Intersection of HIV and Anemia in women of reproductive age: a 10-year analysis of three Zimbabwe demographic health surveys, 2005–2015

DHS uses a two-stage cluster sampling procedure with a stratified sample using probability proportional to size, of enumeration areas (EAs) selected at the first stage and a randomly selected household within the selected EAs for the second stage. All women aged 15–49 in the selected household are interviewed [17].

According to the Zimbabwe Census and Statistics Act, censuses are held every 10 years. Each DHS cycle used the most recent census of non-institutionalized population as the sampling frame excluding those living in army barracks, hospitals, police camps, and boarding schools. The DHS sample was designed to yield representative information for most indicators for the whole country, for urban and rural areas, and for each of Zimbabwe’s 10 provinces or sub-national regions: Manicaland, Mashonaland Central, Mashonaland East, Mashonaland West, Matabeleland North, Matabeleland South, Midlands, Masvingo, Harare, and Bulawayo (Fig. 2a). The three Mashonaland Provinces are in highveld, the rest are in drought-prone lowveld (Fig. 2b). Administratively each province is further divided into districts, each district into smaller administrative units called wards, and each ward into convenient areas (EAs). In 2015 DHS, 400 EAs were selected (166 in urban areas and 234 in rural areas) from a total of 29,365. A sample of 28 per EA with a typical size of 101 households were selected for a total of 11,196 private households.

Sampling weights are computed for the households and women selected for the study based on the probability of selection of each household, and response rates for households and individuals. Sampling weights are computed for the households and women selected for the study based on the probability of selection of each household, and response rates for households and individuals.

DHS computes sampling weight for a particular household as the inverse of its household selection probability multiplied by the inverse of the household response rate in the stratum and for an individual woman for a particular household as the inverse of its household selection probability multiplied by the inverse of the household response rate in the stratum [17]. Additional sampling weights are computed for sample subsets, biomarkers or HIV testing. Finally, the weights are normalized by dividing each weight by the average of the initial weights (equal to the sum of the initial weight divided by the sum of the number of cases) so that the sum of the normalized weights equals the sum of the cases over the entire sample. Using these sampling weights in the analyses allows investigators to generalize results to the target population. Furthermore, DHS has protocols in place to protect the privacy of the surveyed households and participants [17]. Detailed of the methods used to create weights are provided in the DHS manual (https://​dhsprogram.​com/​pubs/​pdf/​DHSG1/​Guide_​to_​DHS_​Statistics_​DHS-7.​pdf accessed on October 2, 2020).

DHS surveys collect data through four main questionnaires: Household questionnaire, Woman’s Questionnaire, Men’s Questionnaire, and Biomarker Questionnaire [17]. The Household Questionnaire is used to collect information on characteristics of the household’s dwelling unit and characteristics of usual residents. It is also used to identify members of the household who are eligible for an individual interview. Eligible respondents are then interviewed based on their gender using an Individual Woman’s to obtain their characteristics including age, education, and geographic location (urban vs. rural). The Biomarker Questionnaire is used to collect biomarker data. The household roster within this questionnaire captures socio-demographic characteristics of each member. The resultant dataset filenames, variable types, names, and coding are uniform across cycles [17]. The questionnaires are updated each cycle to add new questions, but the core questionnaire is barely changed from cycle to cycle. Detailed information regarding DHS questionnaires are provided in the DHS manual (https://​dhsprogram.​com/​pubs/​pdf/​DHSG1/​Guide_​to_​DHS_​Statistics_​DHS-7.​pdf accessed on October 2, 2020).

For each DHS cycle, the most recent population census was used as the sampling frame. According to the Zimbabwe Census and Statistics Act, censuses are held every 10 years. The 2012 Zimbabwe Population Census was used as the sampling frame for the 2015 Zimbabwe DHS. The DHS sample was designed to yield representative information for most indicators for the whole country, for urban and rural areas, and for each of Zimbabwe’s 10 provinces or sub-national regions: Manicaland, Mashonaland Central, Mashonaland East, Mashonaland West, Matabeleland North, Matabeleland South, Midlands, Masvingo, Harare, and Bulawayo. The three Mashonaland Provinces are in highveld, the rest are in drought-prone lowveld (Fig. 2a and b). Administratively each province in Zimbabwe is further divided into districts, and each district is divided into smaller administrative units called wards. Each ward was subdivided into convenient areas, which are called census enumeration areas (EAs).

