Design
This research is a secondary database study of routine data collected from patients with T2D attending primary care and followed from 1991 to 2019. This research was approved by the Institutional Review Board of the Institute of Tropical Medicine Antwerp in Belgium (IRB/RR/ac/149) and the Ethical Committee of the University of Kinshasa in DRC (ESP/CE/153B/2021). This study was guided by the Reporting of studies Conducted using Observational Routinely-collected health Data (RECORD) Statement [14], an extension from STROBE, The Strengthening the Reporting of Observational Studies in Epidemiology, guideline [15].
Setting
According to information and data from the World Bank in 2020, DRC is the largest country of SSA with approximately 89,561,404 million inhabitants and has a gross domestic product (GDP) annual growth of 1.7% [16, 17]. In 2018, it was estimated that DRC has one of the largest populations living in poverty, precisely the third one globally, as approximately 73% of the population lives with less than $1.90 a day [18].
In 2015, the World Health Organization (WHO) has described that DRC has a 3-level health system organisation: the implementation level, each district in DRC has a network of several health centres and district hospitals; the intermediate level focuses on the technical and logistic support and is mainly managed by provincial health departments; the central level has the normative responsibility [19]. Between 2008 and 2012, the health expenditures in DRC were rather low, $12–13 per capita per year [19]. A large part of the challenges of the health system in DRC originates in the political situation of three decades of non-governance resulting in the collapse of the state and economy [19].
In the big capital city of DRC, Kinshasa, adults seek medical help close to their homes when health complaints are present. During medical check-ups at primary health facilities, screening for T2D is typically done through medical history, clinical parameters (e.g. blood pressure and blood glucose) and anthropometric measurements such as body weight and height. When blood glucose is abnormal (random glucose test of > 200 mg/dL or FPG > 126 mg/dL) and/or diabetes-related symptoms are present, for instance, polyuria, polyphagia, or polydipsia, extreme fatigue or blurred vision, patients are referred, in most of the cases, to the endocrinology department of the hospital for further testing. After the diagnosis of T2D is confirmed and the physician has selected the most appropriate treatment, the national standard form for patients with T2D is completed. Developed by the Diabetes National Programme, this paper-based form collects the diabetes history of the patient, demographic details, diagnosis status, clinical evaluation, and anti-diabetic treatment records. Diagnosis of T2D is carried out by a doctor at the hospital or by a nurse at the primary health centres. If the diagnosis was performed at the hospital, from the moment the patient’s glycemia is stable, the continuation of treatment and care is decentralised to the primary health centres.
In Kinshasa, T2D care is often offered by Kin Réseau, a network initiative, which was set up in 1974 by mostly religious organisations, aiming at providing decentralised care for diabetes. This longstanding network, comprises 80 care centres, including hospitals and primary health centres, and has a patient referral system in place. This programme also offers patient education, medication for a subsidised price, and daily insulin administration at the health centre, as well as annual screening for complications of diabetes. More details on this programme can be found elsewhere [20]. Currently, Kin Réseau offers a basic diabetes package that includes one follow-up visit a week at the health centre and a bi-monthly medical examination by a trained doctor for a price of USD 3.5/month.
Data sources
The Kin Réseau network often provides T2D care by offering medical packages. The T2D care package includes disease follow-up visits in which routine measurements such as weight, blood pressure, and foot examinations are assessed. Glycemic control is monitored by measuring FPG. Based on clinical assessment and test results, the physician decides on treatment adjustments. Patients are encouraged to achieve glycemic targets recommended by the IDF of HbA1c < 7.0 % and its equivalent of FPG < 126 mg/dL [21]. The routine data gathered through these visits is collected in a paper-based follow-up form used systematically across DRC (additional file, Image S1). These forms have been collected and stored by the Centre d’éducation diabète & santé covering the years between 1991 and 2019. In the context of Kin-Antwerp’s main objective, supporting Kin Réseau in improving the quality of care and a better follow-up of patients, a database software was developed by the ITM in collaboration with staff from Memisa and the Centre d’éducation diabète & santé. After its development, staff from both Congolese institutions were trained on the use of the database and data entry and, throughout the years, the database has evolved based on data quality controls. Currently, the established electronic database is an independent effort from the Centre d’éducation diabète & santé and is limited to the information collected in the follow-up form (Image S1) of patients with T2D. Paper-based forms with information collected at 6-month follow-up visits, as this cut-off was set to indicate if a patient has not attended for disease follow-up, of patients with T2D have been entered retrospectively into the database from the latest forms received in 2019. Currently, the database contains information on approximately 13,000 patients. Data were checked for missingness and accuracy. Data-cleaning processes were carried out before the analyses and included removing irrelevant data, standardising terms and fixing typing errors, and converting data types. This study did not include any data linkage.
