Hiv Prevalence and Associated Factors: John Chipeta, Erik Schouten, John Aberle-Grasse
Hiv Prevalence and Associated Factors: John Chipeta, Erik Schouten, John Aberle-Grasse
Hiv Prevalence and Associated Factors: John Chipeta, Erik Schouten, John Aberle-Grasse
FACTORS 12
John Chipeta, Erik Schouten, John Aberle-Grasse
AIDS is one of the greatest public health and social problems threatening the human race.
The greatest burden of the HIV/AIDS pandemic is in sub-Saharan Africa. According to the Joint
UN Committee on HIV/AIDS (UNAIDS, 2004), an estimated 38 million people worldwide were
living with HIV in 2003, of which 5 million were newly infected. In 2003, two-thirds of all people
living with HIV/AIDS (25 million) were in sub-Saharan Africa, which has about 10 percent of the
worlds population.
Malawi has one of the highest national prevalence rates in the world. Heterosexual contact is
the principal mode of HIV transmission, while mother-to-child transmission (MTCT) accounts for
about 25 percent of all new HIV infections (NAC, 2004a).
Monitoring and evaluation data for 2004 show a momentous increase in programme
intervention coverage. Subsequently, some positive changes in behaviour, especially among men,
have been observed and documented (NAC, 2004a). The National AIDS Commission (NAC) has
coordinated the development of a National AIDS Framework for 2005 to 2009, which is expected
to galvanise a decentralised comprehensive multi-sectoral national response. With the strengthening
of the multi-sectoral national response to HIV and AIDS, HIV transmission is expected to decline.
However, HIV prevalence will likely remain high or even increase for some time, as antiretroviral
therapy is scaled up. Consequently, deaths due to AIDS are reduced.
ANC sentinel surveillance systems use unlinked anonymous methods for specimen collection
and testing to avoid participation bias which can significantly affect the HIV prevalence rates.
However, other biases are inherent in sentinel surveillance systems: health facilities are not randomly
selected and tend to be urban; pregnant women may be having unprotected sex at a greater rate than
the general population, which could overestimate the prevalence; the prevalence in ANC attendees
may underestimate what is happening in the general population because women with HIV
associated infertility are not captured; and men and non-pregnant women are not included in the
sentinel surveillance sample. To obtain a nationally-representative HIV prevalence estimate for all
adults, sentinel surveillance data should be adjusted based on assumptions about the biases in the
clientele who use the selected facilities and part of the population that does not use antenatal clinic
services.
The 2004 MDHS is the third survey in Malawi conducted as part of the international DHS
program, and the first to anonymously link the HIV results with key behavioural, social and
This chapter presents characteristics of respondents who accepted and refused to take an
HIV test. Findings are presented on HIV prevalence by various demographic and socioeconomic
characteristics. Being the first survey to present estimates of HIV prevalence at the national, urban-
rural, regional, and district levels, data from the 2004 MDHS serve as baseline findings. Trend
analysis can only be done after another national sero-survey is conducted.
As described in Chapter 1, every third households in the 2004 MDHS sample was selected
for individual interviews with male respondents. All men age 15-54 were eligible for individual
interview. In the same households, all women age 15-49 and all men age 15-54 were eligible for
HIV testing. Overall, 4,071 women age 15-49 and 3,797 men age 15-54 were identified as eligible
for testing. Of these, testing was successfully conducted on 2,686 women and 2,581 men, resulting
in a response rate of 70 percent for women and 63 percent for men.
Table 12.1 presents the coverage rates for HIV testing by sex, urban-rural residence, and
region. Based on the reason for nonresponse, respondents who were not tested are divided into four
categories:
those who refused testing when asked for informed consent by the health worker (22 percent
overall)
those who were interviewed in the survey, but who were not at home when the health worker
arrived for testing and were not found on callbacks (less than one percent)
those who were not at home for the testing and were never interviewed (9 percent), and
those who were missing test results for some other reason, such as they were incapable of
giving consent for testing, there was a mismatch between the questionnaire and the blood
sample, or there was a technical problem in taking blood (1 percent).