In 2015 DHS there were 29,365 EAs from which 400 were selected (166 in urban areas and 234 in rural areas). A typical EA size was about 101 households from which a sample of 28 households per EA were selected for a total sample of 11,196 households [17]. The sample was selected with a stratified, two-stage cluster design, with EAs as the sampling units for the first stage. The 2015 DHS cycle sample, for example, included 400 EAs-166 in urban areas and 234 in rural areas. The second stage of sampling included the listing exercises for all households in the survey sample. A complete listing of households was conducted for each of the 400 selected EAs in March 2015 [17]. Maps were drawn for each of the clusters and all private households were listed. Sampling weight and wealth index were derived from the Households Questionnaire. The listing excluded institutional living arrangements such as army barracks, hospitals, police camps, and boarding schools.

To be eligible for anemia test, a woman of age 15–49 years should be a usual resident or slept in the household the night before the household interview. After consenting a health specialist records all data related to biomarkers. Hemoglobin analysis is conducted immediately on the spot using a battery-operated portable HemoCue® analyzer (HemoCue AB, Angelhom, Sweden) that does not require trained phlebetomists or sample storage, producing a result in less than 1 minute [17, 18]. Testing is voluntary and respondents receive results immediately, as well as information on anemia prevention [19]. Anemia was defined as hemoglobin < 11.0 g/dL and < 12.0 in pregnant and non-pregnant women, respectively. Severe anemia was defined as < 7 g/dL; moderate 7–9.9; and mild 10–11.9 [2, 3]. All women with moderate to severe were referred for follow-up care [17]. .The Biomarker Questionnaire and Biomarker Field Manual are provided at (https://​dhsprogram.​com/​pubs/​pdf/​DHSG1/​Guide_​to_​DHS_​Statistics_​DHS-7.​pdf accessed on October 2, 2020).

Optimal sample size for multi-stage stratified design like a two-stage cluster sampling design is dependent on intra-cluster correlation [17, 18, 20]. To estimate the prevalence of anemia for ZDHS 2015, it was assumed the background prevalence was 0.268, SE = 0.006, design effect = 1.388. A weighted sample of n = 9235 is enough to estimate with at least 95% confidence the prevalence of anemia to lie between 0.255–0.281, i.e., a margin of error (E) of 0.013. A similar approach was used for the DHS Cycles conducted in 2005 and 2010.

HIV Testing: Blood specimen collection and analysis was based on the anonymous linked protocol developed for The DHS Program that allows HIV test results to be merged with the socio-demographic data collected in the individual questionnaires after the record is de-identified [17, 21]. The ZDHS biomarker interviewers explained the blood collection procedure, the confidentiality of the data, and the non-availability of the test results to the respondent.

HIV testing protocol provides for informed, anonymous, and voluntary testing making it difficult to provide the results to participants. Eligible women are identified on the Household Questionnaire (if using paper questionnaires) or Biomarker Data Form (if using electronic questionnaires).

Height and weight were measured and body mass index (BMI; kg/m²) calculated.

The DHS uses Principal Component Analysis to construct the household wealth index using a composite measure of a household’s cumulative living standard based on ownership of selected assets, such as televisions and bicycles or a car; materials used for housing construction; and source of drinking water, toilet facilities, and flooring materials. (www.​dhsprogram.​com/​programming/​wealth%20​index/​Steps_​to_​constructing_​the_​new_​DHS_​Wealth_​Index.​pdf). The resulting asset scores are standardized in relation to a standard normal distribution with a mean of 0 and a standard deviation of 1. Individuals in the same household are assigned the same wealth index score. These standardized scores are then used to create the break points that define wealth quintiles as: Lowest, Second, Middle, Fourth, and Highest [17, 21].

