The sex distribution was similar between rural and urban individuals, with a female predominance. Although our study was not a prevalence study, differences in diabetes prevalence by sex have been reported to be variable depending on the population and setting . In some African populations, diabetes prevalence is higher in men , and in others diabetes prevalence is higher in women . In Cameroon, a similar sex distribution in diabetes prevalence has been reported in rural populations, while in urban individuals, a female predominance was noticed . This variability in the sex-specific prevalence of diabetes might be related to differences in exposure to the risk factors for diabetes by sex; for example, the Rwanda NCD risk factors survey revealed that obesity and overweight are more prevalent in women than in men . This variability implies the need for specific population-based assessments of sex differences in terms of diabetes burden for targeted and need-based interventions to address the diabetes burden.
We found that diabetes was diagnosed at a relatively young age in our study population, which was most noticeable in rural individuals. Furthermore, rural residents were younger at the time of study enrolment. The age distribution among rural residents was consistent with patterns reported in other African populations [1,31] and in a south Asian population , and the age distribution was inconsistent with the findings in Western and urban African populations, in which larger proportions of older people were found among people with diabetes [31,33]. The age distribution among rural and LMIC populations with diabetes in general might be explained by higher proportions of misclassified type 1 diabetes or so-called “malnutrition-related diabetes”, or other atypical diabetes subtypes in underserved settings for which the onset has been reported to be in the second and the third decade of life [34–36]. However, the higher proportion of diabetes in rural younger age individuals could also be explained by the poorer survival rate in more disadvantaged populations [37,38], which could lead to short life expectancy in individuals with diabetes living in poverty .
Most rural residents reported being in low-income work, having limited access to running water and electricity, more common use of herbal medicine for high blood glucose symptoms and less fruit and vegetable consumption when compared to urban residents. We were not able to describe the association of socio-economic condition with diabetes prevalence because population data were not available. At later stages of the epidemiologic transition, low socio-economic status is associated with an increased risk of NCDs such as diabetes mellitus, cancers and cardiovascular diseases . Poverty and food insecurity might contribute to the increasing prevalence of diabetes in some rural African settings in which diabetes prevalence is reported to exceed the diabetes prevalence in urban areas . In addition to the fact that poverty might contribute to the onset of diabetes potentially through foetal and childhood under-nutrition or obesity in later life, poverty is reported to be a factor related to unequal access to care . More importantly, even though there was no difference in health insurance coverage and if there were dedicated diabetes clinics in rural hospitals in Rwanda, poverty would make it more difficult for people with diabetes to keep themselves healthy. This is because of limited access to a healthy diet, electricity, a refrigerator to store insulin, and running water to keep injection sites clean as well as the ability to travel for specialist care, such as eye care. The impact of poverty and its consequences for the burden of diabetes and its complications should be explored further in low-income countries.
We found that traditional risk factors for type 2 diabetes, such as family history of diabetes, obesity and physical inactivity, were less prevalent among rural individuals. Central obesity was prevalent in both groups but was less common in rural residents. This result is consistent with the findings of other reports from LMICs in which the increasing prevalence of diabetes did not match the low prevalence of common risk factors for diabetes [32,42,43]. This finding suggests that there might be other factors contributing to the increase in diabetes prevalence in low-income settings.
We found a higher prevalence of reported childhood under-nutrition among rural than among urban residents. It has long been suggested that chronic under-nutrition is associated with impaired insulin secretion . To our knowledge, childhood under-nutrition as a risk factor for diabetes in adulthood has been given limited attention in Sub-Saharan Africa, where its prevalence is reported to be high, especially in East Africa , where evidence suggests that the prevalence of diabetes in the poorest population exceeds the prevalence in less poor populations . In urban settings, traditional risk factors remain the main drivers of the rapid increase in diabetes [27,28].
We observed an unusually high prevalence of type 1 diabetes among our study participants, particularly among rural dwellers. Furthermore, most participants, particularly those from rural areas, reported insulin requirements from diagnosis. Our finding corroborate other studies’ results in LMIC where Type 1 and type 2 diabetes have been reported to be equally common  in contrast with Western countries where type 2 diabetes is considerably more common. The over-representation of type 1 diabetes could reflect the limitation of clinic-based nature of the study; people with type 2 diabetes might have participated in fewer clinic visits or received their care in other settings. In our case, people with diabetes are given appointments to attend the NCD clinics on a monthly basis for prescription renewal and follow up, with active retrieval of those who were lost to follow-up, regardless of the type of diabetes. Furthermore, we recruited participants in various health facilities on different diabetes clinic days over a whole year to overcome potential selection bias. Diabetes classification is usually based on clinical presentations in our clinics, and atypical diabetes with type 1-like phenotypes such as MRDM and ketosis-prone type 2 diabetes could have been misclassified as type 1 diabetes. This misclassification may have negative impact on the necessity to understand the aetio-pathology of the above atypical phenotypes and on the decision making for treatment and prevention.
Although our study population was uniformly well covered by medical insurance, more rural individuals than urban participants reported severe hyperglycaemia at diagnosis and use of herbal medicine, and their diabetes was less well controlled. Limited access to diabetes care, easy accessibility to traditional healers, lack of resources and frequent lack of stock of modern diabetes drugs have been reported to be the reasons for herbal medicine use and poor quality of diabetes care in LMICs [47,48]. There is a need to identify other barriers to quality diabetes care in the setting in which universal medical coverage is maximized and diabetes care decentralization to lower levels of the health system is established to improve equitable access to care.
The characteristics of people with diabetes in rural settings, such as low socio-economic conditions, young age of onset, leanness, lower prevalence of traditional risk factors for type 2 diabetes, higher prevalence of reported childhood under-nutrition and an unusually high prevalence of type 1 diabetes reflect the challenges facing healthcare providers in diabetes diagnosis, classification and clinical care decision making in Africa. Under-nutrition and over-nutrition as well as poverty might play an important role in the burden of diabetes in LMICs. Lean, young individuals with a history of childhood under nutrition and without classic risk factors for type 2 diabetes or classical features of type 1 diabetes do not fit any type of the diabetes classes mentioned in 1999 World Health Organization (WHO) diabetes classification although could be assigned to the unclassified group in the 2019 updated classification .
Further studies are required to assess risk factors for sub-types of diabetes and their aetio-pathology in rural and low-income settings and to identify effective interventions to inform guidelines to prevent and treat all forms of diabetes in LMICs. There is a particular need to establish the characteristics and burden of MRDM and other atypical types of diabetes and appropriate approaches to their primary and secondary prevention.