Retention in care for type 2 diabetes management in Sub‐Saharan Africa: A systematic review

Abstract Objective Diabetes prevalence has risen rapidly in Sub‐Saharan Africa, but rates of retention in diabetes care are poorly understood. We conducted a systematic review and meta‐analysis to determine rates of retention in care of persons with type 2 diabetes. Methods We searched MEDLINE, Global Health and CINAHL online databases for cohort studies and randomised control trials (RCTs) published up to 12 October 2021, that reported retention in or attrition from care for patients with type 2 diabetes in Sub‐Saharan Africa. Retention was defined as persons diagnosed with diabetes who were alive and in care or with a known outcome, while attrition was defined as loss from care. Results From 6559 articles identified, after title and abstract screening, 209 articles underwent full text review. Forty six papers met the inclusion criteria, comprising 22,610 participants. Twenty one articles were of RCTs of which 8 trials had 1 year or more of follow‐up and 25 articles were of non‐randomised studies of which 19 had 12 months or more of follow‐up. A total of 11 studies (5 RCTs and 6 non‐randomised) were assessed to be of good quality. Sixteen RCTs were done in secondary or tertiary care settings. Their pooled retention rate (95% CI) was 80% (77%, 84%) in the control arm. Four RCTs had been done in primary care settings and their pooled retention rate (95% CI) was 53% (45%, 62%) in the control arm. The setting of one trial was unclear. For non‐randomised studies, retention rates (95% CI) were 68% (62%, 75%) among 19 studies done in secondary and tertiary care settings, and 40% (33%, 49%) among the 6 studies done in primary care settings. Conclusion Rates of retention in care of people living with diabetes are poor in primary care research settings.

INTRODUCTION About 1.6 million deaths worldwide are attributed to type 2 diabetes mellitus (hereon referred to as diabetes) annually [1][2][3][4][5][6]. The prevalence of diabetes in Sub-Saharan Africa (hereafter referred to as Africa) has risen rapidly [6][7][8], now standing between 3% and 10%, and the largest global increase in prevalence in the next two decades is expected to occur on this continent [3,9]. Diabetes is affecting a younger and a poorer population in Africa than is the case in high-income countries, with almost 80% of all diabetesrelated deaths occurring before people reach the age of 60 years, and high levels of morbidity from diabetes-related complications [9,10]. The effective control of diabetes is therefore a major challenge facing health services.
Prevention of morbidity and mortality from diabetes requires the disease to be controlled adequately. As a first step, this requires patients to be identified, linked to and retained in care. This is difficult in Africa, where health systems have typically focussed on acute management of infectious diseases and are ill-prepared and ill-resourced to respond to the chronic nature of care required for diabetes [11][12][13][14]. The impact of this is seen in the low achievement of optimal diabetes treatment goals in the region [10]. For patients, the costs of transport and other related expenses can be catastrophic and accessing care regularly for diabetes presents a huge problem [15].
Experience with the HIV care programme suggests that provision of chronic care in Africa is challenging but achievable [16]. Unlike HIV, diabetes care does not have the dedicated resources HIV care has received. Unmet need for diabetes care is high with one cross-sectional study finding that of all people with diabetes, only 40% had been diagnosed and 38% were on treatment [17]. With the rising burden of diabetes, knowledge of the rates of retention in care is essential for planning a response to this epidemic. We conducted this systematic review and meta-analysis to assess the rates of retention in diabetes care for adults in Sub-Saharan Africa.

METHODS
The protocol for the review is registered on the PROSPERO database (registration number CRD42018112400). Institutional review board approval was not required since the information used was published and publicly available. We present our findings according to Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines (Appendix 1).

Search strategy and eligibility criteria
We searched MEDLINE, Global Health and CINAHL for articles published in English up to 12 October 2021 and reporting retention or attrition rates in patients with type-2 diabetes managed in health facilities across Sub-Saharan Africa. The main search terms were: 'diabetes mellitus', 'dysglycaemia', 'hyperglycaemia' and 'hyperinsulinemia' coupled with 'retention in care', 'retention rates', 'attrition rates', 'lost to follow up', 'patient dropout', 'patient adherence' and 'patient compliance'. These were combined using Boolean characters 'OR' 'AND'. We searched the reference lists of included papers for additional papers or crossreferenced studies. A detailed description, including MESH terms, of the search is outlined in Appendix 2.
Two authors independently screened the titles and abstracts for relevance, and a third review author was consulted to resolve disagreements.
We included randomised trials, retrospective and prospective cohort studies. For studies with multiple publications, we selected the article that included outcomes relevant for the review. We excluded letters to the editor, reviews, editorials and commentaries, conference abstracts or unpublished studies, studies with fewer than 30 participants or less than 3 months of follow up, and studies or programme descriptions which followed up patients in diabetes care but did not have enough data to compute an estimate of the retention of patients in care or attrition rates. We included studies that reported on retention in care in type 2 diabetics where patients with hypertension and type 1 diabetes where also included and where possible, separated out the results in type 2 diabetics.

