HIV,
AIDS, and the Changing Burden of Disease in Southern Africa:
A brief note on the evidence and implications.
Amar Hamoudi
June 12, 2000
Discussion around President Mbeki’s
recent AIDS panel has focused on his decision to include the
views of “HIV dissidents” including Peter Duesberg and
others, who maintain that HIV is benign.
“African AIDS,” according to Duesberg, “is
caused by malnutrition, parasitic infection and poor
sanitation. . . [T]here is no scientific evidence for the
correlation between HIV and African AIDS, only guesses.”
(“Peter Duesberg on AIDS” website, http://www.duesberg.org).
Implicit in this claim, of course, is the contention
that the AIDS epidemic has not had a qualitative impact on
the burden of disease in sub-Saharan Africa.
Even in the absence of long-term,
comparable data, however, empirical evidence on burden of
disease in sub-Saharan Africa directly contradicts this
claim. As the
AIDS epidemic has matured, it has clearly had devastating
and appreciable demographic and epidemiological effects.
Here, three pieces of evidence are cited which
demonstrate that higher levels of HIV transmission result in
a qualitative difference in the burden of disease.
First, life expectancies in the countries within
Africa which are hardest hit by HIV are significantly lower
than life expectancies in the rest of the continent and the
rest of the world even after accounting for standard
determinants of longevity including sanitation, nutrition,
and the risk of infection with parasitic diseases. Second,
these countries over the past half-decade have shown
dramatic increases in population-adjusted crude death rates,
significantly greater than the rest of the continent.
Finally, local areas within South Africa have seen
dramatic increases in morbidity and mortality in just the
past few years, which are unlikely to have been accounted
for by infrastructural deterioration or increases in the
burden of parasitic diseases.
Within Africa, the countries in the
southern region are by far the hardest hit by HIV.
Ranking the countries in the world with population of
one million or more in order by HIV seroprevalence rates as
of 1998, the top seven are in the southern African region—
Zimbabwe, Botswana, Namibia, Zambia, Malawi, South Africa,
and Mozambique. Lesotho,
the only other country in this region, is 16th
(prevalence data are from UNAIDS, Report
on the Global HIV/AIDS Epidemic June 1998, http://www.UNAIDS.org,
and are based on direct serosurveys).
Table 1 shows the results of regressions of log life
expectancy on standard determinants of longevity.
The regression in column 1 includes all countries of
population one million or more for which data are available.
Life expectancies in the regions of Africa hardest
hit by HIV are significantly shorter than the rest of the
world (by about two years), and the rest of Africa (by about
one year).
Column 3 includes only countries in Africa south of
the Sahara. Life
expectancies in the region hardest hit by HIV remain
statistically significantly lower than elsewhere on the
continent. Finally,
column 4 examines log life expectancy in 1980— before the
HIV/AIDS epidemic had begun to have a generalized effect.
In this regression, the eight countries in the
southern African region showed no statistically significant
differences in life expectancy from the rest of the
continent.
Can factors other than HIV prevalence fully explain
this difference? Incidence
of malnutrition and the availability of sanitation are not
significantly different between the southern African region
and the rest of the continent (t-test results not shown
here). The most
devastating parasitic disease in the world is malaria, which
causes over a million deaths each year (World Health
Organization, World Health Report 1999, http://www.who.int).
However, for well-documented ecological and
geographical reasons, malaria is much more widely
transmitted in tropical Africa than in much of southern
Africa (particularly Botswana and South Africa, where
conditions are particularly adverse for the malaria parasite
and its vector mosquitoes).
Furthermore, the mortality burden of conditions like
malnutrition, diarrheal diseases attendant on poor
sanitation, and parasitic diseases— including malaria—
is borne primarily among infants and young children (see
UNICEF Annual Report
on the State of the World’s Children 1998). However, the observed differences in life expectancy between
southern Africa and the rest of the continent remain
statistically significant at the 5% level, even after adding
under-5 mortality rates or log under-5 mortality rates to
the regressions in table 1, column 3 (regression results not
shown here). Factors
hypothesized to account for differences in life expectancy,
therefore, must explain differences in mortality rates among
older children, young adults, and adults.
This evidence suggests that the
burden of disease in southern Africa, where HIV transmission
is highest, is qualitatively distinct from the burden in the
rest of the continent and the rest of the world.
This difference has emerged within the past two
decades, after the onset of a generalized HIV epidemic.
Furthermore, the claim that AIDS is caused by
malnutrition and parasitic disease is in contradiction with
this evidence. In
fact, the prevalence of malnutrition (measured as the
prevalence of severely low weight for height among children
under 5 years of age) and availability of sanitation
infrastructure are poor predictors of overall life
expectancy— although they are vital to infant and child
survival. Furthermore,
southern Africa is not significantly different from the rest
of the continent in terms of these metrics (t-test results
not shown here). The evidence further indicates that differences in life
expectancy cannot be attributed simply to differences in
conditions that have measurable effects primarily on infant
or child mortality, like malnutrition or malaria.
AIDS has had a devastating and
appreciable demographic impact throughout the southern
African region. In
Botswana, one of the countries hardest hit by HIV, crude
death rates nearly doubled over the decade between 1992 and
1997, increasing from eight deaths per 1000 population to
over 15. On
average, crude death rates in the region have increased by
about 18 percent over this period, compared with a six
percent decline in the rest of the continent. Unfortunately, detailed time-series data on sanitation,
nutrition, and the incidence of parasitic disease over this
period are sorely lacking, so direct quantitative assessment
of the role of sanitation and nutrition is impossible.
However, any hypothesis to explain the rapid jump in
the population-adjusted number of deaths in the region must
include some factors unique to the region.
