Laboratory Markers and Characteristics of New HIV Infections Among Adolescents in Zambia

We aimed to explore HIV RNA (ribonucleic acid) virologic levels greater than 1,000 copies/millilitre (ml), among HIV-positive adolescents aged 15–24 years, establish the spread of CD4 T- cell counts and inspect characteristics of adolescents presenting with new HIV infections, including co-infections with Hepatitis B virus and syphilis. We analysed data from the Zambia Population-based HIV Impact Assessment 2016 survey. Two-stage stratied cluster probability sample design was used to select the target population. Our study truncated the population to focus on the age-group 15–24 whose biomarker tests and household information were complete. Our primary outcome measure was “New HIV-positive Infections among 15-24-year-olds” dened as HIV-positive biomarker samples presenting with HIV RNA ≥ 1,000 copies/ml without detectable ARVs. We tested associations between new HIV infections and clinical characteristics using negative binomial models adjusting for age, sex, education, marital-status, residence among several covariates. baseline levels of viral load and low CD4 + T-cell count in recent HIV infections among adolescents indicate weak immune repertoire at rst diagnosis, increasing the risk of contagion. As the epidemic continues to spread within the adolescent population, HIV-infection will become more complex and greater proportions of adolescents will likely be infected by regular partners. This suggests growing need for interventions targeted at stable partnerships and intensied public health campaigns specic for adolescents. 15–19 years 1·4 0·71 (0·24–2·15). Controlling for socio-demographic factors shows that being married or cohabiting, having higher than secondary education, belonging to a higher than middle wealth quintile, and being a rural resident were associated with increased risk for HIV incidence. attributes, having previously tested for HIV and reporting one sex partner in the for young CD4 + T-cell count of 479 cells/mm3 (95% 445–513·6) range of between 500 cells/mm 3 and 1500 cells/mm 3 the target of depletion of CD4 + T-cells acutely constrains the capacity. HIV infects activated cells, the T-cells the of the to baseline CD4 + T-cell count, HIV RNA levels and a number of other determinants including and sex (48, 49). In corroboration, our nds that 63% of males aged 15–24 with CD4 + T-cell counts ≤ cells/mm 3 coincided with 59·3% of them having HIV RNA ≥ 50,000 copies/ml compared with 61% and 45·3% respectively, their equivalents. 64·5% and 57·1% of 20–24 and 15–19 separately, CD4 + T-cell count cells/mm results a large proportion of adolescents testing positive for HIV, especially males, have their system potency far deteriorated. In our ndings, and a signicant independent predictor greater CD4 + T-cell count (42). Other HIV


Abstract Background
We aimed to explore HIV RNA (ribonucleic acid) virologic levels greater than 1,000 copies/millilitre (ml), among HIV-positive adolescents aged 15-24 years, establish the spread of CD4 T-cell counts and inspect characteristics of adolescents presenting with new HIV infections, including co-infections with Hepatitis B virus and syphilis.

Methods
We analysed data from the Zambia Population-based HIV Impact Assessment 2016 survey. Two-stage strati ed cluster probability sample design was used to select the target population. Our study truncated the population to focus on the age-group 15-24 whose biomarker tests and household information were complete. Our primary outcome measure was "New HIV-positive Infections among 15-24-year-olds" de ned as HIVpositive biomarker samples presenting with HIV RNA ≥ 1,000 copies/ml without detectable ARVs. We tested associations between new HIV infections and clinical characteristics using negative binomial models adjusting for age, sex, education, marital-status, residence among several covariates.

Conclusion
High baseline levels of viral load and low CD4 + T-cell count in recent HIV infections among adolescents indicate weak immune repertoire at rst diagnosis, increasing the risk of contagion. As the epidemic continues to spread within the adolescent population, HIV-infection will become more complex and greater proportions of adolescents will likely be infected by regular partners. This suggests growing need for interventions targeted at stable partnerships and intensi ed public health campaigns speci c for adolescents.