Data extraction
Using pre-tested data extraction forms, we independently extracted the following general information from each included article: surname of the first author; date of publication; study design; country of study; sample size; method of patient recruitment; type of intervention; age range of participants; sex; follow up period and the proportion of participants retained in care. Any differences between the two entries were resolved by discussion with a third author.
We contacted the corresponding authors of included articles where valuable information (retention rates or attrition rates for diabetes patients in care and sample size) was missing in the text or not in a retrievable format. Reminders were sent weekly and if there was no response after a month, the article was excluded.

Definitions
We defined diabetes care as receiving any of the following interventions: drugs, advice or support, diet and lifestyle modification aimed at controlling blood sugar and reducing risk of complications in patients diagnosed with type 2 diabetes.
Retention was defined as being alive and in care at the end of the study or period of evaluation, or where the patient mortality was known and occurred in care. Retention was also considered as the converse of attrition, which is measured as a combination of loss from care or death. This loss from care is described where a participant has not returned or made any contact with the clinic for about 3-6 months. Rates of retention were derived from attrition by subtracting the reported attrition from 100%.

Assessment of methodological quality and risk of bias
The quality of studies and the risk of bias were assessed independently by two authors using the quality assessment tool for observational studies and randomised control trials (RCTs) of the National Health Institute/National Heart, Lung, and Blood Institute. Any disagreements that occurred between the two reviewers in the course of assessing methodological quality and risk of bias were arbitrated by a third reviewer.

Data synthesis and analysis
The primary analysis for all outcomes was by intention totreat where the denominator was the number of patients enrolled in the study. Participants who were lost to followup were considered as not retained. We excluded multi-site trials where we were not able to separate data for Sub-Saharan African countries.
To calculate pooled retention rates by study arm at different follow-up time points for RCTs, a generalised linear mixed model with a binomial distribution and log link function was used. The model had study arm (intervention arm or control arm), follow-up time, interaction between study arm and follow-up time as fixed effects, study as random effect. For cohort (non-randomised) studies, the pooled retention rates at different follow-up time points were also estimated using a similar generalised linear mixed model but only with follow-up time as the sole predictor. Generalised linear mixed model was estimated using the Proc Glimmix in SAS version 9.4.

RESULTS
A total of 9002 articles were identified initially by the search (Figure 1 and Appendix 1). After removal of duplicates, 6559 studies remained. The titles and abstracts were screened to eliminate 6350 irrelevant articles. The full texts of the remaining 209 studies were scrutinised with 46 articles meeting the inclusion criteria. These comprised 21 RCTs and 25 cohort studies with total of 4557 and 18,053 evaluable participants, respectively. The studies had been conducted in 14 countries in Africa: Cameroon, Democratic Republic of Congo, Ethiopia, Ghana, Kenya, Mali, Malawi, Mauritius, Nigeria, Rwanda, South Africa, Sudan, Tanzania and Uganda.
The characteristics of the RCTs included in the systematic review and meta-analysis are given in Table 1 and the non-randomised studies in Table 2. The range of follow-up for patients retained in care varied from 3 to 24 months in the RCTs and 3 months to 14 years in the cohort studies.
RCTs testing novel diabetes management strategies against comparison groups among people living with diabetes All of the RCTs enrolled persons with diabetes with the aim to test the effectiveness of novel intervention strategies for diabetes management: patient education and support programmes [18-20, 22, 24, 26, 28, 31-34, 36, 38], medications [35,65], facility-based interventions including financial incentives [23], exercise programmes [21,30,37] and electronic reminders of appointments and self-management [25,29]. These approaches were designed to enhance adherence to anti-diabetes medications, improve anthropometric and clinical outcomes (body mass index, glycaemia and HbA1C levels) and to retain people in care. The comparison group was typically standard care. Only one RCT reported a financial incentive partially covering costs of drugs [23] and in the remainder, provision of drugs was left to the health service.
Of the 8 trials with more than 12 months follow-up, 5 were done in secondary and tertiary care settings, with rates of retention ranging from 84% to 93% (Table 3) [18,19,26,30,32] and 3 had taken place in a primary care setting and reported retention rates of 44% [23] and 55% [24] at 12 months and 38% [29] at 2 years.

Non-randomised studies reporting rates of retention in care
There were 25 cohort studies. A total of 19 enrolled participants from or treated participants in secondary or tertiary care health facilities and 6 were conducted in primary care settings.
The 19 studies conducted in secondary and tertiary care facilities had varying durations of follow-up. Five studies had follow-up durations of 3-6 months and they reported retention ranging from 45% to 89% [39,45,47,48,62]. Three studies had 12 months of follow up and reported retention rates ranging from 48% [55] to 74% [54] and 89% [43]. Four studies had 2-3 years of follow-up and reported retention rates between 68% and 90%. [49,53,57,61] Three studies had 4-5 years of follow-up and reported retention rates of 39% [59], 54% [60] and 81% [63]. Retention rates at 6 years of follow-up in two studies that included both type 1 and type 2 diabetics were 47% [44] and 63% [42]. One study in South Africa had a 12-year follow-up and reported a high retention rate of 95% coupled however with a high mortality rate of almost 80% [50]. Finally, one study had a follow-up period of 14 years [52]. It comprised both people living with type 1 and type 2 diabetes and had a retention rate of 65% and a mortality rate of 12%. The 19 studies had a combined sample size of 18,123 participants and the combined pooled retention rate was 68% (95% CI 62%, 75%) ( Figure 2).