While refusal rates for women and men are similar (23 percent and 22 percent, respectively),
women are more likely to be found at home than men; 5 percent of women were absent compared
with 14 percent of men. The difference in nonresponse rates between women and men are more
significant in urban areas and in the Southern Region.
Table 12.1 shows that response rates are consistently higher in rural areas. For both women
and men, urban respondents are more likely to refuse taking the test or to be absent during the
survey. For example, nonresponse resulting from absence for urban men is 20 percent compared
with 13 percent for rural men. Across regions, respondents in the Northern Region are much less
likely than those in the other regions to refuse testing. Overall, the refusal rate in the Northern
Percent distribution of women age 15-49 and men age 15-54 eligible for HIV testing by testing status, according to resi-
dence and region (unweighted), Malawi 2004
Residence Region
Testing status Urban Rural Northern Central Southern Total
WOMEN
MEN
TOTAL
Table 12.2.1 shows that response rates also vary across the respondents background
characteristics. HIV testing coverage among women varies from 65 percent among those age 15-19
to 76 percent among women age 40-44. Women with no education and in the highest wealth
quintile are the least likely to have been tested. The response rate for women in the richest group is
67 percent, with 26 percent of non response due to refusal and 7 percent due to absence. The
response rate for testing in Lilongwe is surprisingly low (39 percent). Field implementation of blood
sample collection was not adequate to provide district-specific estimates (see Section 12.2.2 below
for a modeling approach that provides a prevalence estimate). In other oversampled districts, the rate
ranges from 65 percent in Blantyre to 77 percent in Salima.
Testing coverage among men also varies by age (Table 12.2.2). Men 15-19 are the least likely
to be tested (60 percent) while men age 35-44 years have the highest coverage (67 to 68 percent). It
is interesting to note that response rates among men increases with education ranging from
57 percent for men with no education to 65 percent or men with secondary or higher education. As
Percent distribution of women age 15-49 eligible for HIV testing by testing status, according to background characteristics
(unweighted), Malawi 2004
Testing status
Background Absent for Other/
characteristic Tested Refused testing missing Total Number
Age
15-19 65.3 24.4 8.5 1.8 100.0 835
20-24 70.6 23.9 4.2 1.3 100.0 979
25-29 70.8 21.8 5.2 2.2 100.0 744
30-34 72.9 20.9 3.9 2.3 100.0 532
35-39 68.7 24.6 5.0 1.7 100.0 403
40-44 75.9 18.0 4.6 1.5 100.0 323
45-49 74.9 18.8 3.9 2.4 100.0 255
District
Blantyre 64.7 23.0 10.6 1.7 100.0 235
Kasungu 75.6 18.2 5.2 1.0 100.0 308
Machinga 73.5 21.6 3.8 1.1 100.0 264
Mangochi 67.8 22.7 2.9 6.6 100.0 273
Mzimba 73.8 21.4 4.5 0.3 100.0 332
Salima 77.0 17.1 3.6 2.4 100.0 252
Thyolo 69.0 21.4 7.9 1.7 100.0 290
Zomba 73.2 23.3 3.5 0.0 100.0 257
Lilongwe 39.0 51.5 7.5 2.1 100.0 241
Mulanje 70.5 23.4 5.4 0.8 100.0 261
Other districts 73.3 19.6 5.2 2.0 100.0 1,358
Education
No education 66.9 23.2 7.3 2.7 100.0 1,048
Primary 1-4 70.9 23.4 4.3 1.4 100.0 1,064
Primary 5-8 72.7 21.0 4.5 1.9 100.0 1,389
Secondary+ 70.0 23.3 5.8 0.9 100.0 570
Wealth quintile
Lowest 69.5 23.5 5.4 1.5 100.0 718
Second 70.7 22.2 4.8 2.3 100.0 817
Middle 72.1 20.5 4.8 2.6 100.0 894
Fourth 72.1 21.3 5.3 1.4 100.0 875
Highest 66.6 25.7 6.5 1.2 100.0 767
Total 70.4 22.5 5.3 1.8 100.0 4,071
As in the case of women, men in Lilongwe are the least likely to be tested for HIV (38
percent). Response rates are also low in Mangochi (50 percent) and Blantyre (54 percent). On the
other hand, men in Kasungu have the highest response rates (76 percent). In all districts, absence is
an important reason for nonresponse among men. The highest absence rate was observed in Blantyre
(26 percent) and Thyolo (22 percent). The highest refusal rate is recorded in Lilongwe (49 percent),
while the lowest refusal rate is in Salima (17 percent).