Background
The human immunode ciency virus (HIV) infection is one common cause of adolescent hospitalization because of the continuum of opportunistic infections and high burden of protracted complications due to HIV/AIDS (1). UNAIDS reveals that in 2015, 260 000 [180 000-340 000] new HIV infections were diagnosed among adolescents in sub-Saharan Africa, estimated at 29 adolescents acquiring HIV every hour (2). Estimates further indicate that young girls aged 15-19 accounted for almost 80% of these recent infections (2). The proportion of young people living with HIV rose globally by 30% between 2005 and 2016 and those dying due to AIDS-related illnesses tripled, making it the only age group to have experienced an increase (3). Adolescents' risk of acquiring HIV is closely correlated, among other factors, with age at sexual debut, low condom use, low counseling and testing coverage, legal and structural barriers (2,4).
Irrespective of achievements so far made in responding to HIV, approximately 60,000 persons were diagnosed with new HIV infections in Zambia in 2015 comprising 50,000 adults and 8,900 children (2,5). Youth are the least likely than any other age group to be aware of their infection. For example, only 42% of adolescents (aged [15][16][17][18][19][20][21][22][23][24] in Zambia knew their HIV status compared to 67% among adults 15-59 years in 2016 (6). Also, population-based survey data from sub-Saharan Africa shows that only 9% of young men 15-19 compared to 13% of of their girls contemporaries had tested and received results for HIV in the last 12 months (7,8). Compounding the status quo are sex disproportions in the prevalence of HIV. Within the population aged 20-24, females exhibit higher prevalence levels than males (5.6% versus 1.8%) (5). Adolescent girls who reported being divorced, separated, or widowed had higher (13.4%) HIV prevalence than their currently married and never-married counterparts (6.2% and 4.8% respectively) (9). In spite of growing accessibility to effective HIV prevention tools, methods and substantial scaleup of HIV treatment, attainment of viral load suppression (VLS) remains distinctly low in the younger population: 34% in HIV-positive females and 35.7% among HIV-positive males aged 15 to 24 compared to 73.5% in HIV-positive women and 73% in HIV-positive men aged 45 to 59 years (10). This rate of progress towards slowing the incidence of HIV acquisition, bolstering accessibility to treatment, and stopping AIDS-related deaths particularly among adolescents demands expansion and scaled up in order to reach high-incidence locations and maximize impact (11).
Because several clinical studies of HIV infection (5,(12)(13)(14), often disaggregate the population aged 0-14 years as children and 15-49 years as adults, in so doing, the probability of missing distinctive features of recent infections among a special group of adolescents aged 15-24 years is high. The age range 15-24 years aligns with the concept that adolescence might best be considered as ranging from 10 to 24 years because the transition period from childhood to adulthood continues into the twenties (15), and occupies a substantive part of their life course and shifting patterns of health and wellbeing (16). For this reason, it has been proposed that when adolescence is perceived as the population of young people from 10 to 24 years (15), it would help create opportunities for adolescents to acquire valuable assets and capabilities relevant, among others bene ts, to averting health risk (17). However, for this study, we consider young people aged 15-24 years as adolescents. Herein, we explore the HIV RNA (ribonucleic acid) virologic levels greater than 1,000 copies/millilitre (ml) with undetectable antiretrovirals (ARVs), among HIV-positive adolescents aged 15-24 years, determine the distribution of CD4 + T-cell counts and inspect characteristics of adolescents presenting with new HIV infections, including co-infections with Hepatitis B virus (HBV) and syphilis.

Study Design and Sample
The Zambia Population-based HIV Impact Assessment (ZAMPHIA) 2016 dataset, on which this study is based, was a cross-sectional survey, nationally representative as described in the ZAMPHIA 2016 report (5), the study utilized the two-stage strati ed cluster sample method, a probability proportional to size and an equal probability method to select the target population. The total sampled households were 13,441, comprising 5,205 eligible females and 4,337 eligible males aged 15-24. Among the eligible population aged 15-24 years, 88·1% (4,585/5,205) of women and 80·5% (3,491/4,337) of men completed the interview. The response rate for biomarker testing was 90.6% (4,716/5,205) for females and 90·0% (3,903/4,337) for males. Therefore, the sample size for our study was 7, 320 comprising 4,165 girls and 3,155 boys aged 15-24 who undertook biomarker testing in the ZAMPHIA 2016 study.

