Multiple concurrent illnesses associated with anemia in HIV-infected and HIV-exposed uninfected children aged 6-59 months, hospitalized in Mozambique

Anemia is an increasingly recognized problem in sub-Saharan Africa. To determine the magnitude, severity, and associated factors of anemia among hospitalized children aged 6-59 months, HIV-infected and HIV-exposed uninfected children (a child born to a known HIV-infected mother) with a documented fever or history of fever within the prior 24 hours of hospital admission (N = 413) were included in this analysis. Of 413 children enrolled, 364 (88%) were anemic, with 53% classified as mild anemia (hemoglobin [Hb] 7-9.9 g/dL). The most common diagnoses associated with hospital admission included acute respiratory illness (51%), malnutrition (47%), gastroenteritis/diarrhea (25%), malaria (17%), and bacteremia (13%). A diagnosis of malaria was associated with a decrease in Hb by 1.54 g/dL (P < 0.001). In HIV-infected patients, malaria was associated with a similar decrease in Hb (1.47 g/dL), whereas a dual diagnosis of bacteremia and malaria was associated with a decrease in Hb of 4.12 g/dL (P < 0.001). No difference was seen in Hb for patients on antiretroviral therapy versus those who were not. A diagnosis of bacteremia had a roughly 4-fold increased relative odds of death during hospitalization (adjusted odds ratio = 3.97; 95% CI: 1.61, 9.78; P = 0.003). The etiology of anemia in high-burden malaria, HIV, tuberculosis, and poor nutrition countries is multifactorial, and multiple etiologies may be contributing to one's anemia at any given time. Algorithms used by physician and nonphysician clinicians in Mozambique should incorporate integrated and non-disease specific approaches to pediatric anemia management and should include improved access to blood culture. We conducted a prospective observational study of HIV-infected and HIV-exposed uninfected children in the cities of Maputo and Quelimane, Mozambique, who were prospectively followed up during their hospital stay. This “parent” study was designed to determine the incidence, etiology, antibiotic sensitivity patterns, and molecular characterizations of culture-confirmed bacteremia in representative rural and urban hospitals in Mozambique. All HIV-infected and HIV-exposed uninfected children aged 0–59 months with a documented axillary temperature of ≥ 37.5°C or rectal temperature ≥ 38.0°C, or a history of fever within the 24 hours before hospitalization between April 1, 2016 and December 31, 2018 were enrolled. Patients were recruited from the pediatric urgent care clinics of three hospitals in Maputo: Central Hospital Maputo, Hospital Jose Macamo, and General Hospital Mavalane; and two hospitals in Quelimane: Central Hospital Quelimane and General Hospital Quelimane. All are tertiary referral hospitals supported by the National Health System (Figure 1). Map of Mozambique with study hospitals identified for Maputo and Quelimane. This figure appears in color at www.ajtmh.org. Using data collected during the course of the parent study, we conducted this anemia analysis on our patient population aged between 6 and 59 months. Children were considered HIV positive if they had a documented proof of HIV infection by polymerase chain reaction (PCR) or HIV rapid antibody tests (if ≥ 9 months at time of test), or if they were taking ART in the absence of documented HIV test results. Children admitted with a fever, or a history of fever, and no documented history of HIV, but with documented maternal HIV exposure by self-report, were offered a PCR (if < 9 months old) or an HIV rapid antibody test (if ≥ 9 months old). Those with negative test results were then considered HIV-exposed uninfected. Children admitted with a fever, or a history of fever, and no documented history of HIV and no documented maternal HIV exposure by self-report, although with high clinical suspicion of HIV, were enrolled and offered HIV testing based on the aforementioned. Those with negative test results were considered HIV negative, and there were 18 subjects who were excluded from analysis because of a final designation of being HIV negative or unknown. For all eligible patients, study staff collected a single blood specimen for bacterial culture before the initiation of antibiotics. Additional diagnostic testing was performed if there was clinical suspicion based on the history or signs/symptoms. Readily available tests included complete blood count, blood chemistries, dipstick urinalysis, chest X-ray, lumbar puncture for chemistries and bacterial culture, stool culture, stool ova and parasites, HIV rapid antibody testing or DNA-PCR, and malaria antigen rapid testing. For patients with a positive malaria rapid test, thick and thin blood films were prepared for confirmation and to quantify Plasmodium falciparum parasitemia. There was no available laboratory diagnostics for TB in this age-group. Severe anemia was defined as a hemoglobin (Hb) concentration of < 7 g/dL, moderate anemia as Hb of 7–9.9 g/dL, mild anemia as Hb of 10–10.9 g/dL, and no anemia if Hb was > 11 g/dL. All patients were followed up until their discharge from the hospital. Possible final disposition included discharged patients, patients who died during hospitalization, or patients who abandoned treatment, meaning they left the hospital against the wishes of the treating clinician. Data for the parent study, including this anemia analysis, were collected by study clinicians using a paper-based study instrument and then uploaded into a password protected, tablet-based, online database maintained by the Research Electronic Data Capture consortium (www.project-redcap.org). This allowed for the recording of demographic information, medical and medication history, and information on clinical course while hospitalized. Data quality control was conducted by study investigators who reviewed all completed paper-based study instruments and confirmed the accuracy of data entered into the electronic database. Descriptive statistics were used to summarize the participants’ sociodemographic characteristics using frequencies and proportions (for dichotomous or categorical variables) or medians with interquartile ranges for continuous variables. In univariate analysis, we compared factors by anemia status defined as > 11 g/dL, 10–10.9 g/dL, 7–9.9 g/dL, and < 7 g/dL using chi-squared test. In regression analysis, for our response or outcome variable that proxies anemia, we used Hb level (g/dL) without categorization. Using the continuous response of the Hb level maximizes our regression power by using all information levels. We assessed the association of prespecified diagnosed conditions of interest that included bacteremia, malaria, gastroenteritis (GE)/diarrhea, acute respiratory illness (ARI), or malnutrition, with the Hb level as a biomarker response variable using separate regression analyses. Because anemia can be secondary to one’s ART treatment, we also examined the association of ART treatment and Hb level among HIV-infected patients. We examined the interaction effect of concurrent diagnoses using separate regression and a cross-product term (bacteremia and malaria, bacteremia and malnutrition, and malaria and malnutrition) on Hb levels as an exploration analysis. Among our subset of children who were HIV infected, we examined the association of diagnosis with the outcome of in-hospitalization death (yes versus no) using multivariable logistic regression. Each multivariable regression analysis included the child’s age, gender, and health facility as covariates for adjustment. We had case-wise deletions on ∼4% of subjects because of missing data for infant age (n = 18). We performed multiple imputations (MIs) to account for missing age data in multivariable regression as sensitivity analyses.26 We present beta coefficient-associating factors with anemia from complete case analysis. Statistical significance was determined using a 2-sided, 5% significance level. Statistical analysis was performed using R version 3.4.0 software (R core team, 2015, R Foundations for Statistical Computing, Viena, Austria, http://www.r-project.org). The Mozambican National Bioethics Committee for Health (Comité Nacional de Bioética para Saúde) (404/CNBS/14) and the Institutional Review Board of Vanderbilt University Medical Center (IRB#141167) approved this analysis. Informed consent was obtained from the parent or legal guardian of all children enrolled in this study.