Percent distribution of men age 15-54 eligible for HIV testing by testing status, according to background characteristics
(unweighted), Malawi 2004
Testing status
Background Absent for Other/
characteristic Tested Refused testing missing Total Number
Age
15-19 59.8 25.0 14.7 0.5 100.0 761
20-24 61.6 22.5 15.7 0.3 100.0 690
25-29 64.5 21.5 13.2 0.7 100.0 710
30-34 63.7 21.0 14.9 0.4 100.0 557
35-39 67.0 18.0 13.8 1.2 100.0 333
40-44 68.1 18.5 12.5 0.9 100.0 335
45-49 61.8 24.6 12.6 1.0 100.0 207
50-54 64.7 21.1 13.7 0.5 100.0 204
District
Blantyre 53.9 19.6 25.7 0.7 100.0 280
Kasungu 76.2 16.9 6.3 0.6 100.0 332
Machinga 66.5 19.1 13.9 0.4 100.0 230
Mangochi 50.4 29.1 18.8 1.7 100.0 234
Mzimba 73.2 12.8 12.5 1.6 100.0 313
Salima 68.8 14.9 15.3 0.9 100.0 215
Thyolo 56.5 21.9 21.6 0.0 100.0 269
Zomba 67.1 21.1 11.8 0.0 100.0 237
Lilongwe 38.2 49.0 12.0 0.8 100.0 259
Mulanje 62.6 20.9 15.6 0.9 100.0 211
Other districts 66.6 21.0 12.2 0.2 100.0 1,217
Education
No education 56.7 25.1 18.0 0.2 100.0 467
Primary 1-4 60.4 24.9 13.9 0.8 100.0 961
Primary 5-8 66.8 19.7 12.9 0.6 100.0 1,413
Secondary+ 64.5 20.6 14.3 0.5 100.0 950
Wealth quintile
Lowest 58.9 24.4 16.5 0.2 100.0 509
Second 64.6 22.2 12.4 0.8 100.0 765
Middle 67.2 19.3 13.3 0.2 100.0 865
Fourth 67.3 18.6 12.9 1.2 100.0 851
Highest 56.5 26.3 16.7 0.5 100.0 807
Total 63.3 21.9 14.2 0.6 100.0 3,797
The 2004 MDHS indicates that 12 percent of the population age 15-49 in Malawi is living
with HIV/AIDS (Table 12.3). HIV prevalence among women is higher for women than for men
(13 percent compared with 10 percent). Prevalence peaks at 19 percent for women and men age 30-
34, 18 percent for women and 20 percent for men. Women start getting the infection at a younger
age than men; the prevalence among women age 15-19 is 4 percent compared with less than 1
percent for men of the same age. HIV prevalence among women is higher than that for men until
age group 30-34 and 35-39. At ages 40-49, the prevalence among men is again lower than the
prevalence among women (Figure 12.1).
Surveillance of AIDS cases indicate that very few children who were infected through
mother-to-child transmission survive up to 15 years of age. Therefore, prevalence among the youth
represents more recent HIV infections and is recognised and used as a proxy indicator for tracking
incidence. Overall, HIV prevalence among women and men age 15-24 is 6 percent. Prevalence
among women in this age group is more than four times higher than that for men 15-24 (9 and
2 percent, respectively). These figures are useful in measuring progress towards the National HIV
and AIDS Action Framework 2005 to 2009.
