Socioeconomic disparities and the distribution of HIV in Cameroon: a multilevel logistic regression analysis

Background: There have been controversial debates on the relationship between socioeconomic status and the distribution of HIV in Cameroon. We aim to illustrate the vulnerability of socioeconomic disparities and the risk of getting HIV for public health interventions. Methods: Descriptive statistics was conducted to quantify the socioeconomic gradients of HIV. A Multilevel logistic regression model was used to study the relationship between socioeconomic factors and HIV. The effect of the factors was presented as odds ratios (OR), with 95% con�dence intervals (CIs). P-value less than 0.05 was considered to be statistically signi�cant. We further mapped HIV prevalence in ArcGIS to visualize the regional distribution of HIV. Results: HIV was signi�cantly associated with age (p<0.001), sex (p<0.001) and varies signi�cantly by geographic region (p<0.001), level of education (p=0.001), wealth status (p<0.001), religion (p=0.042), ethnicity (p<0.001) and residence (p=0.001). HIV positive participants were more likely to be women, people with higher educational level, live in urban areas, practice the protestant religious belief, belong to the ethnicity of Kako/Meka/Pygmy and distributed in the East, South, and Yaoundé regions. Age, sex, region, education level, and ethnicity were signi�cantly associated with the odds of having HIV from the multilevel regression model. Conclusion: Our �nding recommends for novel intervention programs that will target the various socioeconomic factors associated with the odds of having HIV for proper public health management of the disease in Cameroon.

There have been controversial debates on the relationship between the socioeconomic status and the distribution of HIV in Cameroon and other Sub-Sarah Africa (SSA) countries.Most studies argued that an increase in socioeconomic statuses such as wealth and education will increase the risk of getting HIV due to a uent lifestyles and many sexual partners (5,6) while others reported poverty and no education as potentially vulnerable factors (5)(6)(7)(8)(9)(10)(11).In Cameroon, HIV prevalence among women was higher among those from wealthier households.Moreover, HIV prevalence was lowest among women and men who have had no education.The Demographic and Health Survey (DHS) report demonstrated that, among women, HIV prevalence was higher among those with higher education at the national level, though other studies have reported lower prevalence at the City level (6,12).Socioeconomic disparities in population and global health outcomes are emphasized because social conditions, such as knowledge, wealth, education, power, prestige, and bene cial social associations, allow individuals to avoid health-related risks and to adopt protective strategies, and access to quality health care services (9).Understanding these differences in socioeconomic status with respect to HIV prevalence and distribution is very vital for strategic HIV prevention programs in Cameroon and SSA.
Although the DHS program in Cameroon has highlighted some general trends of the relationship between socioeconomic disparities and the risk of getting HIV and their distribution (12), detailed assessment of these differences and other related factors are limited or underemphasized in reported studies(6) as there are substantial variations in these relationships across, regions, tribes, religions, ethnicity, residence, wealth, educational level and employment status.In order to elucidate these variations in Cameroon, we aim to illustrate the vulnerability of socioeconomic disparities and the risk of getting HIV for public health interventions.  1) and survey design are described elsewhere (13,14).

Materials And Methods
Analyses in this paper were restricted to participants with non-missing information on key variables of interest (i.e.sex, age, education level, ethnicity, place of residence, employment status, wealth quintile, region, and religion).Participants were excluded if they missed the self-report of HIV.

Ethical approval
Data access for downloads was obtained through a written request from the DHS data discharge team of the USAID.The linkage of DHS HIV test results (HIV Biomarker) to the full DHS survey record allows for an in-depth analysis of the sociodemographic and behavioral factors associated with HIV infection (13).
Datasets, which include test results that can be linked to other variables in the DHS are available for research and study.All DHS studies in Cameroon and data release have been approved by the Cameroon government.

HIV Prevalence Testing.
HIV prevalence estimates are usually obtained from sentinel surveillance systems that monitored HIV rates among pregnant women and high-risk populations.By collecting blood for HIV testing from representative samples of the population of women and men in a country, the DHS Program can provide nationally representative estimates of HIV prevalence rates for over 60 countries.Both sentinel surveillance and population-based sources of prevalence data are used to track HIV epidemics.Population-based testing is usually carried out every 3 to 5 years in most countries, because of the size and cost of the surveys (13).Sentinel surveillance testing is often reported annually and provides a good benchmark for measuring progress over short time periods.Furthermore, population-based testing is dependent on the population's willingness to be voluntarily tested for HIV.