Study population and variables
Kin-Antwerp gathers information from 65 health centres across Kinshasa, out of which 32 centres had updated information on patients with T2D. For this study, code was developed to retrieve information based on the following eligibility criteria. Data on adults (≥ 18 years old) diagnosed with T2D were included (N = 9700; 41,353 observations). Patients’ information was excluded if any of the following variables were not available at the first visit: date of the visit (0.3%), sex (0.0%), glycemic value (1.9%), and treatment (7.8%) leaving information for a total of 8976 patients with multiple follow-up visits representing 37,548 observations. The index date was defined as the date of the first prescription of anti-diabetic medication for a patient meeting the inclusion and exclusion criteria in the database.
Demographic and clinical information described in the database included sex, age, and values at each follow-up visit for weight and height (the latter only at the first visit), and the clinical parameters of systolic and diastolic blood pressure (SBP, DBP; in mmHg) and FPG (mg/dL). Weight was measured in kilogrammes with a digital or mechanical scale placed on a firm and flat surface. Height was assessed using a measuring board positioned against a wall and taken in centimetres. For both measurements, weight and height, standard guidelines developed by each health centre in collaboration with the Centre d’éducation diabète et santé were followed. BMI was calculated as body weight in kilogrammes divided by height in metres squared and classified based on the WHO classification of adults (normal: BMI < 25 kg/m2; overweight: BMI ≥ 25 – < 30 kg/m2; obesity: BMI ≥ 30 kg/m2) [22]. SBP and DBP were measured using either a digital automatic blood pressure monitor or a sphygmomanometer, depending on the resources of the health centre. FPG was measured by pricking the skin with a lancet to obtain a drop of blood which is placed on a disposable test strip, followed by inserting it in the glucometer to estimate the glycemia in blood.
In this study, the operational definition of T2D relies on the physician’s written diagnosis by which patients were referred for follow-up in primary care centres using the standard national forms developed by the National Programme of Diabetes in DRC. Our primary outcome is the odds/probabilities of achieving glycemic target defined as FPG < 126 mg/dL. All available FPG values per patient were used to estimate the achievement of the glycemic target. Secondary outcomes entailed identifying demographic characteristics (sex and age) that could potentially influence the odds/probabilities of achieving glycemic target.
Statistical analysis
All analyses were conducted using STATA (Release 16/SE. College Station, TX: StataCorp LP). Demographic characteristics were reported as measures of central tendency for continuous data, mean ± standard deviation (SD) and medians and interquartile range (IQR) if not normally distributed, and counts and percentages for categorical variables. To assess the achievement of glycemic target (FPG < 126 mg/dL) over the years of follow-up, a multilevel mixed-effects logistic model (command melogit) was conducted. To account for patients nested within health centres, a random intercept was added for the health centres and to consider repeated measurements, we fitted a random intercept at the patient level and a random slope of the variable representing years of follow-up varying by patient. The results of this model are expressed as odds ratios (OR) and 95% confidence intervals (95% CI) and are conditional to the random effects. From this model, we derived probabilities by predicting the average marginal effects (AME) (command margins) and illustrated the average marginal effects at specific time points. AME indicates the average change in the probability, in this case, glycemic control, when x, years of follow-up, increases by one unit. We adjusted our model at baseline (first visit of follow-up) for sex, age categories (< 40 years, 40-65 years, > 65 years), BMI (normal, overweight, obesity), SBP (normal, elevated), treatment (oral glucose lowering drugs (OGLDs), insulin, insulin + OGLDs, or diet), and interactions between the years of follow-up and the mentioned covariates. These interactions are referred to as time interactions in the manuscript. We assumed the missingness mechanism was ‘missing at random’. Missing values in the covariates of the model were handled by listwise deletion in a long format, while a direct likelihood approach dealt with missing values in the outcome (i.e. the default strategy for regressions in STATA). To identify subgroups of patients that may do worse in terms of achieving the glycemic target, exploratory stratification models for sex and age categories were carried out. A p-value < 0.05 was considered statistically significant for the main model and a p-value < 0.008 for the stratified exploratory analyses after applying Bonferroni correction for multiple comparisons. As part of our objective was to assess the achievement of glycemic target, measured as odds/probabilities, over the years of follow-up, we chose a multilevel mixed-effects logistic regression, with random intercept and slope as this model allows for patient’s observations to be analysed as a cluster, hence allowing each participant to have its own starting point (intercept) and time of follow-up (slope). We favoured this approach in comparison to a survival analysis which implies selecting an event (achieving glycemic target) at a specific time, for example, time to the first or the last event. A mixed-effects logistic regression accounts for repeated evaluations of glycemic target over the follow-up time, reflecting what has happened in real practice, and taking into account the potential correlation between them, aside from also allowing adjusting the estimate for relevant covariates.