Percentage HIV positive among women age 15-49 and men age 15-49 (54) by age, Malawi 2004
na = Not applicable
25
20
15
Percent
Women
Men
10
0
15-19 20-24 25-29 30-34 35-39 40-44 45-49
Age
MDHS 2004
Table 12.4 shows that urban residents have a significantly higher risk of HIV infection than
rural residents. While 18 percent of urban women are HIV positive, the corresponding proportion
for rural women is 13 percent. For men, the urban-rural difference in HIV prevalence is even
greater; urban men are nearly twice as likely to be infected as rural men (16 and 9 percent,
respectively). Since 85 percent of Malawis population live in rural areas, the greatest burden of HIV
infection is in the rural population.
The HIV epidemic shows regional heterogeneity. The prevalence among women in the three
regions is similar to what has been seen in ANC sentinel surveillance estimates, high in the Southern
Region (20 percent), and low in the Northern (10 percent) and Central (7 percent) Regions. The
regional differential in HIV prevalence for men is somewhat different than that for women, high in
the Southern Region (15 percent) and lower in the Northern (5 percent) and Central (6 percent)
Regions.
ANC surveillance system data and VCT data for Malawi show that women with secondary
or higher education have higher infection levels than women with less education (NAC, 2004a).
Data in Table 12.4 show that HIV prevalence is somewhat constant across education levels, but
higher among women with secondary or higher education. For men, however, education has a
positive relationship with the risk of infection; the rate of infection increases with education.
Work status is related to the HIV rate for both women and men. Fifteen percent of working
women are HIV positive compared with 12 percent of women who are not working. For men, the
difference is more dramatic, 13 percent for working men, compared with 6 percent for men who are
not working. Rates of HIV infection also increase with the wealth quintile; overall, the infection rate
in the highest quintile is two times that in the lowest quintile (16 and 8 percent, respectively). This
Percentage HIV positive among women and men age 15-49 who were tested, by background characteristics, Malawi
2004
Across religions, HIV prevalence varies by gender. Anglican and Muslim women have the
highest infection rate (18 and 17 percent, respectively). For men, those who are Seventh Day
Adventists have the highest rate (17 percent).
Because of the low response rate for HIV testing in Lilongwe (see Tables 12.2.1 and 12.2.2),
and the implausible pattern of infection where male prevalence is higher than female prevalence,
additional analysis of the Lilongwe results was undertaken. A statistical model was developed using
the questionnaire information from individuals who were tested for HIV in Malawi outside of
Lilongwe. A nationally common set of predictor variables, including background and behavioural
characteristics, were used to predict HIV status for women and men separately. Where individual
interviews were not carried out, information from the household questionnaire was used to predict
HIV status. The model parameters were then applied to the Lilongwe sample to predict HIV status.
The resulting predictions, or adjusted HIV rates, for Lilongwe are substantially higher than
the observed prevalence. For women, the observed HIV prevalence of 1.6 percent is raised to 11.5
percent by the adjustment (Table 12.5). For men, the observed rate of 5.5 percent is increased to 9.2
percent. The resulting adjusted figures for Lilongwe are much closer to the expected HIV levels
based on the ANC sentinel surveillance results. In addition, the adjusted prevalence for women and
men in Lilongwe are consistent with the patterns by sex observed in other districts and regions in
Malawi.