Of the 7634, 8169, and 9235 women surveyed in 2005, 2010, and 2015 DHS cycles, the crude prevalence (95% CI) of anemia was 37.8(35.9–39.7), 28.2(26.9–29.5), and 27.8(26.5–29.1), respectively. The mean age for women across the 3 cycles was 28 years and 76, 74 and 72% of women in each cycle were younger than 35 years (Table 1). The prevalence of HIV decreased from 20.4% in 2005 to 16.3% in 2015. A typical woman in each cycle had a BMI of 24 kg/m², however, overweight and obesity increased by 10% over the 10-years, from 25.8% in 2005 to 31.8 and 35.4% in 2010 and 2015, respectively. Underweight (BMI < 18 kg/m²), decreased from 5.6 to 4.1% and 3.5%, in 2010 and 2015, respectively (Table 1).

The distribution of anemia status within participant characteristics in each DHS cycle is presented in Table 2. Prevalence of anemia decreased in all women and was generally higher (> 41%) in women aged 30+ years in 2005 than those under-30 (< 37%). HIV-positive women across the 3 cycles had 20% higher prevalence than HIV-negative women. There was an inverse relationship between BMI categories and prevalence of anemia. For underweight women (BMI < 18 kgs/m²) the prevalence decreased from 41.4% in 2005 to 28.0% in 2015, whereas in obese women (BMI > 30 kg/m²) the prevalence decreased from 33.6 to 23.4% for the same period. Matebeleland South and Bulawayo, and Midlands had the largest burden of anemia across the 3 cycles. Notably, while 9 provinces experienced declines in prevalence between 2005 and 2015 ranging − 5.7% to − 13.5%, the prevalence in Matebeleland South (45.0% in 2005; 44.6, 43.1% in 2015) the prevalence remained essentially unchanged. Mashonaland West (− 22.6%), Mashonaland Central (− 13.5%), Mashonaland West (− 11.9%) and Mashonaland East (− 11.3%) had > 11% declines in prevalence (Table 2).

Table 3 shows multivariable adjusted ORs and 95% CIs for the association of anemia with participant characteristics within each DHS cycle. In 2005, there was a significant association between anemia and women in older age-group [40–44: 1.66(1.28–2.16),; 45–49: 1.80(1.35–2.41)] which did not exist in 2010 and 2015 (Table 3). When restricted to HIV negative women, the odds for anemia were elevated in 2005 [40–44: 1.71(1.34–2.20); 44–49: 1.88(1.37–2.60)]. A negative linear trend between the odds for anemia and age-groups was observed in 2005 but not in 2010 or 2015.

Similar analyses are shown by subgroup status in Supplemental Table 1 (restricted to HIV negative women), and Supplemental Table 2 (restricted to HIV positive women). There was no association between age-group and anemia in HIV negative women in 2010 and 2015 (Supplemental Table 1). Upon restriction to HIV positive women, results show an association between anemia and age-group 20–29 [1.41(1.00–1.99)] in 2010. In 2015 results show elevated odds for anemia ranging from 43 to 57% higher in women age 25–44 years. Specifically, adjusted ORs by age-group were 1.57(1.09–2.25) for women 25–29, 1.43(1.04–1.95) for women 30–34, 1.50(1.00–2.24) for women 35–39, and 1.68(1.16–2.45) for women 40–44 (Supplemental Table 2).

In all three DHS cycles, HIV positive women had greater than 2-folds the odd for anemia: OR = 2.40(2.03–2.74) in 2005, 2.35(1.99–2.77) in 2010, and 2.48(2.18–2.83) in 2015 (Table 3).