HIV testing protocol and measurement
The DHS HIV testing protocol provides for informed, anonymous, and voluntary testing of women and men, usually age 15-49.The testing protocol undergoes a host country ethical review.Biomarker agents test participants on the eld using full blood samples.Alere Determine is used as the rst-line test and unigold is used as the second line.Results are handed to the participants after post-test counseling and positive cases are referred to treatment sites.Indeterminate cases are referred for con rmations by ELISA in designated facilities within their clusters.Dried blood spot samples are also collected from participants and transferred to the reference laboratory (CPC) for quality control of HIV results via Western blot and ELISA techniques.

Statistical analysis.
Categorical variables were described using frequencies and proportions accounting for the survey design.Sample weights, mean, and standard deviations were conducted to quantify the socioeconomic gradients of HIV.The χ 2 test was performed to assess the proportional difference of categorical variables and t-test was used for continuous variables to contrast the Cameroon DHS participants by HIV status.A multilevel logistic regression model was used to investigate the association between socioeconomic factors and HIV, adjusting the in uence of survey clusters in Cameroon.Firstly, we constructed an empty model without any covariates to determine whether it was necessary to account for the multilevel structure of the data.Secondly, we constructed the multilevel logistic regression to study the relationship between socioeconomic factors and HIV disease while considering the effect of the clusters.The effect of factors was presented as odds ratios (OR) with 95% con dence intervals (CIs), derived using Markov Chain Monte Carlo methods, accounting for the survey design and sampling weights.Prevalence estimates for the whole study population, population subgroups, and regression analysis were also accounted for by the complex survey design and sampling weights.
Descriptive statistics and prevalence rate analyses were derived using SAS 9.2.A multilevel logistic regression model was constructed using Stata version 14 (Stata-Corp.,College Station, TX, USA).P-value of less than 0.05 was considered to be statistically signi cant.To further visualize the distribution of HIV prevalence in the different DHS geographic regions, we mapped HIV prevalence in ArcGIS 10.3 (ESRI, Redlands, California, USA) software.

General characteristics according to HIV status.
A summary of the socioeconomic and demographic characteristics according to HIV status is presented in Table 1.Among the 10,478 subjects, 49.4 % were male, 50.6% were female and 493(4.7%)participants had HIV.The mean age of the whole study population was 31.4.HIV was signi cantly associated with age (p<0.001) and sex (p<0.001).Moreover, HIV signi cantly varied by geographic region (p<0.001),level of education (p=0.001),wealth status (p<0.001),religion (p=0.042),ethnicity (p<0.001) and residence (p=0.001).Compared to the Non-HIV participants, the HIV participants were more likely to be women, within the higher education level, within protestant religion, lived in urban areas, within poorer to richest wealth index, distributed in the East, South, and Yaoundé DHS regions, and belong to the ethnicity for Kako/Meka/Pygmy.The incidence of HIV increases with the improvement of socioeconomic level.

Multilevel logistic regression analysis of factors associated with HIV
Under the empty model, there was a signi cant variation in the odds of having HIV (τ = 0.456, 95% CIs: 0.301-0.691)across clusters in Cameroon.The interclass correlation coe cient indicated that 12.2% of the variance in the odds of having HIV could be attributed to the difference between clusters.Table 2 shows the result of the multilevel logistic regression analysis of factors associated with HIV.Age, sex, region, education level, and ethnicity were signi cantly associated with the odds of having HIV while considering the effect of the clusters.The odds of HIV increased by 1.03-fold (95% CIs: 1.02-1.04)for every 1-year increase in a participant's age.Male participants had lower odds of having HIV than female participants (OR=0.39,95% CIs: 0.33-0.46).Participants who lived in the Far North had 76% (p<0.001) less likely to have HIV than in the Adamawa region.Participants with educational level of primary and secondary were 105% (p<0.001) and 117% (p<0.001) more likely to have HIV respectively, than those with no education.Participants belonging to the ethnicity of Bamilike/Bamoun had lower odds of having HIV than those of Arab-Choa/Peulh/Houssa/Kanuri (OR=0.53,95% CIs: 0.29-1.00).
Figure 2 presents the Geographic Information Systems (GIS) spatial distribution analysis and it indicates that the South region had the highest HIV prevalence followed by Yaoundé, Center, East, Southwest, and Northwest.HIV prevalence was higher in women from all the regions than men except for the West region.