Observed Adjusted
Geographic area prevalence prevalence
WOMEN
MEN
TOTAL
Because nonresponse for HIV testing may bias the results, HIV prevalence rates among non-
tested women and men in the rest of Malawi were predicted using the same multivariate statistical
models. The results of this analysis, including the above adjustment for Lilongwe, show that the
Percentage HIV positive among women and men age 15-49 who were tested, by sociodemographic characteristics,
Malawi 2004
The relationship between HIV prevalence and circumcision status is not in the expected
direction. In Malawi, circumcised men have a slightly higher HIV infection rate than men who were
not circumcised (13 percent compared with 10 percent). In Malawi, the majority of men are not
circumcised (80 percent). The practice of circumcision varies greatly across ethnicity, ranging from
82 percent among the Yao and 30 percent among the Lomwe to 2 percent among the Tumbuka (see
Chapter 11). As in Table 12.4, ethnicity is also significantly associated with HIV infection. It is
interesting to note that women and men in ethnic groups with high proportion of circumcision such
as Yao and Lomwe, the prevalence of HIV infection is also high. For example, 20 percent of Lomwe
women and 18 percent of Lomwe men as well as 18 percent of Yao women and 12 percent of Yao
men are HIV positive. While Ngoni men are not customarily circumcised, they also have a higher
prevalence compared with other ethnic groups (15 percent). These observations suggest that the
relationship between circumcision and HIV sero status is not straightforward. Further analysis is
needed to determine the relationship between male circumcision and the risk of HIV infection.
In the 2004 MDHS, male respondents were asked whether they spent any time in past 12
months away from home, and in the same time period, whether they were away from home for more
than one month. The survey results show that in general, men who stayed home have the lowest
HIV prevalence. Men who were away from home for more than one month have a higher risk (13
percent) of HIV infection than those who were away for less than one month at a time (11 percent).
Table 12.7 examines the prevalence of HIV infection by sexual behaviour indicators among
respondents who have ever had sexual intercourse. In reviewing these results, it is important to
remember that responses about sexual risk behaviours may be subject to reporting bias. Also, sexual
behaviour in the 12 months preceding the survey may not adequately reflect lifetime sexual risk.
For women, there is a clear pattern of higher HIV prevalence with earlier sexual debut.
Women who started having sex at an early age (before age 15) have higher HIV prevalence than
those with a later sexual debut (18 percent compared with 15 percent or lower). This pattern is not
evident among men.
Percentage HIV positive among women and men age 15-49 who ever had sex and were tested for HIV, by sexual
behaviour characteristics, Malawi 2004
Men who paid for sex in the period prior to the 12 months preceding the survey have a
higher HIV prevalence (18 percent) than either those who have never paid for sex (10 percent) or
those who paid for sex in the past 12 months (11 percent).
Condom use does not make much difference in the likelihood of a woman being infected
with HIV. HIV prevalence among women who said that they never used a condom and those who
used a condom at some time is 15 percent each. In contrast, men who never used a condom have a
lower prevalence of HIV than those who did use a condom at some time (9 and 14 percent,
respectively).
Table 12.8 presents HIV prevalence by other characteristics related to HIV risk among men
and women who have ever had sex. As expected, women and men with a history of a sexually
transmitted infections (STIs) or STI symptoms have much higher rates of HIV infection than those
with none. Women and men with STIs are twice as likely to be HIV positive as those who have no
STI. For example, 26 percent of women who report having an STI or symptoms of an STI are HIV
positive, compared with 13 percent of women who did not have an STI or STI symptoms.
Percentage HIV positive among women and men age 15-49 who ever had sex and who were tested, by other
characteristics, Malawi 2004
The uptake of HIV testing in Malawi remains below 25 percent in the adult population and
data on HIV testing indicate that the most common reasons for seeking testing is concern regarding
infection risk and illness (MACRO, 2004). As might be expected from this finding, women and
men who have been tested for HIV have higher rates of HIV infection than those who have never
been tested. For example, 13 percent of men who have been tested for HIV are HIV positive,
compared with 11 percent of men who have never been tested.
Although the individuals HIV status is associated with prior HIV testing, the results in
Figure 12.2 show that four of five of those infected with HIV (85 percent of infected women and 80
percent of infected men) do not know their HIV status, either because they were never tested or, to a
small extent, because they were tested and did not receive their results. Men are more likely than
women to know their sero status. This is particularly true for HIV-positive individuals.
