The risk of death of older persons in differing populations is generally dependent on specific pre-existing conditions such as cognitive impairment, depression, loss of physical function, poor nutritional status, loneliness and limited social networks [1–6]. Physiological decline and diminished physical capacity invariably accompany ageing in all societies. Functional decline and social isolation increase vulnerability in old age, and the ensuing neuro-endocrine stress responses can precipitate a further decline in health [2]. Determinants of mortality in old age in Western societies are well documented [7–10]. In Africa the determinants of mortality in old age have not been investigated. It could be argued that mortality and its predictors in older Africans would differ little from what is already known in the West, but this would be to discard different disease patterns, and cultural and lifestyle influences. Non-communicable diseases of old age are frequently discarded in development of public health strategies of developing countries; yet non-communicable diseases, mental health problems included, are among the most important causes of sickness, disability and premature mortality in these settings [11, 12].

Cognitive impairment has been found in 9 % of older Botswana subjects (aged 60 years and over): this increased with age. 12 % of older Batswana were also categorised as 'hazardous' drinkers whilst 7 % were depressed [13, 14]. Mental disorders of old age such as dementia and depression contribute in turn to diminished physical function [15].

In the developed world population ageing and the loss of function in older citizens is addressed through the provision of targeted systems of health and social care. Social care systems may differ between nations, but require the financial support of Governments; usually to provide a mix of cash transfers and formal care services, such as pension schemes, meals, home based care and ultimately institutionalisation [16].

By contrast, the long held belief that the informal, but 'traditional' support from the extended family in African countries such as Botswana, has allowed a presumption that this can be depended upon for provision of care to older persons. Perhaps unsurprisingly formalised care is uncommon or often non-existent in Africa; indeed Botswana is one of only a handful out of 54 African countries with some form of old age pension programme [17]. Older persons in Botswana are entitled to the same level of public health care as any of its citizens, yet no old-age-specific health care facilities exist to meet this requirement. There are no personnel specialising in geriatric care in the country, and in 1999 there were 82 psychiatric nurses out of the total of 5000 registered nurses [18]. A decentralised general primary health care system has recently been created, with the goal of ensuring that a local clinic, run by nurses, is within 15 km (considered to be within walking distance) of all citizens, the aged included. Current socio-demographic trends and socio-medical pressures such as those caused by HIV/AIDS, urbanisation, and labour migration are contributing to the dissolution of the 'extended family' and thus contribute to the isolation and marginalisation of older Africans. At the same time, the burden older persons also carry for the provision of care to their dependants increases paradoxically [19]. Older Africans are compelled to care for their HIV infected children and eventually their orphaned grandchildren. Older persons' increasing contributions to civic stability and development in Sub-Saharan Africa cannot be ignored in a context of population ageing which will gain momentum in Africa over the next decades, accompanied inevitably by the increasing burden of non-communicable diseases and mental disorders of old age [20].

Informal care, social support, mental disorder and functional capacity require further investigation as possible predictors of premature mortality in older Africans [13, 21, 22].

This study assesses mortality rates and defines predictors of short survival in a national longitudinal sample of community dwellers aged 60 years and older in Botswana.

The median age was 71 years for both the baseline and follow-up population.

The median number of people in the household was 6 in both the baseline and follow-up populations. Living alone is not rare as 8% live unaccompanied and a total of 20 % live alone or with only one other person in the household.

The majority of subjects lived in rural areas and more than half had never attended school.

About 50 % were married or lived with a partner; whereas the other half were single, dominated by widowed persons (2/3 of single were widowed).

During follow-up 249 subjects were located alive and interviewed, and 16 subjects were confirmed to have died by family members or neighbours, totalling to 265 subjects.

Age distribution of crude probability of death among those with confirmed information at follow-up was 6 % in total; (3.6% (60–69 yrs); 5.9% (70–79 yrs); and 10.1% (80+ yrs). A more conservative estimate of crude probability of death (n = 16) during follow-up based on the total baseline sample (n = 372), assuming all deaths are registered, was 4.3% (2.4% (60–69 yrs); 4.5% (70–79 yrs) and 7.5% (80+ yrs).