Field-based Biomarker Testing
All results for eld-based tests were provided to clients and those testing positive were referred for treatment. As detailed in the ZAMPHIA 2016 report (5), the following tests were conducted in the eld.
HIV Home-Based Testing and Counselling (HBTC) -This was performed on all sampled households. First, a screening test using Determine™ HIV-1/2 (Abbott Molecular Inc., Des Plaines, Illinois, United States) was performed. Participants diagnosed with a non-reactive result were reported HIV-negative. A con rmatory test using Uni-Gold™ (Trinity Biotech, plc. Wicklow, Ireland) was conducted for participants with reactive test results (18).
Hepatitis B Testing -As with HBTC, testing for HBV was done in every sampled household on individuals of all ages. The serological hepatitis B surface antigen rapid diagnostic test, Determine™ HBeAg, was used to determine acute or chronic HBV infection (5).
Syphilis Testing -All sampled participants in the age range 15-59 were tested for syphilis on DPP Syphilis Screen and Con rm Assay (Chembio, Medford, NY) in order to concurrently detect antibodies against non-Treponemal and Treponema pallidum antigens. Further con rmatory test was done on SD BIOLINE Syphilis 3.0 (Abbott Molecular Inc., Chicago, Illinois, United States) (5).

Laboratory-based Biomarker Tests
The following tests were performed in reference laboratories in Lusaka and Ndola.

Estimating New HIV Infections
To facilitate estimation of new HIV infections, the ZAMPHIA made use of two laboratory-based testing algorithms. 1) HIV-1 LAg Avidity enzyme immunoassay (EIA) (Serodia Biosciences Corporation, Portland, Oregon, United States) and VL (see Fig. 1) HIV-1 LAg Avidity EIA, VL, and ARV detection (5). Samples for which the median normalized optical density (ODn) was ≤ 1.5 were categorized as newly infected specimens precipitating VL testing. Following this investigation, outcomes of VL < 1,000 copies/ml were labelled as old infections and VL ≥ 1,000 copies/ml -new infections (5). When the ARV-controlled algorithm was considered, samples with VL ≥ 1,000 copies/ml in which the presence of ARVs was detected were also grouped as old infections. Conversely, samples for which VL ≥ 1,000 copies/ml with absence or undetectable ARVs were then categorized as new infections (5,19). The serological testing algorithm for recent HIV seroconversion (STARHS) estimates the immunological response against the virus on the basis of certain HIV antibody concentration, proportion (BED), isotype or avidity (20, 21). The time lapse from the point of acquisition (when antibodies can be detected) to the cut-off value de ning con rmed infection status, or the window period, should be de nitively established which is fundamental to the STARHS assays' capacity to specify the rate of incidence for a given population (20, 22).

Measures
The study aimed to nd evidence to answer the following questions; Among the population aged 15-24 years, what proportion of HIV-positive samples indicated HIV RNA ≥ 1,000 copies/ml with undetectable ARVs? What proportion of HIV-positive specimens with HIV RNA ≥ 1,000 copies/ml without detectable ARVs, tested positive for HBV and syphilis antibody test? What is the CD4 + T-cell count distribution of HIV-positive samples with HIV RNA ≥ 1,000 copies/ml without detectable ARVs, for the population 15-24 years? The CD4 + T-Cell count pro les how the immune system is functioning. The higher the T-Cell count, the better. As HIV infection progresses, the number of T-cells falls. The standard range for CD4 + T-cells runs between 500 and 1,500 cells/mm 3 (Pantaleo & Fauci, 2005). However, most opportunistic infections (OIs) are found among patients with CD4 counts < 200 cells/mm 3 (23). To investigate other determinants of new HIV infection in adolescents, our study endeavoured to answer the following questions. What underlying factors are associated with new HIV infections in adolescents. (e.g. new HIV infections with HBV and/or active syphilis antibodies)? Are there differences in new infections in adolescents depending on their background characteristics? A priori, we expected associations between new HIV infections in the population and the predictor variables. Our primary outcome measure (recent HIV infections in the population 15-24) was therefore, regressed on the four main predictor variables (age, sex, positive syphilis test and positive HBV test) across all models.

Outcome
The outcome variable was "New HIV-positive Infections among 15-24-year-olds" de ned as HIV-positive biomarker samples with HIV RNA ≥ 1,000 copies/ml without detectable ARVs". It was a count variable. The focus for analysis was on samples that tested positive with HIV RNA ≥ 1,000 copies/ml with undetectable ARVs as these samples were classi ed as new infections. HIV-1 RNA is the response marker for antiretroviral therapy (ART). HIV RNA tests measure the amount of HIV in the blood. A low RNA means a person is less likely to transmit HIV. A patient's pre-ART RNA level and the extent of RNA reduction following commencement of ART offers predictive insight about the prospects of disease progression (24).