HIV Positive HIV Negative HIV Positive HIV Negative HIV Positive HIV Negative
0%
100
15 13 20 16 17 14
10%
Percentage of women and men
2
80 2 2
2 2
20% 2
60
30% 83 86 78 82 81 84
40
40%
20
50%
60% 0
Women Men Total
Ever tested, knows result Ever tested, does not know result Never tested
Table 12.9 presents HIV prevalence among youth by background characteristics. Young
people living with HIV are more likely to have been more recently infected compared with adults.
Consequently, statistics on variation of HIV prevalence among youth is critical in understanding the
patterns of recent HIV infections. HIV prevalence among younger people does not reflect the
cumulative burden of AIDS because it does not take into account AIDS-related mortality in the
general population.
Overall, 6 percent of youth are infected with HIV. Prevalence of HIV is more than four
times higher among young women than among young men (9 percent compared with 2 percent).
Youths in the Southern Region have the highest HIV prevalence compared with those in the
Northern and Central Regions (9 percent compared with 6 and 3 percent, respectively).
HIV prevalence in youth in the urban areas is similar to that in rural areas (7 and 6 percent,
respectively). In the past seven years, HIV prevalence in urban areas was estimated to be substantially
higher than in rural areas (NAC, 2004b). The 2004 MDHS result suggests that incidence of HIV in
rural areas has reached that in urban areas. The highest HIV prevalence among young women is
found among women in the urban areas, in the Southern Region, and women who are divorced or
separated. Prevalence is consistently higher among female youth compared with that among male
youth.
Percentage HIV positive among women and men age 15-24 who were tested for HIV, by background characteristics,
Malawi 2004
Note: An asterisk indicates that an estimate is based on fewer than 25 cases and has been suppressed.
The high prevalence among youth who are in a union (9 percent) compared with those who
have never been in a union (2 percent) indicates that early marriages may be linked to early sexual
debut and other risks.
HIV prevalence increases with increasing number of sexual partners in the past 12 months.
Women and men who report having two or more partners are more likely to be HIV positive than
those who had only one partner in the past 12 months (10 and 7 percent, respectively). For women
and men who have higher-risk sex, the corresponding proportions are 7 and 5 percent, respectively.
Among the 1,324 cohabiting couples who were tested for HIV in the 2004 MDHS, for
83 percent both partners are HIV negative and for 7 percent both partners are HIV positive. Ten
percent of the couples are discordant, that is, one partner is infected and the other not (Table
12.10). The variations in the level of HIV infection of both partners by background characteristics
generally conform to the patterns observed in the variations in womens seroprevalence rates.
Infection rates are highest among couples in urban areas and in the Southern Region, and among
those with higher education and in the higher wealth quintiles.
Among cohabiting couples who were tested, percent distribution by results of HIV testing, according to background characteris-
tics, Malawi 2004
Note: An asterisk indicates that an estimate is based on fewer than 25 cases and has been suppressed.
The inclusion of HIV testing in the 2004 MDHS provides the basis for a more precise
estimate of the burden of HIV in Malawi and permits the calibration of estimates of HIV prevalence
based on sentinel surveillance in pregnant women. Malawi has a heterogeneous HIV epidemic, with
significant differences in the disease burden by region and ethnicity.
The linkage of biological and behavioural data in this survey has strengthened the validity of
this survey for allowing multivariate analyses. The measurement of HIV prevalence in the 2004
MDHS should prove useful in calibrating HIV prevalence estimates of the general population from
sentinel surveillance in pregnant women.
This link between HIV test results and demographic and behavioural data also enhances the
understanding of the distribution, patterns, and risk factors for HIV in Malawi, with the potential
for improved planning and implementation of programs as a result of this information. Finally, the
prevalence of couples that are discordant for HIV underscores the need for knowledge of both ones
own HIV status and that of ones partner to prevent the continued spread of HIV. Subsequently,
some positive changes in behaviour, especially among men, have been observed and documented
(NAC, 2004b).