The burden of disease profile of residents of Nairobi's slums: Results from a Demographic Surveillance System

Africa is urbanizing at a faster rate than any other region in the world and by 2030, more than half of the population of sub-Saharan Africa will live in urban areas [1]. The rapid growth of urban centres in many sub-Saharan African countries has occurred largely in an environment of poor economic performance and lack of urban planning and regulation. This has resulted in an increase in the number and size of informal settlements or slums in many cities. Current estimates from the UN HABITAT suggest that more than 70% of urban residents in sub-Saharan Africa live in slum or slum-like conditions. In Kenya, this percentage is about 71% [2]. Because of the informal nature of these settlements, they are underserved by health, education, water and sanitation, and garbage collection services.

Apart from lack of social amenities, slums are also characterized by high unemployment, overcrowding, insecurity, greater involvement in risky sexual practices, social fragmentation, and high levels of mobility [3‐5]. As a result, slums have become the new hubs of poverty and ill health. Studies from different sub-Saharan African countries have shown that slum residents have worse health indicators than their counterparts in rural areas [5‐9]. For example, children under the age of five years living in urban poor areas in Zambia and Malawi had higher mortality than those in rural or peri-urban areas [10, 11]. The worse health status of slum children could be explained by the continuous exposure to environmental hazards coupled with lack of social services. Desperate living conditions and lack of livelihood opportunities could predispose residents to risky health-related behaviours such as high alcohol consumption, unsafe sex, smoking and other substance abuse. In the short and medium term, these behaviours may lead to higher mortality from external causes and communicable diseases such as HIV/AIDS, and in the long term – to higher risk of non-communicable diseases.

In order to plan for and address the health needs of a population, information is needed about its health status. However, information for health planning is largely unavailable in most developing countries partly due to lack of vital registration. Other data on health status are often derived from small hospital-based studies; yet most illnesses and deaths occur outside the health care system. Estimating the burden of disease in a population is one way to generate health information for planning. Burden of disease projects have been undertaken and estimates have been made for the whole world [12, 13] and world regions including sub-Saharan Africa [12, 14]. While these estimates are useful for making national and international policy decisions, they mask inter-country and intra-country differences and may not be relevant for local health planning [15]. The situation is even worse for marginalized populations such as the urban poor, which have hitherto not received much attention in terms of research and service delivery. A Burden of Disease (BoD) approach for district and national planning has been successfully used in Tanzania with a positive impact on childhood mortality [16].

This paper therefore aims to describe the causes and quantify the burden of mortality among residents of two Nairobi slums, using a BoD approach with data generated from a Demographic Surveillance System (DSS).

We analysed data collected by the longitudinal surveillance system of the NUHDSS which covers a population of about 56,000 residents in two slums. The population has a high disease burden especially among children under the age of five years. Pneumonia, diarrhoeal diseases and stillbirths account for more than half the burden among children under five, while HIV/AIDS, tuberculosis, interpersonal violence injuries and road traffic accidents account for more than two thirds of the burden among people aged five years and older.

The age structure of the study population is remarkably similar to the population structure of Nairobi province [20] and is characteristic of a predominantly labour migrant population which has more people in the productive age groups most likely accompanied by some of their younger offspring. The distribution of YLL by age group reflects this age structure. However, the large shares of the mortality burden borne by children under five and the 15–49 years age group reflect the effect of childhood illnesses, maternal causes and HIV/AIDS in these age groups. The underlying age structure is controlled for by calculating the age shares per capita or the mortality burden per 1,000 person years. This shows the tremendous mortality burden among children under the age of five years compared to the rest of the population, as well as the expected high burden among the elderly. The findings of a high mortality burden among the study population reinforce the findings from previous studies that have shown that the urban poor have poor health status, which in some instances, is worse than that of rural dwellers. These studies have however only shown this through high childhood mortality [10, 11]. Since there are no similar estimates for other parts of Kenya, we compared our findings with estimates for the Tanzania coastal sentinel DSS site in 2003.