Pregnant or breast-feeding women had elevated odds for anemia in 2005 and 2015 with adjusted OR of 1.31(1.16–1.47) in 2005 and 1.23(1.09–1.34) in 2010 (Table 3). When restricted to HIV positive women only, the adjusted odds ratios were not elevated (Supplemental Table 2). However, when restricted to HIV negative women, the adjusted odds ratios were OR = 1.43(1.296–1.58) in 2005, and 1.28(1.15–1.43) in 2010. [peak of economic meltdown, and successive drought) (Supplemental Table 1). Not taking iron supplementation during pregnancy was associated with higher odds for anemia in all 3 cycles: OR = 1.17(1.03–1.33) in 2005, OR = 1.23(1.09–1.40) in 2010, and OR = 1.24(1.08–1.42) in 2015 (Table 3). Similarly, not taking iron supplementation elevated the odds for anemia in HIV negative women in all 3 cycles OR = 1.16(1.02–1.33) in 2005, OR = 1.22(1.07–1.39) in 2010, and OR = 1.26(1.07–1.48) in 2015 (Supplemental Table 1). We explored the interaction of anemia and taking iron supplementation with anemia in HIV positive women, the interaction term did not attain statistical significance.

While results presented earlier suggest negative effects observed between the prevalence of anemia and independent variables, the association between anemia and BMI were indicative of a protective balance. Being obese had a protective effect was against anemia with odds ratios for anemia of OR = 0.68(0.54–0.86) in 2005, OR = 0.61(0.52–0.72) in 2010, and OR = 0.75(0.65–0.88) in 2015, respectively (Table 3). In all the 3 cycles there was a significant negative linear trend in the odds for anemia with BMI (all 3 p-value for trend < 0.0001). The protective effect of obesity was also observed in both HIV negative and HIV positive women (Supplemental Table 1 and Supplemental Table 2).

Relative to women residing in urban locations, rural women had elevated odds for anemia in 2005 and 2010 but not in 2015 with odds ratios OR = 1.33(1.08–1.65) in 2005, and OR = 1.26(1.03–1.53) in 2010 (Table 3). After stratification by HIV status, the odds ratios of anemia in 2005 and 2010 were 0R = 1.26(1.01–1.58), and OR = 1.28(1.08–1.52), respectively. For HIV positive women, the odds were more elevated in 2005 comparatively, OR = 1.71(1.17–2.49). There were no differences in rural vs. urban odds of anemia in HIV positive women in 2010 and 2015, respectively (Supplemental Table 1 and Supplemental Table 2).

With the lowest prevalence of anemia of 30.7and 21.7% in 2005 and 2010, respectively, Manicaland was used as the referent province in the logistic regression analysis. In 2005, 6 provinces had significantly higher odds for anemia than Manicaland. The two provinces with the highest odds ratios were Masvingo, OR = 2.23 (1.90–2.62) and Matebeleland South, OR = 1.86(1.65–2.10). In 2010 the pattern flipped, and except for Matebaleland, which had higher odds for anemia, OR = 1.7-(1.34–2.17), four provinces Mashonaland Central, Mashonaland West, Masvingo, and Harare had significantly lower odds with odds ratios ranging from as low as 0.61(0.53–0.71) in Mashonaland West to 0.70(0.60–0.82) in Harare. In 2015 the pattern changed again, with four provinces having elevated odds ranging from the highest of 2.46(2.21–2.75) in Matebeleland South to 1.57(1.40–1.77) in Midlands, and the lowest being 1.48(1.26–1.74) in Harare (Table 3). Similar patterns were observed in the 3 cycles upon restricting the analysis to the HIV status of being negative (Supplemental Table 2) and/or being HIV positive (Supplemental Table 2).

The high anemia rate among women of child-bearing age is similar to findings from other LMICs. A meta-analysis reported an anemia rate of 42.7% (7.0, 48.4%) among pregnant women in LMICs [24, 25]. The 2015 burden of anemia in Zimbabwe (25.9% overall and 26.8% in pregnant or breastfeeding mothers) was substantially lower than the 39 and 46% (13.1 and 19.2% lower), respectively reported globally in 2016 [5]. A 2009 study of adult women across 40 DHSs from 32 countries found that although anemia rates do decrease as income increases, the decrease is modest [25]. We purposely did not account for education. While education and health infrastructure probably contribute to reduce the burden of anemia, such factors are likely to be positively correlated with wealth index. In our study, there was no statistically significant linear trend across quintiles of household Wealth Index. There was little or no difference in the prevalence of anemia between the lowest and highest quintiles of wealth index across the 3 cycles. This observation is reinforced by the assertion that even with a robust economy, it takes decades of sustained income growth to double national per capita income [25].