Predictor
Predictor variables were clustered into laboratory markers (CD4 + T-cell count distribution, testing positive for hepatitis B virus and syphilis antibody test) and; demographic parameters (age and sex). The two laboratory markers, presence of hepatitis B virus and syphilis, were appropriate for assessing co-infection in HIV-positive individuals. HIV-positive persons who are concurrently positive for hepatitis B 'e'-antigen (HBeAg) usually present with higher hepatitis B viral load and fail to respond to antiviral treatment as positively as those with HBeAg-negative hepatitis B (25). Similarly, syphilis demonstrates an adverse impact on HIV infection, often revealing increases in RNA and corresponding decreases in CD4 + T-cell counts during active syphilis infection. Thus, individuals with HIV-syphilis co-infection are at increased risk of neurological complications and treatment failure (26).

Covariates
The third set of variables we considered were socioeconomic covariates deemed as in uential determinants of new infections. They included marital status, education [highest level achieved], residence [rural, urban], wealth index, HIV testing history, and awareness of one's HIV-positive status.

Statistical Analysis
We summarized incidence of new HIV infections and immunological pro les according to demographic and socioeconomic characteristics. The likelihood ratio chi-square χ 2 test aimed at testing whether the model containing the full set of predictors ts signi cantly better than a null (intercept only) model.; differences between predictor variables were assessed using Pearson chi-square χ 2 test of independence or Fisher's exact test for variables with an expected frequency of cells of ve or less (e.g. education). Continuous predictors were analyzed with a Student's t test for normally distributed variables, and the Wilcoxon-Mann-Whitney test (for age) or the Kruskal-Wallis test (for wealth quintile) for variables that did not have normal distributions. Univariate and bivariate analysis of variables reports proportions and binomial exact output at 95% con dence intervals (CI).
The number of new HIV infections is a count variable and closely follows a poisson distribution. We modelled the number of recent infections among young people using negative binomial regression. The negative binomial relative to count models such as the poisson or zero-in ated models, was considered suitable because the outcome variable was over-dispersed and did not have so many zeros. We further used generalized estimating equations and robust standard errors to determine factors independently and mutually associated with new HIV infections. When exponentiated, negative binomial regression coe cients provide the ratio of expected count per unit increase in exposure, referred to as the incident rate ratio (IRR) (27). We started by examining independent associations of each predictor variable and covariate (one at a time) with the outcome variable -new HIV infections (model 1). In model 2, all predictors and covariates with a P-value of up to 0.2 in model 1 were included. Associations were adjusted for potential confounders: education, marital status, wealth quintile, awareness of HIV-positive status, residence (rural/urban), number of sex partners in past 12 months and history of HIV testing. Model 3 considered all variables that were statistically signi cant at p ≤ 0.05 in model 2. Probabilities for removal and entry of predictor variables into the models were set at p-values of 0.20. IRRs at 95% con dence intervals are reported, and p-values less than 0.05 were considered statistically signi cant. All statistical analysis was done in STATA 14.2 software (28).

Knowledge of HIV-positive Status among Newly HIV Infected Adolescents
Even if 85·5% of participants did not know they had HIV, older adolescents were relatively less aware ( Figure:4b shows that adolescents in urban areas were less inclined to knowing they had the virus at 50% (83/166; 95% CI: 42·2-57·8) than those residing in rural areas (35·5%, 95% CI: 28·3-43·3). It examines association between each predictor variable and the outcome variable. † †Model 2 is a full controlled model with robust standard errors (SEs) in which inclusion into the model for all variables and covariates was contingent on passing the 0.2 p-value inclusion criterion in model 1are included. Hence, the variable, aware of one's HIV status was excluded. † † †Model 3 with robust SEs estimation include only statistically signi cant covariates. The restriction is relaxed for the four main predictors. Three variables, wealth quintile, ever tested for HIV and aware of one's HIV-positive status are not included. We found evidence that incidence of new HIV infections is signi cantly different between males and females. Increased likelihood of new HIV infection was signi cantly associated with adolescent girls than boys (IRR 2·30; 95% CI: 2·27-2·28), p < 0·001; Table 2, model1).