Estimates from the coastal DSS sentinel site in Tanzania using the same approach show a total per capita burden of 181 YLL/1,000 person years in 2003 (unpublished data) compared to 205 YLL/1,000 person years in our study population. Despite the fact that the burden in our population aged over five years is high (160 YLL/1,000 person years compared to 133 YLL/1,000 person years in the Tanzanian site), children under the age of five years still have more than four times the mortality burden of the rest of the population. In comparison, estimates from the DSS sentinel site in Tanzania show that children under five have about 3.6 times the burden of the rest of the population. Therefore, not only does our study population have a higher mortality burden than the Tanzanian DSS site, children in Nairobi slums have an even greater mortality disadvantage, compared to the rest of the population, than their Tanzanian counterparts. This disadvantage is also reflected in the infant and childhood mortality estimates from the Tanzania coastal DSS site, which are much lower that the estimates from our study population. In 2003, the infant mortality rate (1q0) in the coastal DSS site was 46.3, while the under-five mortality ratio (5q0) was 75.4 per 1,000 live births [21] compared to 96 and 139 respectively in our study population.

Perinatal causes account for 28% of under-five YLL which is rather high. According to estimates of global childhood mortality, in countries with high under-five mortality rates, neonatal causes account for about 20% of all childhood deaths [22, 23]. Our study population is in a country with high childhood mortality (115 per 1,000 live births) [24], and estimates for the slums of Nairobi and the NUHDSS slums show an under-five mortality rate(151 and 139 per 1,000 live births respectively) higher than the national estimate [5]. The finding of a high proportion of neonatal deaths (33%, including stillbirths) therefore shows the increased vulnerability of children around the time of birth and is a reflection of the lack of good quality antenatal and delivery services in the study area [25]. We chose to include stillbirths in the YLL calculation in order to highlight their substantial contribution to the poor health status of the population.

The top ten causes of mortality among children under the age of five years in our study population are similar to the top ten causes of mortality for sub-Saharan Africa, but the order is different. Malaria plays a less significant role in the study area than in other parts of sub-Saharan Africa. This is because of the high altitude and cold climate in Nairobi which are unsuitable for malaria transmission. The small contribution of malaria is however in contrast to its position as the second most common diagnosis in health facilities in Nairobi province based on routine health information system reports [26], and the single most common self-reported illness in the study population [27].

The significant role played by pneumonia is unsurprising given the poor housing and overcrowding conditions in the slums. Another contributing factor may be the common practice of using the same room for cooking and sleeping coupled with the widespread use of kerosene for cooking (APHRC, unpublished data). A morbidity survey in 2000 showed that 46% of children aged below 36 months living in the slums of Nairobi had had cough in the two weeks preceding the survey, while 32% had had diarrhoea [28]. Our findings therefore reinforce those from the morbidity survey which show the relatively high importance of these two conditions in influencing the health of slum children. On the other hand, given the high mortality burden from AIDS in the adult population, it is possible that some paediatric HIV-related deaths may be misclassified as diarrhoeal diseases or pneumonia, as has been the case in other studies [29]. In most cases, there are no death certificates to complement the oral interview. Most people die outside the formal health care system and the tool does not capture HIV status in any other way except what is volunteered by the respondent. The verbal autopsy approach therefore has limitations in this case and is unable to capture the likely high proportion of AIDS-related deaths.

Malnutrition contributes less than 10% of the YLL among children under the age of five years. This may be because, in the verbal autopsy approach, respondents are more likely to report the more dramatic events surrounding the death of a child such as diarrhoea and cough, rather than the more insidious underlying malnutrition. It is therefore likely that the malnutrition burden is underestimated in the study population since it is known that malnutrition is an underlying cause for more than half of childhood deaths. Given the levels of poverty and deprivation in the slums, one would expect malnutrition to play a more significant role than we found. On the other hand, although malnutrition may be prevalent, it may not necessarily contribute directly to mortality since chronic types of malnutrition are rarely fatal.

