than one publication from the same study, 20,26,35,36 leading a total of nine unique studies. These studies yielded a total of 2743 participants, 67% being female (n = 1838).

| Duration, intensity, frequency, types
The duration of interventions ranged from 3 to 6 months (n = 5) to 1 year (n = 4). Fully 32.48% (n = 891) of patients were lost to followup at the end of the intervention, mainly from one study. 35 The intensity of interventions was very similar across the different studies. The duration of education session varied from 45-60 minutes (n = 4) to 1-2.5 hours (n = 4) and was unspecified in one study (Table 1). 38 The frequency of the interventions varied widely from one study to another. They were classified as not frequent (n = 3) with  (Table 1). 37 Eight interventions were clinic-based, and one was community-based. 39 Interventions were diabetes self-management education (n = 3), 35,41,42 pharmacist-led intervention (n = 2), 38,40 lifestyle education programmes (n = 2), 24,26 cognitive behavioural coaching 37 and peer-led intervention. 39

| Strategy
Two strategies were adopted: group education (n = 5) and individual education (n = 4).The healthcare professional responsible for this education varied by the nature of the intervention (nurse, physician, pharmacist, dietician, psychologist, peer and health promoter). One intervention 24 had four types of healthcare professionals (physician, nurse, dietician and psychologist), while several other interventions had only one type of healthcare professional responsible for patient education (Table 1). [37][38][39]41,42

| Topic and educational support
All the selected studies included education on diabetes and related factors, plus self-management of the diseases. Additionally, the interventions provided patients with educational support in terms of material (pamphlets, booklets, etc), immaterial (phone call, text message, etc) and financial compensation (free charge of phone counselling, FBS test, etc); these supports were not provided in one intervention (Table 1). 37

Primary outcomes
HbA1c and FBS as indicators to measure the blood sugar level were mentioned in six studies 24,26,35,39,41,42 and one study, 38 respectively; the two other studies measured none of these parameters ( Table 1).
The effects of interventions on HbA1c were mixed. The studies reported results that favoured the intervention groups, with a statistically significant difference in the improvement of HbA1c (n = 2), 24,39 results that favoured the intervention but were statistically insignifi- 26,42 and no significant difference between control and intervention groups (n = 2). 35,41 For FBS, the study reported a significant difference within the intervention group only. 38 Six studies with 1549 participants contained enough data to be included in a meta-analysis 24,26,35,39,41,42  Four studies were identified with different measures of self-management in disease control. Two studies reported diabetes knowledge score, one with a significant difference between control and intervention groups 20 and the other one without. 39 One study 35 reported self-management in disease control in terms of psychological factor scores (self-efficacy, internal locus of control, external locus of control, chance locus of control) with no significant difference between control and intervention groups. One study 20 also reported a significant difference in patient autonomy in their disease management between groups. Depressive symptom scores were used in one study 37 to measure the self-management in disease control with a significant difference between control and intervention groups.

Secondary outcomes
Seven studies reported blood pressures 24,26,35,38,39,41,42 : three reported a significant difference between groups in the control of systolic blood pressure (SBP) and diastolic blood pressure (DBP) 24,35,42 ; one reported a significant difference within the intervention group for SBP and DBP 38 ; one found a difference for groups with SBP but not in DBP 39 ; and two reported no significant difference between groups for SBP and DBP. 26,41 Seven studies with SBP and DBP contained enough data to be included in a meta-analysis of 1699 participants as shown in  Figure S6, Appendix S7). A sensitivity analysis, excluding the heterogeneous studies, 26,41 reported no significant difference between groups for BP, followed by the study that reported high losses to follow-up of all participants at the end of the intervention, 35 respectively, dropped the heterogeneity for SBP to 75% with higher overall effect size of −7.29

Authors, publication year
Country (region of residence), type of place of residence Population size (N) and number of women (n)

Type of the facilitator Main outcomes: (1) primary outcomes and (2) secondary outcomes
Group education (not specified)

Type of the facilitator Main outcomes: (1) primary outcomes and (2) secondary outcomes
Group education
For lipid profile parameters, two studies measured the total cholesterol with no significant difference between groups. 26,35 For physical parameters, four studies measured BMI 26,38,39,41 and only one of them reported a statistical significance. 39 The four studies contained enough data to be included in a meta-analysis of 468 participants as shown in Figure S7 (Appendix S10). The pooled results indicate no statistically significant difference in outcomes of mean difference (MD) −0.82 [95% CI: −1.71, 0.08] (P = .07) with moderate heterogeneity in the effects of the interventions (I 2 = 50%; Figure   S7, Appendix S10). Two studies reported weight, one with statistical significance, 24 and the other without. 35 One study measured the cost-effectiveness of the intervention in terms of incremental cost-effectiveness ratio (ICER) and reported 1862 $/quality-adjusted life-year (QALY) gained, 35 but fail to show significant differences between groups.
The medication adherence was measured in three studies with significant differences, 38,40 except in one study 35 where there was no significant difference between groups.
One study reported hospital admissions with significant reductions in the intervention group compared to the control group. 40 Lifestyle was evaluated in three studies in terms of adherence to diet plan with significant differences between control and intervention groups in one study 38 and no significant difference in two other studies, 26,35 except for starchy foods servings/day that displayed significant differences between groups. 26 Physical activity plan was reported in two studies, with significant differences between groups in one. 38 One study 35 also evaluated the reduction of the frequency of smoking and found no significant difference between groups; they also used some parameters to evaluate the lifestyle (physical functioning, role functioning, social functioning, mental health, general health, pain) but no significant difference was found between groups.