Zimbabwe will also fall short of the revised UN 2030 Sustainable Development Goal (SDG) Target 2.1 Goal 2.2 which aims to end all forms of malnutrition by 2030 (the UN Inter-Agency and Expert Group on the SDG Indicators) [12]. With only 5 years left to attain Resolution 65.6 (before 2025), and compounding this with the current prevalence of 27.8% (and 43.7% for HIV+), Zimbabwe will have to achieve declines equivalent to 2.8% (4.4% for HIV+) annually if the Global Nutrition Target is to be met. The flattening of prevalence experienced between 2010 and 2015 does not suggest this level of annual decline to be achievable or indeed sustainable; with this in mind, there could be other possible explanations for the stagnation in prevalence of anemia between 2010 and 2015.

Firstly, the 2005 to 2015 DHS cycles overlapped a period during which Zimbabwe was among the top 5 countries in SADC most burdened with HIV/AIDS [16]; it is also a period camouflaged by an extension of a serious economic downturn. Figure 1a shows 27-year trends of prevalence and incidence of HIV/AIDS in women 15–49 years for Zimbabwe, South Africa, Botswana and OECD countries, 1990 to 2017. Figure 1b shows the rates of prevalence and DALYs associated with dietary iron deficiency per 100,000 population in women 15–49 years for Zimbabwe, neighboring South Africa and Botswana, and the Organization for Economic Co-operation and Development (OECD) countries; between 1990 and 2017 iron deficiency anemia declined for the three countries and OECD countries, the prevalence and DALYs rates per 100,000 for Zimbabwe flattened after 2005, a pattern that mirrors our study findings. The patterns of HIV burden for Zimbabwe was a sharp contrast with the burden for Botswana and South Africa. The prevalence for Zimbabwe flattened at about 20% between 1995 and 2005 but increased for the neighboring countries. Similarly, incidence progressively declined in Zimbabwe from ~ 0.65% to ~ 0.1% unlike for Botswana and South Africa. Figure 1a and b suggest that Zimbabwe, unlike its neighbors and OECD countries, was entangled in an intersection of increasing to decreasing levels of HIV/AIDS burden coupled with the stagnant burden of dietary iron deficiency – a situation critical in giving context to our current study findings. Anti-retroviral treatment (ART) has improved survival and prevention of mother-to-child transmission of HIV. However, individuals with HIV infection are significantly more likely to develop anemia independently of HIV treatment [13]. Moreover, other studies suggest, many drugs used to treat HIV can also cause anemia [13, 14]. .In Zimbabwe, most individuals, 79% (75–83%) with HIV were on antiretroviral therapy (ART), a 30% increase, 49% (42–57%) from 2009. Note that the burden of HIV/AIDS in OECD countries was negligible (Figs. 1a and b). Our study showed that women 15 to 45 years who were HIV+, regardless of their breast-feeding or pregnancy status, had greater than 2.35 the odds for anemia than HIV- women. Our findings, taken together with the GBD study [26], provide associative but not causal evidence of elevated risk of anemia in HIV+ women. HIV status-specific analysis results shown in Supplemental Tables 1 and 2 revealed elevated odds for anemia for all characteristics between HIV positive and HIV negative women. Women 40–49 years of age without HIV have higher odds for anemia than similar women with HIV; women with BMI < 30 kg/m² with HIV had higher odds for anemia than similar women without HIV; household wealth index was a predictor for anemia in women without HIV but not in women with HIV; current pregnancy or breastfeeding, and not taking iron supplementation were protective in women without HIV but not significantly so for women with HIV. For policy, the findings in Supplemental Tables 1 and 2 suggest that screening for anemia by measuring hemoglobin level should be prioritized for all women especially if they have HIV. Furthermore, leaner women with BMI < 30 kg/m² with HIV should equally be prioritized for hemoglobin level screening for anemia.