Total observation time at follow-up was 1762 months (n = 265). Overall mortality of the follow-up sample combined is 10.9 per 100 person years.

Age adjusted risks for death during follow-up are, as a trend, lower, but still similar to the calculated crude odds ratios' and the confidence intervals are relatively wide due to the low numbers of deceased.

Cognitive impairment and physical functional decline and limited social support are associated significantly with increase in risk of death in the short term (follow-up period was on average 6.8 months). Depression shows a non-significant trend of increased risk of mortality.

At least 20 % of older persons in Botswana live in relative isolation. This renders false the notion that older Africans are always taken care of by their extended family. The combined effects of absence of "helping hands" in old age and increasing physical and mental frailty increase the risk of a premature death.

Community dwelling older persons in Botswana with cognitive impairment, reduced lower extremity strength, solitary eating and/or living in a household alone or with only one other person have a significantly increased short term risk of death. Attention to critical functional decline as a precursor of rapid death has also previously been reported in Zimbabwe, but no specific determinants for this association were described [29].

Estimated mortality at nearly 11 per 100 person years for the entire study population is more than twice the mortality found among age-peers in several European countries [10].

Older persons in Western societies who eat alone have been found to be at nutritional risk [30, 31], and this is also the case with older persons in Botswana [25]. Eating patterns are indicative of social interaction (which can be seen as proxy indicator for support) and most elders share meals with others if possible [32]. Older persons in Botswana are left increasingly with no one to share their meals and hence "eating alone" may be seen as an indicator of social isolation with its attendant risks.

The shorter survival of older persons with cognitive impairment and physical disabilities in Botswana may be a function of lack of care at the stage of critical function decline which in turn may lead to rapid deterioration and death. Short survival as a result of function decline will ultimately lead to a lower prevalence of mental disorders or physical disability in old age in Africa. Thus differential survival may be one explanation and contributing factor explaining the variation in dementia prevalence in different regions of the world, with lower rates in Africa, [12, 33, 34]. Similarly, this could explain the relatively low level of physical disability and dependence found among older Africans [13, 21]. Another possible explanation is that older persons in Africa today are "strong survivors" through tough selection mechanisms during their lifetime. These strong survivors may be less prone to chronic diseases of old age which might explain a lower prevalence in Africa of conditions such as dementia.

It is likely that future generations of older Africans will experience less extreme selection as a result of improvement in general standards of living and health care; at least in countries such as Botswana.

A mortality rate of around 11 per 100 person years among older persons in Botswana is high compared with other older populations, although no comparative data from elsewhere in Africa exist.

How can it be that older Africans with documented lower prevalence of chronic age related diseases die at more than twice the rate of European age peers?

The explanation may be that when older Africans cross the dividing line between independence and dependency they are left with little or no public or informal support; and thus they literally die. There may be no specific disease causing this high mortality; rather the sum of mortality from any cause resulting from lack of care and support when functional abilities are below a critical level.

Older community dwelling persons in Botswana with reduced cognitive or physical function, have a significantly increased risk of death. The 'traditional' extended family fails to cater for support and care for their older family members in a significant number of instances.

Policy makers should target and strengthen formal care mechanisms and encourage informal care towards older citizens, especially those with mental disorders, functional decline and limited social support, in order to reduce avoidable premature mortality.

The total burden of mental disorders and especially declining cognitive function in old age constitute a major portion, expected to increase in the decades to come, of a population's total health burden, yet these seem to receive disproportionately little attention from health planners and policymakers in the developing world and a disproportionately poor allocation of health care resources.

The project was funded by the Norwegian Council of Universities/Centre for International University Cooperation (NUFU). TCs' doctoral grant was funded through the Norwegian Research Council.