| Risk of bias in included studies
All included studies were RCTs: most of those in 'selective reporting' were at high risk of bias, whereas those in 'random sequence generation' and 'incomplete outcome data' were at low risk of bias. Most studies did not provide details about the allocation concealment and blinding process (participants and outcomes); thus, it is difficult to make a judgement about how biased some of the studies may be ( Figures 2 and 3).
F I G U R E 2 'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies

| D ISCUSS I ON
Nine studies with 2743 participants were included in the review.
These studies included a wide spectrum of interventions covering clinic-and community-based interventions distributed into DSME, cognitive behavioural coaching, pharmacist, peer-led and lifestyle interventions.
The result of the glycaemic control indicated small but statistically significant differences in the mean difference of outcomes between intervention and control groups. The subgroup analysis showed that long-term interventions seem to be more effective than short-term interventions. Indeed, since diabetes is a chronic NCD and patients are likely to carry it for the rest of their lives, their experiences become rich sources of knowledge to use in developing long-term interventions. 14  improve patient knowledge about the disease (intelligibility) and help better participate in the process of decision-making as well as improve his self-efficiency. 46 A long-term intervention with a much-reduced frequency (one session every 3 months) even if the duration of each session is long (1-2 hours) 35,39 may not be long enough for the patient to develop the self-management ability that will help him control the disease. Empowering patient will likely improve his life quality or positively change his health behaviour such as physical activity, diet, smoking and alcohol. 17 Unfortunately, the two studies that showed a significant improvement in self-management in this review did not evaluate their impact on behavioural change. 20,37 Nonetheless, the nonsignificant improvement in patient self-management had no positive effects on behavioural change (such as diet, physical activity, smoking, medication adherence, quality-of-life score) and no significant change in glycaemic control. 35 Only four of the nine studies evaluated the self-management of the disease by patients after the intervention and only one study indicated the behavioural change that followed; this makes it difficult to know whether the observed effect actually derived from the intervention.
For the secondary outcomes, two studies with nonsignificant difference in BP after the intervention appeared to be the ones with low frequencies although the duration of education was long; the topic was well related to the self-management of diabetes with substantial support. 26,41 The other secondary outcomes were evaluated narratively; they mainly had a mixed effect which makes it difficult to conclude on the effectiveness of the interventions.
Components of patient empowerment intervention have been shown to be particularly effective when delivered by a multidisciplinary team. 47,48 As such, intervention with more than three types of healthcare professionals, especially with a psychologist, who is important for the 'meaningfulness' component of patient empowerment, seemed to be effective in all evaluated variables. Indeed, Antonovsky defined the 'meaningfulness' as the motivator that guides the other components of patient empowerment (intelligibility and management); intervention that includes this component in terms of psychosocial support will have a greater impact in the management of the diseases. But in this review, we cannot fully ascertain whether multidisciplinary settings led to the effectiveness of patient empowerment intervention since some included studies consisted of multidisciplinary teams which did not significantly impact the outcomes. 26,35 All the frameworks identified here were useful to explain how patient behaviours changed to become empowered and responsible of his own health care. This was particularly well with the case of the healthcare patient-centred approach, 16 which considers the characteristics, values and experiences of patients. However, the motivational interviewing framework 36 appears to have more appeal than others because it is considered as a motivational component, which serves as a driving force to combatting diseases and recover lost health. 17 Meaningfulness, which is central to the three components of the SOC, is also represented by this framework characterized by a collaborative approach that evokes ideas and solutions from the patients, is based on their experiential knowledge, and respects their choices and sense of control while attempting to empathically understand their perspective. 36 There are several strengths of this review. First and to our knowledge, this review is the first comprehensive review of evidence on patient empowerment interventions for diabetes patients in SSA.
Second, we use a structured procedure of data collection according to PRISMA, 30  Overall, we believe that the findings from this review are of importance to clinicians, researchers, patients and policymakers directly or indirectly involved in the prevention and control of type 2 diabetes in sub-Saharan African countries as well as similar resource-limited settings elsewhere in the world.

| SUMMARY
This review supports the findings that interventions based on patient empowerment may improve HbA1c and BP in patients with diabetes. The long-term or the lifestyle interventions appear to be the most effective in terms of improving glycaemic control. It was not possible to determine the effectiveness of intervention in all selected outcomes; those classified as most frequent, that utilized support for patient education, and multidisciplinary teams were associated with improved outcomes. However, more evidence from high quality of studies on most interventions with large sample sizes is required to support future patient empowerment programmes. Existing interventions are poorly implemented in the context of SSA, and there is a need to contextualize and standardize their implementation, by using the same definition of patient empowerment and by using the same indicators to evaluate the effects of the intervention.

ACK N OWLED G EM ENTS
We thank Sylvie Fontaine for her assistance in the search review process and Isambert Leunga for his assistance in statistical data analysis. We would also like to thank the 'Fonds de Recherche du Québec-Société et Culture (FRQSC)' for funding this project. The funder did not contribute or influence the findings or conclusions.

CO N FLI C T O F I NTE R E S T
Nothing to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author, (AM), upon reasonable request.