Taken together, the above may explain our findings. Whether the HIV/anemia association is mediated by ART should be studied further. Since DHS 2005, 2010, and 2015 cycles not collect data on ART, we had no way to explore whether indeed the HIV/anemia association has been mediated by the ART factor. We therefore recommend, that future DHS cycles develop IRB modalities for providing individuals tested for HIV with their test results and immediately link those with positive results with appropriate care. Once a woman tests positive, it is logical to also offer the test to their spouse and if negative, the couple is offered information on how to live together as a discordant couple. Furthermore, notwithstanding that our study, because of its observational nature, is insufficient to establish causality of HIV and anemia, we recommend that all individuals with HIV should be provided iron supplementation as prevention for anemia. Pharmaceutical industries are encouraged to develop a once weekly, or once monthly iron-supplement tablet to keep the pill burden low for HIV positive individuals who could also be on ART – which, as pointed out earlier, do cause anemia [13, 14].

SADC countries experience more intense droughts linked to changes in El Niño/La Niña-Southern Oscillation patterns. This climate change pattern, that occurs across the tropical Pacific Ocean approximately every 5 years cause food shortages and food insecurity. Droughts occur annually in Zimbabwe. The three Mashonaland Provinces are in highveld, the rest are in drought-prone lowveld [27]. “see elevation map at: https://​ocw.​jhsph.​edu/​index.​cfm/​go/​find.​browse#images/​topicID/​16)”. The frequency and intensity of droughts have increased but severity and spatial distribution varies by province affecting livelihoods of the people largely because of the Zimbabwean economy’s dependence on agriculture [27]. The socio-economic impacts of drought are varied, ranging from famines, and food insecurity, caused by crop failure which could explain the high burden of anemia we observed [28]. The western regions, Matabeleland North and South Provinces, as well as Midlands provinces, with average rainfall less than 650 mm annually, are more likely to be affected by severe droughts. Droughts range from severe to mild, with average rainfall greater than 650 mm annually in the three Mashonaland provinces (East, West, and Central) to the north, and Masvingo province in the South East. In a drought year, the rain in these regions allow some form of crop production that is adapted to drought conditions like small grains [27]. Zimbabwe’s climate is projected to become more erratic, with some projections suggesting that widespread crop failure will occur every three out of 5 years. United States Agency for International Development (USAID) 2020 Climate Risk Profile Zimbabwe (2020) https://​www.​climatelinks.​org/​sites/​default/​files/​asset/​document/​2020_​USAID_​ATLAS_​CRP-Zimbabwe.​pdf) The Zimbabwean government has been relying on food aid and food redistribution [27]. Between 1991 and 2016, Zimbabwe experienced six moderate-to-severe droughts. The drought severity in Zimbabwe between 2005/2006 and 2009/2010 seasons, the 2007/2008 agricultural season was the “worst” (i.e., had the largest proportion of the country under drought) whilst the 2008/2009 agricultural season was the “best” (lowest proportion of the country under drought) [27]. These patterns match the high prevalence of anemia we observed in 2005 and the steep decline observed in the two later cycles.

Zimbabwe experienced the first hyperinflation of the twenty-first century for the period 2006–2010. At the peak (2008–2009) of this sharp economic downturn, Zimbabwe’s inflation was estimated at 79.6 billion percent month-on-month, 89.7 sextillion percent year-on-year by mid-November 2008 [29]. .In the history of the world, only Hungary (1946) had ever experienced such a worse stratospheric inflation [29]. Such inflation, primarily caused by political instability, created chaos in all spheres of the economy, food became scarce and health outcomes suffered substantially [30‐32]. Other external parameters such as the Zimbabwe Democracy and Economic Recovery Act of 2001, (ZIDERA) enacted into law by the Senate and House of Representatives of the United States of America in Congress (www.congress.gov/bill/115th-congress/senate-bill/2779/text) reversed any gains made in health and stifled affected countries from attaining the UN-led SDGs [32]. On paper, such economic sanctions ideally meant to target politicians, as our study and others alike show, have deeper and far-reaching health consequences and unnecessary loss of life affecting common people instead [31, 32].

