Bulletin of the World Health Organization

Local-level mortality surveillance in resource-limited settings: a case study of Cape Town highlights disparities in health

Pam Groenewald a, Debbie Bradshaw a, Johann Daniels b, Nesbert Zinyakatira a, Richard Matzopoulos c, David Bourne c, Najma Shaikh d & Tracey Naledi d

a. Burden of Disease Research Unit, South African Medical Research Council, PO Box 19070, Tygerberg, 7505, South Africa.
b. Information and Technology, City of Cape Town Health Department, Cape Town, South Africa.
c. School of Public Health and Family Medicine, University of Cape Town, Cape Town, South Africa.
d. Department of Health, Provincial Government of Western Cape, Cape Town, South Africa.

Correspondence to Pam Groenewald (e-mail: prenaud@compnet.co.za).

(Submitted: 04 July 2009 – Revised version received: 27 October 2009 – Accepted: 02 November 2009 – Published online: 08 January 2010.)

Bulletin of the World Health Organization 2010;88:444-451. doi: 10.2471/BLT.09.069435


It is increasingly recognized that cause-of-death data are an essential component of health information systems.1,2 Mortality data are required to identify the health needs of a community, monitor progress in the implementation of programmes and track changes over time.3 Numerous authors have highlighted the need to strengthen national statistics systems and improve the comparability of mortality data and of international collection and reporting of data.48 While the importance of such data at national level is well established, recent evidence of heterogeneous mortality patterns within countries and districts911 – and the health inequities that they reflect – suggests that it is also important to compile such data at subnational level.12 In contexts of resource constraints and health disparities, subpopulation data are critical for identifying and monitoring inequalities in health status and for prioritizing interventions, services and research at the local level. Consistent and reliable cause-specific data are also essential for evidence-based health planning.13,14

Few low- or middle-income countries have national cause-of-death data, and when they do, the quality of the data is often questionable.2,4 The official mortality statistics produced by Statistics South Africa are classified as low-quality owing to incomplete reporting and large proportions of deaths attributed to ill-defined causes and injury deaths with undetermined intent.2 Furthermore, since 1997 these statistics have only been collected for the national and provincial levels. Cause-of-death profiles for the local level are generally unavailable.15 The City of Cape Town, however, has collected cause-of-death statistics for more than 100 years as part of its public health programme.16 During the transition of the local government from 6 municipalities to a single municipality with 8 health subdistricts in the late 1990s, the method of collecting and collating statistics was reviewed and enhanced through training and standardization.17 Since 2000, statistics have been analysed to subdistrict level with support from the South African Medical Research Council.18 In 2003, the first subdistrict analysis was carried out and the findings revealed wide differentials at the local level. On the basis of these findings, trend analysis was done for the Cape Town Metro district for 2001–2006 and for the 8 new subdistricts for 2003–2006.

This article presents the key findings of the trend analysis and demonstrates how local mortality surveillance can identify the major causes of premature mortality and highlight the differential health needs of subpopulations. The findings were particularly relevant for health managers in Cape Town, who were decentralizing health services and developing subdistrict health plans at the time of the study.


Cape Town, one of 6 major metropolitan areas in South Africa, has a population of approximately 3.5 million people. Since 2003 the city has been divided into eight health subdistricts. The City of Cape Town has a well established system for routinely compiling cause-of-death statistics. The Cape Town Health Department collates data from copies of death notifications from the local offices of the Department of Home Affairs and collects data from local mortuaries. Cause-of-death coding and identification of underlying cause is done by trained clerks at the Health Department using a shortlist based on the International statistical classification of diseases and related health problems, 10th Revision (ICD-10) (table 2 in Bradshaw et al.)17 This list includes the most prevalent conditions in Cape Town, as well as diseases of public health importance.

For this study, interpretive coding was used where euphemisms for HIV/AIDS such as “retroviral disease,” “RVD”, “immunodeficiency” and “immunocompromised,” were written on the death certificate. These causes were coded to HIV as the underlying cause. For unnatural deaths, where the manner of death was missing from the death certificate, the relevant mortuary records were identified using personal identifiers such as name and identity number to obtain this information.