Discussion
In this study, we investigated how socioeconomic differences could be associated with the risk of having HIV in Cameroon.Our study used HIV biomarkers and household sociodemographic data that were collected from eligible participants.These ndings have contributed to the vast majority of studies in SSA that have investigated the association of socioeconomic status and the prevalence of HIV (5,10,15,16).
Overall, HIV prevalence was 4.7 % which is a decrease from 5.5% during the Cameroon 2004 DHS.The South region had a higher prevalence of HIV as compared to the East that saw higher HIV prevalence during the 2004 surveys (12), while the Far North region still maintains a low HIV prevalence in all the cross-sectional studies.The change in the HIV prevalence pattern by region may be due to rapid urbanization that has seen a subsequent increase in HIV prevalence in the urban areas as compared to the rural areas.Our descriptive ndings con rm with that of Hajizadel ( 15) who had similar results with increase HIV prevalence in urban areas though common among the poorest group of people in Uganda, Kenya, Zimbabwe, and Swaziland.Comparing with those without HIV, HIV was common in those who have attained a high educational level, rich, and residing in urban areas.
HIV prevalence was signi cantly associated with age, sex, region, education level, and ethnicity.The odds of having HIV increases with age.This is true for the DHS participants as most of those enrolled in the survey is usually 15 years and above.
HIV prevalence by gender in Cameroon was higher in women than men.Women seen as a high-risk group calls for targeted prevention programs.For example, in 2013 Kisesa and Chamla (17) reported that twothirds of the 250,000 new HIV infections were common among adolescent girls between 15-19 years.In this report, Cameroon, India, Malawi, Haiti, Namibia, and Zimbabwe contributed to 50% of adolescent's HIV transmission; high prevalence of intimate partner violence, age-desperate sex, and gender-based inequality were regarded as the three common factors responsible for higher HIV prevalence among girls (17)(18)(19).With support from our spatial HIV distribution map (Figure 2) that has been used to manage HIV prevention programs elsewhere (20)(21)(22)(23)(24)(25), HIV prevalence by gender in Cameroon base on the risk factors of intimate partner violence, age-desperate sex, and gender-based inequality in women and girls can be further investigated in regions of alarming female HIV prevalence such as the South, East, Southwest, Northwest, and Yaoundé to inform for intervention and outreach programs.
Although the male participants had lower odds (OR=0.39,95% CIs: 0.33-0.46) of having HIV than females in our study, prevention measures such as those including HIV knowledge, attitude, and selftesting should be encouraged among men.In a study on "Perceived cost advantages and disadvantages of purchasing HIV self-testing kits among urban Tanzanian men: an inductive content analysis", men were not willing to go for counseling or visit clinics for HIV testing because they never wanted to stay away from work as they were the only provider of income for their families(26).This is not different from Cameroon as most male participants would be away at work during the time of the DHS survey and thus a lower number of male participants.Further studies are needed to compare the socioeconomic gradients between male and female and their association with HIV in Cameroon.
Educational level had a signi cant association with having HIV in Cameroon.Participants with primary or secondary education have higher odds of contracting HIV than those with no education.This is not the rst time more educated people have been proven to be more likely to contract HIV in Cameroon.A multilevel logistic analysis of the 2004 DHS data set for Cameroon by Jane (6), demonstrated that highly educated participants especially women, were more likely to have had premarital sex making them vulnerable to the disease.These ndings were also true for Burkina Faso, Kenya, and Ghana, whereas years of schooling were positively associated with rst intercourse for men in Kenya and Cameroon and thus increase sexual behaviors, practices, and risks of getting HIV.Having a high level of educational attainment has been known to be a protective factor against HIV infections and having a general knowledge of the disease (10,27).However, it is a reverse paradigm in Cameroon and many SSA counties.This may be due to the fact that high level of education increases the socioeconomic status such as wealth, better jobs, and overall social status which have been associated with premarital sex, unprotective sexual behaviors, and multiple partners which are risk factors of contracting HIV.Contrary to our study, ndings on the socioeconomic status and the vulnerability to HIV infection in Uganda demonstrated that those with higher educational attainment were less likely to be HIV positive than those with no education (10).Furthermore, higher educational attainment was positively associated with HIV testing among pregnant women in Zambia 2014 DHS study (28).Despite the disparities between Cameroon and other Africa countries, the current evidence from our study and that for the Cameroon 2004 DHS a rms that HIV prevention campaigns in Cameroon should pay more attention to those with higher level of education.HIV prevention programs should implement self-testing, rapid diagnostics technique, counseling, treatment and linkage to care policies that will encourage those with higher education level to participate without fear of stigmatization (29)(30)(31).
Our analysis found ethnicity to be associated with HIV.For instance, participants from the Bamilike/Bamoun ethnic groups were less likely to get HIV than those from the Arab-Choa/Peulh/Houssa/Kanuri.There was no clear statistically signi cant difference between HIV association with other ethnic groups in Cameroon.However, Odimegwu (32) showed that determinants of HIV stigma and discrimination vary by ethnicity in Nigeria and that signi cant ethnic differentials in HIV/AIDS stigma and discrimination by Secondary School education exist among the Hausa and Igbo ethnic groups.Further studies to understand the distribution of HIV among ethnic groups in Cameroon to ease the establishment of intervention programs at the level of ethnicity, customs, tradition, and beliefs are required as the descriptive analysis demonstrated disparities among the ethnic groups though this study could not account for the differences within ethnicity.
HIV was positively associated with region.Participants living in the Far North were 76% less likely to have HIV than those living in Adamawa.Living in a particular region could not be demonstrated to be associated with the odds of having HIV in our study.Other socioeconomic characteristics and their distribution among other factors like the rapid growth of urban cities could be the reason while HIV prevalence varies across the different regions and not necessarily because of one staying there.Moreover, our multilevel model did not nd residences in these regions to be signi cantly associated with the odds of having HIV.Residing in an urban or rural residence has been however strongly associated with the odds of having HIV (9,10,33).Some studies argue that the positive association of urban residences with HIV prevalence was not due to increasing incidence, but improved access to HIV treatments, provision of services, and volunteering counseling in urban areas compared to rural areas (15).
Wealth index did not have any signi cant association with HIV prevalence in our model.The association of household wealth with HIV prevalence has observed mix debates (5,7).Fox (5), highlighted a body of evidence at the national and individual level of the association of the rich with high risk of HIV in SSA.This study also demonstrated some disparities of wealth within the different regions, urban and rural areas in Cameroon and other SSA.The differences were accounted for base on the distribution of other socioeconomic factors such as more employment opportunities in urban than in rural areas.Our descriptive study clearly supports this claim as HIV prevalence was highest among the rich, and those currently working and leaving in the urban areas.Jane(6) again demonstrated that wealth and HIV were related in Cameroon (DHS 2004) and Tanzania (DHS 2003).Preventive measures should also target the wealth gradient disparities in Cameroon for effective HIV/AIDs control.
We did not nd any statistically signi cant association of belonging to a particular religion and having HIV.Faust (7), found that a liating to a particular religion had signi cantly higher odds of low HIV-related knowledge among respondents with traditional religious beliefs in Nigeria and that it was necessary to consider awareness programs targeting these groups, an important characteristic that should be investigated in Cameroon.This study had the following limitations: First, the ndings are based on the Cameroon 2011 crosssectional study and may not present the current situation of some of the socioeconomic variables used although our ndings presented similar trends with previous studies in Cameroon.A recent data set or study is required to con rm the socioeconomic differences observed in our study.Second, our study did not consider participants with missing values for the socioeconomic variables used.HIV biomarkers data was only used for participants presented as HIV positive or negative.Those with intermediate test results or having HIV I and II were not included in the analysis because of missing values for some participants.This may have altered the ndings from studies that used the same datasets such as the DHS general report of the socioeconomic differences and HIV distribution in Cameroon.Despite these limitations, our ndings have made a modest contribution to the gap of knowledge of the association between HIV prevalence and the socioeconomic status and investigated on additional factors that were limited in previous studies in Cameroon for public health intervention programs.