Among the population aged 5+ years and above, the devastating effect of the HIV/AIDS epidemic is evident in the large contribution of AIDS and tuberculosis to the mortality burden. As previously indicated in the results section, these two illnesses were combined in the analysis because close to one third of deaths from these two causes could be attributed to both conditions, i.e. the code "AIDS and Pulmonary TB" and some deaths (seven) were classified as "unspecified AIDS/TB." It is also known that most TB cases in sub-Saharan Africa are related to underlying HIV/AIDS and that TB is a common opportunistic infection in HIV-related deaths in Nairobi [30]. Teasing one from the other without additional information such as medical records is difficult, hence the decision to combine the two. While this masks the role of other opportunistic infections, it still highlights the tremendous impact which both conditions have on the study population. Estimates from sub-Saharan Africa show that these two conditions contribute only about 30% of the YLL burden [31]. While it is possible that the YLL burden may have increased since 2002 when these estimates were calculated, our study suggests that slum residents in Nairobi have been hit by the HIV/AIDS epidemic in perhaps greater terms than other subpopulations in sub-Saharan Africa.

Injuries due to interpersonal violence are the second leading contributor to the mortality burden in the population aged five years and older. The contribution of all injuries to the overall burden in the whole population is also higher than in sub-Saharan Africa. The most common modes of injury are gunshot wounds, road traffic accidents and blunt trauma during mob justice. The modes of injury reflect the high levels of insecurity and violence in the population. Informal settlements are characterized by living and social conditions which are known risk factors for violence. At the society level, they include reduced inhibitions against violence, and the creation and sustenance of gaps between different segments of society [32]. At the community level, social disintegration, high residential mobility, high population density, heterogeneity and lack of social cohesion play a role. At the individual level, low educational attainment, drug and alcohol abuse, and unemployment are important [32]. Most, if not all, of these factors are present in informal settlements and hence the high level of interpersonal violence is unsurprising. Additionally, the marginalization of residents in informal settlements means that they are rarely served by social institutions such as the police and justice system. Perceived indifference of such institutions to injustice in slum communities often results in community measures which usually take the form of "mob justice" meted out to people suspected of committing crimes, even petty ones.

The high contribution of road traffic accidents to the mortality burden reflects the growing importance of this cause in sub-Saharan Africa [33], as well as the effect of rapid urbanization. Growing traffic volumes accompanied by poorly planned cities, poor state of roads and ineffective law enforcement are known factors which contribute to the increasing burden of road traffic accidents in sub-Saharan Africa, and Nairobi is no exception.

Our study's strength is in its longitudinal nature which enabled us to follow up a large population, record deaths soon after they occurred, have an accurate reference population and the time at risk, and conduct verbal autopsies in the majority of cases. We were therefore able to calculate both the overall mortality burden and the age-specific burden per capita which identifies subgroups at higher risk than others.

The verbal autopsy approach fills the gap left by the absence of vital registration systems and other cause of death ascertainment mechanisms despite its shortcomings. Since the majority of deaths occur outside the formal health sector, we could not use official death certificates, more so since the quality of the existing system in Kenya is unknown.

The sensitivity of the verbal autopsy is however dependent on how prevalent a certain disease is [34]. Causes of death like malaria and AIDS are more likely to be assigned by coders in areas with a high prevalence of the illness compared to coders in areas with low prevalence. Conversely the same causes are less likely to be assigned where the prevalence is truly or perceived to be low. The sensitivity of the verbal autopsy also depends on how distinct a disease's signs and symptoms are compared to other diseases. Diseases with common symptoms such as fever may be more frequently misclassified compared to those with distinct symptoms such as a stiff neck [17]. Misclassification errors are even more common in coding childhood deaths. This is because children often have more than one illness and it may be difficult to determine which of these the main cause of death is. While these misclassification errors may not affect the overall YLL estimates, they may lead to underestimation of YLL contributed by specific causes. Our inability to assign causes of death in about 7% of deaths is a limitation of the study which was mitigated by redistributing these deaths to other cause of death categories.

The high disease burden among slum residents, especially among children under the age of five years, the high burden of fatal injuries, and the devastating effect of HIV/AIDS in a context of rapid growth of urban settlements and high proportion of poor people in these settlements raise serious challenges for governments in sub-Saharan Africa. Large subpopulations with high disease burden may threaten progress towards national and global development goals such as the Millennium Development Goals, especially those aimed at reducing childhood mortality and the impact of the HIV/AIDS epidemic, unless drastic steps are taken to address the health and social needs of the urban poor. The successful utilization of a BoD approach for district and national planning in Tanzania and its impact on child mortality offer a way out for marginalized slum dwellers in Nairobi City. Based on data from the NUHDSS, health planners could adopt a similar planning approach that would benefit slum dwellers.