After data cleaning and exclusion of duplicate records by means of personal identifiers, the shortlist cause-of-death codes were aggregated into three main cause groups according to the South African National Burden of Disease Study, based on an adapted version of the 1990 Global Burden of Disease Study.19 These groups were: group 1: pre-transitional causes (communicable diseases, maternal causes, perinatal conditions and nutritional deficiencies); group 2: noncommunicable diseases; group 3: injuries. HIV/AIDS is part of group 1 but statistics on the disease are recorded separately because it accounts for such a large share of the disease burden in South Africa. Deaths attributed to ill-defined causes were redistributed proportionally by age and sex across the appropriate cause groups as described in a previous report.18

The data were analysed by age, cause and sex for the Cape Town Metro district from 2001 to 2006 and for the 8 new health subdistricts from 2003 to 2006. Completeness of death statistics was assessed through a comparison with estimates of the number of deaths per household from the 2001 South African census, as described by Zinyakatira.20 During the period 2001–2006 completeness of death registration for adults was estimated to be 96%, except in 2005 when it was 84%, owing to a problem with collection of death notifications from some offices. Data for 2005 were therefore excluded from further analysis. Completeness for children is difficult to assess but is assumed to be lower than for adults.

Premature mortality was calculated in years of life lost (YLLs) using the Global Burden of Disease approach, with age-weighting, discounting of 3% per year and standard life expectancies based on the West model life table,21 with an average expectancy at birth of 80 years for men and 82.5 years for women.19 These assumptions were chosen to allow for national and international comparison. Population estimates for the Cape Town Metro district were projected by means of the Actuarial Society of South Africa (ASSA) 2003 model22 and were used to calculate age-specific rates per 100 000 population. Those rates were then age-standardized using the World Health Organization (WHO) world population standard23 and compared across subdistricts. Mortality differentials by subdistrict and sex were assessed using 95% confidence intervals around the age-standardized mortality rates.24


A total of 25 251 deaths was recorded in the Cape Town Metro district in 2006 – an increase of about 7.0% with respect to the number in 2001 (23 681). Natural deaths from ill-defined causes accounted for 7.7% of total deaths analysed in 2001 and 6.0% in 2006. Among unnatural deaths, deaths due to undetermined causes (i.e. where intentionality could not be determined) increased significantly from 5.7% to 11.0% over the period (χ2 = 78.0: P < 0.001).

The broad pattern of mortality by age and cause in Cape Town Metro district in 2006 is shown in Fig. 1 and Fig. 2. There were considerable sex differences, with young adult males experiencing much higher numbers of deaths from injury than females in absolute terms. HIV/AIDS deaths were common among young adults and to some extent children. Noncommunicable diseases accounted for large numbers of deaths over the age of 40 years. In age groups over 75 years, there were more deaths among females.

Fig. 1. Age distribution of female deaths (n = 11 132) by cause and age group, Cape Town Metro district, South Africa, 2006
Fig. 1. Age distribution of female deaths (<em>n</em> = 11 132) by cause and age group, Cape Town Metro district, South Africa, 2006
Fig. 2. Age distribution of male deaths (n = 14 119) by cause and age group, Cape Town Metro district, South Africa, 2006
Fig. 2. Age distribution of male deaths (<em>n</em> = 14 119) by cause and age group, Cape Town Metro district, South Africa, 2006

The overall age-standardized mortality rate was 1011 per 100 000 (95% confidence interval, CI: 983–1039) in 2001 (males, 1263, 95% CI: 1241–1286; females, 799, 95% CI: 783–815). The rate decreased slightly, to 939 per 100 000 (95% CI: 941–964), in 2006 and male mortality remained significantly higher (males, 1159, 95% CI: 1140–1180; females, 756, 95% CI: 742–771). During this period there was a significant increase in deaths from HIV/AIDS and a significant decrease in deaths from noncommunicable diseases among both males and females and in injury deaths among males (Table 1).

Significant differences in age-standardized mortality rates were observed between health subdistricts within the Cape Town Metro district (Table 2). The lowest rates were in the Northern and Southern subdistricts (679 and 713 per 100 000, respectively) and the highest in Khayelitsha subdistrict (1619 per 100 000), where the rates were almost 2.5 times higher than in Northern (Table 2). Khayelitsha had the highest mortality rates for all cause groups.

Age-standardized death rates for selected noncommunicable conditions showed marked variations by subdistrict (Table 3). Ischaemic heart disease mortality rates were very high in Eastern and Tygerberg but low in Khayelitsha, while stroke, hypertension and diabetes mortality rates were very high in Khayelitsha and Mitchell’s Plain. Lung cancer and chronic obstructive pulmonary disease death rates were higher among males than females and highest among males in Tygerberg and Mitchell’s Plain. However, there was variability in these rates, the most extreme variation being ± 30% (hypertensive heart disease death rates in males in the Northern subdistrict ranged between 9.5 and 17.5 per 100 000, with the point estimate being 13.5).