Discussion
Population-speci c studies on new HIV infections are of great public health importance because they offer policy makers and planners de nitive insights into the prevalence of the disease burden otherwise shrouded in population aggregation. Studies like ours aid in fully describing the epidemic, monitoring transmission patterns and prioritizing HIV prevention efforts targeted to speci c groups such as the adolescent population. Controlling for socio-demographic factors shows that being married or cohabiting, having higher than secondary education, belonging to a higher than middle wealth quintile, and being a rural resident were associated with increased risk for HIV incidence. Among behavioral attributes, having previously tested for HIV and reporting one sex partner in the past 12 months were associated with increased likelihood for infection.
Our study was not short of limitations. Accuracy in estimating HIV incidence and acute HIV infection in cross-sectional studies is sharply queried for suboptimal performance. The EIA and BED assays used in this study in the determination of recent HIV infections are used only on HIV seropositive specimens. This method employed the Serologic Testing Algorithm for Recent HIV Seroconversion (STARHS). However, this methodology and serologic assays therein have previously shown to contain substantial limitations, including biological, epidemiological, and statistical confounders (33). Additionally, although the EIA and BED assays used in our study are widely known, acceptance and use of these assays has been intensely disputed for its tendency to overestimate incidence (20). Therefore, we, may have overestimated the incidence of new HIV infections in adolescents and resulting associations with other parameters. Although we controlled for socioeconomic factors that have previously been linked to HIV prevalence; it is possible that residual confounding remains for parameters not considered in model estimation.
Future studies should consider expanding confounders to augment appropriateness in determining factors associated with new HIV infections in sub-population groups. Nevertheless, granting that our study focuses on the population 15 to 24 years, the method used for new HIV infection estimation draws its strength from global recommendations stating that tendencies in prevalence among adolescents 15 to 24 years be utilized as proxy measures for calculating HIV incidence (21,34). This is because sexual debut in this age group is expected to be recent such that prevalence closely re ects recent infections.
During the rst few weeks following HIV-1 seroconversion, HIV RNA viral load surges, which poses considerable risk of HIV transmission (35,36).
Our study results show that the mean HIV RNA was 164,183 copies/ml (64,886 − 263,480 copies/ml, Fig. 2b.), three times more than the 50,000 copies/ml HIV RNA limit in someone not taking treatment. Which suggests that newly infected adolescents, on average, had high HIV RNA levels at point of rst diagnosis. There is strong evidence from research that disease progression is substantially escalated in patients with HIV-1 RNA levels > 100,000 copies/ml, regardless of CD4 + T-cell count (37). Strongly associated with baseline viral load levels is treatment e cacy and response to therapy. HIV RNA levels greater than 150,000 copies/ml correspond to 1.5 times increased likelihood of treatment failure which is the ability to decrease the viral load to less than 50 copies/ml (37).
Variations in viral load between males and females were also clearly noticeable. In congruence with several studies (38- copies/ml than males, model estimation results also con rm signi cant independent associations for both age and sex. Increasing age was positively associated with new infections (P < 0·001, model 1), although the correlation with age was abated in controlled models (model 2 & 3).
Irrespective of viral load levels, female adolescents were 2·3 times (P < 0·001, model 1) at increased risk of HIV seroconversion in relation to males. These differentials are substantively documented in previous studies highlighting up to 50% lower HIV RNA viral load and higher CD4 + Tcell counts in HIV-1 infected women than men soon after seroconversion (39,40,(42)(43)(44). However, other studies con rm attenuation of the sex effect in advanced stages of infection (41, 45). Moreover, increasing age at seroconversion has been associated with increased risk of speedy immunologic deterioration and high virologic replication (35,37). Touloumi and colleagues veri ed that older age at the time of HIV antibody seroconversion was associated with shortened period to AIDS indicated by steepest (most negative) HIV RNA level and CD4 + T-cell slopes in the younger population groups like adolescents (46).
Extensive loss of mucosal CD4 + T cells occurs in the early stages of acute HIV infection, once this biomarker of immunologic potential falls below 500 cells/mm3, much of the immune reserve is wrecked and infected persons become susceptible to opportunistic infections. In this study, we found that 62% (95% CI: 54·65-69·46) of recently HIV infected adolescents had CD4 + T-cell count below 500 cells/mm3 with half of these young people aged 15-24 years diagnosed with CD4 + T-cell count of 479 cells/mm3 (95% CI: 445-513·6) from the normal functional range of between 500 cells/mm 3 and 1500 cells/mm 3 . As the primary target of HIV, depletion of CD4 + T-cells acutely constrains the host response capacity. HIV infects activated cells, causing the T-cells directed against the virus to be at greatest risk of infection (35,47). From the point of acquisition to AIDS, disease progresses has formerly been linked to baseline CD4 + T-cell count, HIV RNA levels and a number of other determinants including age and sex (48,49). In corroboration, our analysis nds that 63% of males aged 15-24 with CD4 + T-cell counts ≤ 500 cells/mm 3 coincided with 59·3% of them having HIV RNA ≥ 50,000 copies/ml compared with 61% and 45·3% respectively, of their female equivalents. Additionally, 64·5% and 57·1% of adolescents aged 20-24 and 15-19 years separately, had CD4 + T-cell count below 500 cells/mm 3 . These results imply that a large proportion of adolescents testing positive for HIV, especially males, already have their immune system potency far deteriorated. In line with our ndings, Bosch and colleagues found that younger age was a signi cant independent predictor of greater CD4 + T-cell count (42). Other studies have veri ed that age at seroconversion and HIV RNA level are associated with the CD4 + T-cell count at baseline and its subsequent slope to the ultimate disease syndrome (36,46,47). Further estimates highlight that within two years of contracting HIV, older individuals and those with the highest HIV RNA levels during early infection experience the most severe depletion of CD4 + T-cell s (43).
Co-infection, particularly with HBV, was one factor associated with incidence of HIV in young people. We identi ed that 2% (95% CI: 0·41-5·70) of new HIV infections in the population 15-24 years also tested positive for HBV. In our confounder-adjusted models, HBV -positive adolescents were 8·6 times (P < 0·001, model 2) at increased risk of new infection relative to HBV -negative adolescents in the same age stratum. Because HBV is more infectious and adolescence is a period of increased sexuality, most infections have been found to occur in adolescents and young adults (50). Our ndings align with several studies in Africa that have established increased Hepatitis B virus vulnerability in HIV-positive persons (51)(52)(53)(54)(55). Clinically, HBV belongs to a variety of heterologous viruses that have been shown to enhance HIV replication (36). Therefore, individuals who are HIV-positive and also test positive for HBeAg are more likely to transmit both viruses (25,50). The HIV-HBV co-infection augments risk of morbidity, antiretroviral therapy-related hepatotoxicity and mortality beyond those caused by either infection alone (56, 57). Similar to HBV results, syphilis co-infection was also estimated at 2%. Adolescents who tested positive for syphilis antibodies were 1.2 times (P < 0·84, model 2) more likely to contract HIV even though correlation was not statistically signi cant. Both HBV and syphilis co-infections were observed to occur at HIV RNA > 10,000 copies/ml RNA suggesting increased susceptibility during acute infection. Syphilis has been associated with high-risk sexual behaviour, increasingly prevalent in adolescence (58) and reported to be one of the more frequently occurring infection among in HIV infected people (54).