DISCUSSION
This systematic review shows that retention in research studies on diabetes care is poor in primary care settings in Sub-Saharan Africa. The review involved studies done in research settings, where patients will have had good access to care, education about diabetes and medicines adherence support. Four of the primary care studies were randomised trials and even among these, the retention rate was low. If retention is a challenge in research settings, then good retention rates will be impossible to achieve in real-life settings where access to services is limited: particularly for rural populations, shortages of drugs and diagnostics are common and patients have high out-of-pocket costs. Indeed, one author commented 'it was quite tough and expensive to have sustained the participants through the study' in reference to a RCT with 6 months of followup [22].
Only about 5%-10% of people living with diabetes are estimated to be in care and given the limited routine glycaemia testing, almost all are identified after complications, that is, these patients will have had undiagnosed disease for some time [1]. Hence it is possible that patients who come into care (and were part of the studies in our systematic review) were relatively more motivated and that the low rates of retention we identified are over-estimates. In contrast, about 80% of people living with HIV now know their HIV status. The majority are identified when they do not have symptoms and start treatment promptly [66]. Over 80% of those who are HIV positive are in care and virally suppressed.
This large inequity in care between different chronic conditions raises major ethical dilemmas [67].
Our study also showed that retention in care was much better for people treated in tertiary and secondary care centres, particularly in randomised trials, reaching the levels seen in HIV care. The reasons for this higher retention are not clear. It is possible that these patients are experiencing complications and are symptomatic and therefore more likely to attend health services. It is also possible that these are a select group of patients or that there is a higher quality of care delivered at these settings. Health systems in these countries do not have the infrastructure to deliver chronic disease management at the primary care level, where the vast majority of people with diabetes will be treated [68]. From the experience with HIV and with diabetes in tertiary and secondary care, high retention is achievable, and primary care in low-resource settings needs urgent strengthening to deliver this.
Achieving high retention in diabetes care is a fundamental first step to effective control. This is evident from the gains in HIV control that have been possible because of the remarkable improvements in retention in HIV care over the 20 years [69]. We do not know why retention in diabetes care is so low. Suggested factors include lack of availability of trained healthcare workers, costs of access to care, interruptions in drug supplies and poor provision of health education and counselling [17]. In settings where there are high out of pocket payments, patients may attend intermittently or may move between private and public providers. However our evidence included randomised trials which provide comparatively good quality care, including covering some of the costs of care. In a large study conducted in Malawi, people with diabetes were identified after systematic population-based testing for diabetes. Those diagnosed were given free access to care at their local health centres, including free medicines and a dedicated clinician in a setting where medicines for diabetes were otherwise unavailable. Despite this, retention in care was less than 30% after 12 months [40]. Research is needed to better understand the reasons for poor retention, and approaches and interventions that could achieve good retention in a range of populations attending primary care. It is likely that education, support and empowerment focussed on patients and communities will be essential to increase understanding of the effects of diabetes, and to support patients to remain in care, as was done for HIV control.
The low retention in care we identified for diabetes is likely masking mortality due to diabetes, and it is essential that as well as interventions to improve retention in care, there is an urgent need to enhance medical recording systems at primary care level.
Limitations of this analysis include that we have used retention in research studies to estimate retention in diabetes care. Conducting research, particularly RCTs, at the primary care level in this context is difficult. As a result there was only a small number of studies we could use to estimate retention in primary care. In research studies the care delivered is more resourced, staff will be trained and emphasis is put on retaining people in the study, reflecting what might be achieved in a routine setting with more resources and likely overestimating true retention in care. The definitions of retention in care did differ from study to study, and we also used loss to follow up/attrition to indicate retention, the definitions of which will also have varied by study. Retention in care is influenced by a number of factors including a lack of availability of services, drugs and well-trained health care workers, accessibility of services and associated costs, and patient understanding of their condition; and these may differ from place to place. Our premise is that retaining people in care is the minimum and first requirement to achieve control of a condition. In summary, our systematic review shows low level of retention of care in studies of people living with diabetes and attending primary care services. We showed this in a number of settings, including in settings where highquality care was provided, such as randomised trials.
Research is needed to understand reasons for poor retention and to develop interventions to improve retention in diabetes care, as retention is an essential first step to improve health outcomes of people with diabetes. ORCID Anupam Garrib https://orcid.org/0000-0003-2305-3749