Infrastructural or economic collapse, famine, and
changing age structure (e.g., aging of the population) are
untenable as explanations of this change.
Southern Africa is not distinct from the rest of the
continent on any of these metrics (t-test results not shown
here). Rather,
this change is most likely accounted for by the fact that
the region has seen the longest and widest transmission of
HIV in the world.
Finally, somewhat more specific data
are available within South Africa which testify to an
increased and changing burden of disease.
For example, figure 1 shows the incidence of workers
who retire early due to ill health at one large South
African parastatal.
A steady rise is evident, with a dramatic increase
between 1997 and 1998.
This burden of absenteeism has been distributed among
firm offices across the country—in malarious as well as
non-malarious areas. The data indicate a dramatic rise in morbidity among formal
sector wage earners in urban areas of South Africa, for whom
well established sanitation and primary health care
infrastructure are fully accessible.
Similarly, figure 2 shows the steady and dramatic
increase in burials and cremations in the city of Durban
over the past several years (data collected by Professor A.N.
Smith, Department of Virology, University of Natal).
The observed 247 percent increase in less than half a
decade is suggestive of a massive increase in the burden of
mortality in metropolitan Durban, in the absence of a
massive epidemic of malnutrition or parasitic and diarrheal
disease. In
addition, a survey in rural KwaZulu Natal province in 1999
showed an 81% rise in hospital attendance at Hlabisa
hospital between 1991 and 1998.
Much of this increase was attributable to increased
incidence of tuberculosis; TB admissions were up 400% over
the period (appraisal by Sean Drysdale and others of the
state of HIV/AIDS in rural KwaZulu Natal).
The positive correlation between tuberculosis disease
and HIV infection has been well documented around the world.
Furthermore, in a regression of TB incidence against
log GDP/capita, urbanization, malnutrition, sanitation, and
BCG vaccine coverage, southern African incidence rates are
significantly higher than the rest of the world and the rest
of the continent. Adding HIV seroprevalence rates to the regression eliminates
this “southern Africa effect” (regression results not
shown here).
Therefore, a massive increase in
southern Africa’s burden of disease, and changes in the
structure of this burden, are readily observable even
despite the paucity of detailed and intertemporally
comparable data. Furthermore,
southern Africa’s burden of mortality is qualitatively
different from the burden on the rest of the continent and
the rest of the world, and this difference has only emerged
in the past two decades.
The hypothesis that this burden is solely determined
by malnutrition, poor sanitation, and parasitic diseases
like malaria fails to account for these observations.
These observations are most likely the early
manifestations of a generalized, advanced and advancing AIDS
epidemic. More
detailed data are urgently needed in order to assess and
understand the changing burden of disease in southern Africa
in the face of this epidemic.
For example, the evidence cited here indicate that
the burden of mortality in the region is disproportionately
greater among older children, young adults, and adults
compared to the rest of the world.
However, this burden is likely to shift increasingly
toward infants as the epidemic progresses.
Furthermore, only the broadest and most catastrophic
effects of any epidemic are manifested by changes in crude
death rates and life expectancy.
Other demographic effects, including changes in age
structure, population distribution, and household dependency
ratios have massive social, economic, and welfare
repercussions, although they cannot be assessed using crude
national-level measures.
|
Log
life expectancy
|
1
All countries, 1995
|
2
All countries, 1995
|
3
Africa only, 1995
|
4
All countries, 1980
|
|
Log,
GDP per capita 1995 (1980 for column 4)
|
0.0288
(1.75)*
|
0.0371
(4.37)***
|
0.0238
(0.630)
|
0.0311
(2.73)***
|
|
Log,
per capita calorie supply, 1995 (1980 for column 4)
|
0.128
(2.99)***
|
0.0279
(0.726)
|
0.158
(1.27)
|
0.132
(2.48)**
|
|
Literacy
rate (%) among adult females, 1995 (latest year
available 1977-1985 for column 4)
|
0.00209
(6.32)***
|
0.00178
(6.55)***
|
0.00316
(4.91)***
|
0.00270
(9.51)***
|
|
Prevalence
(%) of stunting, children under 5, latest year
available 1990-1998
|
-0.000242
(0.336)
|
|
0.0000457
(0.025)
|
|
|
Proportion
(%) of population with access to sanitation
infrastructure, latest year available 1987-1995
|
-0.000247
(0.738)
|
|
-0.000254
(0.410)
|
|
|
Proportion
of population living within 100 km of coast or
navigable river (%)
|
.0610
(2.40)**
|
0.0411
(2.63)***
|
0.697
(0.70)
|
0.00468
(0.213)
|
|
Index:
risk of infection with falciparum malaria (0 no risk,
3 high risk), 1994 (1982 for column 4)
|
-0.110
(4.80)***
|
-0.124
(5.98)***
|
-0.105
(2.01)*
|
-0.0612
(1.81)*
|
|
Dummy:
sub-Saharan Africa
|
-0.0602
(3.03)***
|
-0.0767
(3.47)***
|
|
-0.0871
(2.74)***
|
|
Dummy:
southern Africa
|
-0.0623
(2.54)**
|
-0.0632
(2.61)**
|
-0.0814
(2.63)**
|
-0.00981
(0.321)
|
|
Constant
|
2.82
(8.72)***
|
3.55
(13.47)***
|
2.51
(2.63)**
|
2.68
(7.23)***
|
|
N
|
81
|
137
|
35
|
92
|
|
R-squared
|
0.90
|
0.90
|
0.67
|
0.92
|
*- significant at 10% level; **-
significant at 5% level; ***- significant at 1% level
Absolute value of robust t-statistics in
parentheses