Conclusion
Our study suggests new HIV infections in adolescents are rst diagnosed at high HIV RNA and low CD4 + T-cell count increasing the risk of transmission and likelihood of treatment failure with reduced ability to achieve viral suppression. Sex differentials highlight disproportionate susceptibility of female adolescents to infection than males. Additionally, confounder-adjusted models ascertained that adolescents who were married or cohabiting, were positive for hepatitis B virus, were rural residents and had attained higher than secondary education emerged as the strongest correlates of new infections. While the epidemic continues to spread within the adolescent population, HIV transmission and acquisition will become more complex and greater proportions of adolescents will be infected by their regular partners, especially rural ones. Which implies an increasing need for interventions targeted at stable partnerships, intensi ed public health campaigns speci c for rural adolescents and population segmented preventive health services. However, adolescents being the less prioritized population group in health response and routinely overlooked in national plans, the advent of the COVID-19 pandemic poses a major threat to reversing minimal preventive gains so far achieved in adolescent HIV prevention.

Abbreviations
ART Antiretroviral Therapy CD4

Declaration of interests
All authors declare no competing interests.

Data sharing
Data used in this article are available to bona de researchers on request from Zambia's Ministry of Health (www.moh.gov.zm) through email (info@moh.gov.zm).

Funding
No funding was received to produce the manuscript.
Authors' contributions TNM and NM developed the manuscript concept. TNM and NM sourced data, re-coded data, conducted statistical analysis and interpreted analysis results. TNM drafted initial manuscript. XZ, SK, CG, XQ and KS reviewed methods, analysis results and interpretation. All authors edited and approved the nal content of the manuscript before submission.

Ethics declaration
Since this study was based on secondary individual de-identi ed dataset, no ethical approval was required. However, ethical approval for the protocol for ZAMPHIA Population-based survey on which the study is based was obtained from the Tropical Diseases Research Centre (TDRC) Zambia (FWA00003729), the Centers for Disease Control and Prevention (CDC), Columbia University IRB (FWA # 00002636), and WESTAT (FWA # 00005551).

Consent for publication
Not applicable