Trends in the leading causes of premature deaths (presented as a percentage of total YLLs) over the period 2001–2006 are shown in Fig. 3. The relative contribution of violent deaths to premature mortality increased between 2001 and 2002 but declined between 2002 and 2004. However, violent deaths increased again in 2006, reaching 2003 levels. Deaths due to HIV/AIDS increased markedly between 2001 and 2003, levelled off in 2004 and 2005 and declined in 2006. The trends for HIV/AIDS, homicide and tuberculosis death rates are similar to the trends in percentage of total YLLs for these causes, although it is not possible to estimate rates for 2005 owing to incomplete data. HIV/AIDS has now replaced violence as the leading cause of death. The four leading causes in the Cape Town Metro district – namely, homicide, HIV/AIDS, tuberculosis and road traffic injuries – accounted for 44.2% of all premature mortality in 2006. Table 4 shows the ranking of conditions based on premature mortality (YLLs) for each health subdistrict in 2006.

Fig. 3. Trend in percentage years of life lost (YLLs) for selected leading causes of death, Cape Town Metro district, South Africa, 2001–2006
Fig. 3. Trend in percentage years of life lost (YLLs) for selected leading causes of death, Cape Town Metro district, South Africa, 2001–2006


The above data show that although mortality rates are lower than the national average, the pattern of mortality observed in Cape Town reflects the quadruple burden of disease observed in the national cause-of-death profile,25 with high infectious disease mortality among young children; high mortality from violence and injuries among young adults; high noncommunicable disease mortality in older age groups; and rising HIV/AIDS mortality among young adults and young children. The use of premature mortality (YLLs) facilitates the identification of public health priorities such as HIV/AIDS and violence. Analysis of emerging trends in mortality between 2001 and 2006 shows that HIV/AIDS has overtaken homicide as the leading cause of premature mortality, partly as a result of a decline in homicide rates during the analysis period but also as a result of an increase in HIV/AIDS mortality rates.

Despite interpretive coding, the observed HIV/AIDS mortality rates are likely to be understated as there is a tendency among doctors to certify only the immediate causes of death to avoid disclosing HIV status. Nonetheless, HIV/AIDS was either the first or second most common cause of premature mortality in all subdistricts in 2006 and accounted for 16% of the total burden of mortality in the Cape Town Metro district. Mortality from HIV/AIDS followed a distinct age pattern that has been consistently observed,26,27 with the majority of deaths concentrated among young children, women aged 25–39 and men aged 30–49. While the peak mortality rates for males and females were similar in magnitude, the peak occurred approximately 10 years earlier among women. The rapid increase in HIV/AIDS mortality between 2001 and 2004 appears to have levelled off, possibly demonstrating the impact of antiretroviral programmes implemented in the Western Cape in 2003.

In contrast with the rising trend in tuberculosis mortality seen nationally, tuberculosis death rates in Cape Town remained fairly constant over the period. Several factors might account for this. First, a large proportion of tuberculosis deaths at the national level are misclassified as HIV/AIDS deaths,26 but this occurs less frequently in Cape Town because of the interpretive coding used. In addition, in recent years control efforts have been intensified in high-tuberculosis areas such as Khayelitsha, which has led to vast improvements in cure rates and thus probably also reduced mortality rates.

Although Cape Town homicide mortality rates declined between 2001 and 2004, they increased thereafter and remain among the highest in the world (59.8 per 100 000).28 Homicide mortality rates among males in Cape Town in 2006 were almost 8 times the global average (109 per 100 000 versus 13.9 per 100 000), and those for females 3 times the global average (12.5 per 100 000 versus 4.2 per 100 000).29 Data from the National Injury Mortality Surveillance System show that age-standardized homicide rates in Cape Town were also higher than in other South African cities.30

Homicide occurs most frequently among young adults. The rates among males were 2–4 per 100 000 in childhood and peaked sharply, at about 260 per 100 000, in the 15–24 year age group. The rates then declined steadily with age. A similar age pattern was seen among females except that the peak was lower (24 per 100 000) and occurred about 10 years later, between 25 and 34 years of age. Homicide rates were highest in the Khayelitsha subdistrict, where the rate for males (242.4 per 100 000) was 17 times the global average and that for females (36.6 per 100 000) almost 9 times the global average. Overall, our data show that approximately 40% of homicides in the Cape Town Metro district are committed with firearms.

Unlike homicide rates, deaths from road traffic injuries remained fairly constant in Cape Town between 2001 and 2006, ranging from 25 to 29 deaths per 100 000. However, road traffic fatality rates for males in Cape Town are 30% higher than the global average for males (45 per 100 000 versus 32 per 100 000).29 Prevention of injuries, particularly violent injuries, thus remains a pressing provincial priority that will require intersectoral and multilevel coordination.31

The wide differentials in levels of mortality across the city highlight the importance of subdistrict-level information. Khayelitsha subdistrict, a large, impoverished township on the outskirts of Cape Town, had a considerably higher burden of premature mortality for all three main cause groups in comparison to the other subdistricts. While this difference was particularly marked for pre-transitional conditions, HIV/AIDS and injuries, noncommunicable disease rates were also highest in Khayelitsha, which confirms a previous finding that noncommunicable diseases are prevalent among the urban poor in South Africa.32 Of the cardiovascular diseases, only ischaemic heart disease rates were lowest in Khayelitsha. Hypertensive disease, stroke and diabetes mortality rates in Khayelitsha were among the highest, however, which suggests that this population is in an earlier phase of the cardiovascular epidemiological transition than populations in other subdistricts. Rates of death due to tobacco-related conditions such as lung cancer and chronic obstructive pulmonary disease were also higher among males in Khayelitsha than in the Southern, Northern and Western subdistricts.