Conclusions
HIV was prevalent in women than men and comparing with HIV negative individuals, HIV positive individuals attained higher level of education, were richer and residing in urban areas.Age, sex, region, education level, and ethnicity were signi cantly associated with the odds of having HIV.Our nding recommends for novel intervention programs such as counseling and self-testing that will target the various socioeconomic factors associated with the odds of having HIV for proper public health management of the disease.

Declarations
Tables

2. 1 .
Data SourceData for this study was obtained from the DHS website(https://dhsprogram.com)(13).The Cameroon DHS and Multiple Indicators Cluster Survey (DHS-MICS 2011) datasets were used and this survey was conducted from January to August 2011 by the National Institute of Statistics in collaboration with the Ministry of Public Health.The DHS has been asking questions about HIV/AIDS knowledge, attitudes and behaviors since 1988.Indicators include knowledge of HIV prevention methods, attitudes towards those living with AIDS, beliefs about AIDS transmission, and experiences with higher-risk sex, including paid sex.The DHS-MICS was implemented with nancial support from the Government of Cameroon, the U.S. Agency for International Development (USAID), the United Nations Children's Fund (UNICEF), the United Nations Population Fund (UNFPA), and the World Bank.Usually, some institutions provide their expertise in the implementation of the survey, in particular, the Centre Pasteur of Cameroon (CPC) for the HIV testing component.ICF International provided assistance through the MEASURE DHS program funded by USAID.Details on the DHS program, household cluster points sampling (Figure

Table 2 .
Factors with HIV identified by the multilevel logistic regression model