The simultaneous tri-factor of a) adverse climate changes causing more frequent droughts, b) high prevalence of HIV infection and most people with HIV on ART, and c) political and economic sanctions created maelstrom favoring increased propensity for anemia among women of child-bearing age. Climate change caused food shortage and food insecurity leading to malnutrition; high prevalence of HIV infection created demand of HIV drugs some of which are known to trigger anemia; and a tough economic environment worsened by economic sanctions caused the healthcare system to collapse; all of which jointly generated a vortex for anemia.

The WHO Global Nutrition Targets 2025, Anemia Policy Brief offers a range of recommendations and actions for policymakers to consider in achieving the global nutrition targets of 50% [5]. These recommendations range from improving the identification, measurement scaling up coverage of prevention, control and treatment activities; partnerships between state and nonstate actors for financial commitment to implement comprehensive food and nutrition policies; developing programs beyond the health sector including agriculture and education sectors; and to monitor and evaluate the implementation of anemia control programs [5]. The policy brief also provides case studies for successful presentation of anemia in Viet Nam, Venezuela, and India. Others have recommended iron and folic acid supplementation, fortification of wheat and maize flours [5, 36, 37]. Although such programs have been found to be cost-effective, there are comparatively few large-scale delivery mechanisms to reach the rural poor, who are often reliant on their own or local food production and have limited access to health services [25]. Governments should increase awareness about the benefit of adhering to consumption of nutritional diet among people living with HIV/AIDS taking ART as consuming appropriate diet helps the body in proliferating enough amounts of red blood cells and other granulocytes [35]. The food industry should invest in healthy and inexpensive food. For women with HIV taking ART, hemoglobin level should be monitored on each of the follow-up visit and appropriate action taken to combat any detected anemia [35].

While some of these strategies may have been already adopted in Zimbabwe, our findings emphasize that more expansion in terms of this policy and other developing world issues are needed. For example, the Coronavirus Disease (2019) (COVID-19) pandemic which caught the world by surprise is a challenge affecting everyone. Of the of 534 COVID-19 patients admitted to hospitals in Spain, fewer than 1% (0.92%; 95% CI: 0·39–2.14) were HIV positive suggesting that HIV-positive people appear more likely to contract the new coronavirus or to become seriously ill [38]. There is anecdotal evidence that people with sickle cell anemia are more likely to have serious complications from COVID-19. Obesity predisposes individuals with SARS-CoV-2 infection to increased severity of the disease [39]. All stakeholders, i.e., the general population government authorities, and health care professionals should recognize that obesity is a chronic disease which requires a variety of effective prevention and management strategies both in response to COVID-19 and in the longer term. A comprehensive societal approach involving effective strategies will help protect people with obesity from COVID-19 complications, prevent a further increase in the prevalence of overweight and obesity [40]. The association of anemia and COVID-19 in the presence of HIV needs further investigation.

Our study findings suggest a strong association between being HIV+ and anemia, a fact devoid from the WHO Anemia Policy [5]. Given that only 5 years are remaining before setting the next anemia targets, it would be in the interest of the next policy to consider the inclusion of HIV in addressing interventions around anemia especially for populations of women in sub-Saharan Africa where HIV is endemic and significantly higher compared to the rest of the world. Another recommendation not quite related to our study findings is about the DHS HIV protocol which is designed to not disclose HIV test results with the participant. The protocol should be revised to allow for test results to be made available to the participants. Not sharing test results can be viewed as both unethical and a missed opportunity to treat infected participants before the disease gets worse.

It is believed that about 50% of all anemia cases are due to iron deficiency [41], and other major contributors include malaria [42]; infection with bacteraemia-causing organisms [43‐45]. Our findings reinforce the need for further studies to understand the impact of environmental factors on anemia burden at subnational scales to help in the design of cost-effective delivery of programs and interventions. Finally, as our results show, the state of anemia in Zimbabwe remains paradoxical: the anemia burden was lower (13.1% lower for all and 19.2% lower) than the global burden (WHO) regardless of Zimbabwe simultaneously experiencing a cocktail of events favoring increased propensity for anemia among women, all which were not experienced by other countries, at least not at the scale manifested in Zimbabwe [6, 7, 27‐32].