There are some concerns about the quality of the data used in this study. A recent study of cause-of-death certification revealed errors that could affect identification of the underlying cause of death,33 in particular failure to specify an underlying cause, which results in the attribution of deaths to ill-defined causes. In 2006, 6.0% of Cape Town Metro district deaths were attributed to ill-defined causes. This was substantially lower than the 12.0% observed in the national data for 2005,34and is only slightly higher than the international target of 5.0%.19 Nevertheless, such data are not useful and were therefore redistributed to defined causes for purposes of this study. Because the ill-defined deaths accounted for a relatively small percentage of the total and occurred across all age groups, this redistribution did not influence the results significantly.

The relatively low percentage of deaths attributed to ill-defined causes and the relatively high level of data completeness in Cape Town may reflect access to better health services than in other districts. The completeness of death registration may also be a result of the system of triangulation of mortality data from various sources, although lower completeness for child deaths may imply underreporting of diarrhoea and lower respiratory infections, which may have affected the data on leading causes of premature mortality, particularly in Khayelitsha. In any case, improvement in the quality of cause-of-death certification is needed to enhance the reliability of cause-of-death statistics. This need is currently being addressed through a collaborative initiative to institutionalize death certification training at medical schools, training hospitals and other medical facilities. In addition, the surveillance system is being enhanced to allow for electronic transfer of cause-of-death data from mortuaries. This will provide full information on the manner of death for injury fatalities, which is usually missing from death certificates. While the use of a shortlist for coding was expedient in the early phases of the surveillance system, this approach is not ideal and the system is currently being upgraded to make use of automated ICD-10 coding, in line with international standards.

A challenge in establishing an effective mortality surveillance system and assessing trends in mortality rates in small areas is the need for accurate population data. In South Africa, no consistent official population estimates are available for subdistricts. We used estimates based on the ASSA model, which were calibrated to the 1996 and 2001 censuses. Migration may not have been fully captured, however, and mortality trends may therefore have been influenced by errors in the population figures. Nevertheless, since the rates do not show systematic trends across all causes of deaths, such an error is unlikely to have had a major effect.

Despite its limitations, the Cape Town routine local mortality surveillance system provides a wealth of information on the health of the population that is useful for identifying priority health problems and vulnerable groups and for planning and implementing targeted interventions. For example, data from 2006 show that noncommunicable diseases accounted for 55.8% of deaths, the majority from cancers, cardiovascular and respiratory disease and diabetes mellitus. From a health service perspective, noncommunicable disease mortality rates can be reduced through health promotion and improved risk factor management at the primary care level. This has important implications for existing primary health care systems, which were designed mainly to manage acute infectious diseases.35

While macro-level interventions are needed, community-based interventions, such as promotion of a healthy diet and regular exercise, aimed at reducing chronic disease risk factors at population level are also essential.36 In particular, tobacco control efforts should be strengthened to reduce high smoking rates among men of all ethnic groups and women of colour.37 Intersectoral coordination is also needed. The Health Department needs to engage sectors such as education, housing, transport, safety and security, and traffic control in addressing multifactorial problems such as HIV infection, tuberculosis, homicide and road traffic injuries, which together accounted for 44% of premature mortality in the Cape Town Metro district.

Dissemination of mortality information to relevant stakeholders is of the utmost importance. To that end summaries of key findings38 have been produced and disseminated, and presentations have been made to health policy-makers and programme managers at various levels. It would be ideal to link the mortality surveillance system to the national statistical system. However, this needs to be done without losing the valuable ownership of the information by the local health officials who need to use it to allocate resources.

In a limited-resource setting, prioritization of health interventions is essential. Local level mortality surveillance can help by identifying the leading causes of premature death and the subpopulations suffering the highest levels of premature mortality. It can also provide useful data for monitoring the effectiveness of programmes and interventions. The Cape Town experience has demonstrated that it is possible through concerted collaborative effort to implement and strengthen mortality surveillance systems, and that such systems provide useful information for health planning and policy-making, particularly for urban poor communities.

Competing interests:

None declared.