Association of Respiratory Syncytial Virus Infection and Underlying Risk Factors for Death among Young Infants Who Died at University Teaching Hospital, Lusaka Zambia

Background: Respiratory syncytial virus (RSV) is a leading cause of acute lower respiratory tract infections and child mortality. While RSV disease burden is highest in low- and middle-income countries, most knowledge about risk factors for fatal RSV disease comes from high-income settings. Methods: Among infants aged 4 days to <6 months who died at University Teaching Hospital in Lusaka, Zambia, we tested nasopharyngeal swabs obtained postmortem for RSV using reverse transcriptase-quantitative polymerase chain reaction. Through a systematic review of death certificates and hospital records, we identified 10 broad categories of underlying medical conditions associated with infant deaths. We used backward-selection models to calculate adjusted and unadjusted risk ratios (RRs) for the association between each underlying condition and RSV status. Results: From 720 infant deaths, 6% (44) were RSV-positive, 70% were <4 weeks old, and 54% were male. At least 1 underlying condition was found in 85% of infants, while 63% had ≥2. Prematurity/low birth weight (53% [384]) and complications of labor and delivery (32% [230]) were the most common conditions. Congenital cardiac conditions were significantly associated with an increased risk of RSV infection (4%, 32; adjusted RR: 3.57; 95% CI: 1.71-7.44). No other underlying conditions were significantly associated with RSV. Conclusions: Other than congenital cardiac conditions, we found a lack of association between RSV and underlying risk factors. This differs from high-income settings, where RSV mortality is concentrated among high-risk infants. In this population, birth-related outcomes are the highest mortality risk factors. Improved neonatal care remains crucial in the fight against neonatal mortality. ZPRIME study researchers identified infants aged 4 days to younger than 6 months who died in Lusaka, Zambia, from August 2017 through August 2020. The present analysis utilizes a subset of the full cohort and includes data collected through December 2019. In Zambia, a burial cannot occur without a burial permit. Therefore, all deaths must first be cleared for burial at the medical examiner’s office at the University Teaching Hospital (UTH) or one of a select number of smaller clinics. For this analysis, we focus on deaths that occurred at the University Teaching Hospital in Lusaka, for which we had access to medical history data from the death certificates and clinical charts (Figure 1). In a separate analysis within this issue of Clinical Infectious Diseases, we present explanatory data pertaining to community RSV deaths (see Murphy et al in this Supplement issue). Flowchart showing the entry pathway for final enrollment into ZPRIME including the sources of noninclusion. Between the August 2017 and August 2020 ZPRIME we enrolled 2286 infants aged 4 days to <6 months. For infants enrolled from the UTH morgue, we collected “long form” data, which included demographic and clinical data that we extracted from the medical charts and/or the medical certificate of the cause of death. For this analysis, we focus on facility deaths that were enrolled between August 2017 and December 2019 for which we had collected “long form” data (highlighted in yellow). The present analysis includes ~98% of all these deaths. Abbreviations: UTH, University Teaching Hospital; ZPRIME, Zambia Pertussis Infant Mortality Estimation Study. For all deaths, researchers approached family members accompanying the body to the UTH morgue and, after obtaining informed consent, enrolled them in the study, and noted their age, sex, and date of death. We obtained postmortem nasopharyngeal (NP) samples from each infant using flocked nylon swabs (Copan Diagnostics, Murietta, CA). The swabs were placed in universal transport media on ice, transported to the microbiology laboratory on site at UTH, and stored at −80°C until RSV testing. Nucleic acid extraction was performed using the NucliSens easyMAG system (bioMérieux, Marcy I’Etoile, France), a system for automated isolation of nucleic acids from clinical samples based on silica extraction technology. Screening for RSV used a singleplex reaction specific to the dominant M protein on the virus using reverse transcriptase–quantitative polymerase chain reaction (RT-qPCR), a following the RSV protocol from the respiratory viruses branch at US CDC [7]. In order to demonstrate that the NP swab made effective contact with the respiratory mucosa, each sample run included primers/probes specific to the human constitutive enzyme RNAseP, which is expressed in all human cells (including the nasal epithelium). Its presence, therefore, validates the adequacy of the sample collection process. A positive RSV signal was defined as having a cycle threshold (CT) value of less than 40. All runs included positive and negative controls. Using the death certificates and hospital records of deceased infants, we identified sections of these forms where conditions related to the infants’ health before death were recorded. Fields analyzed from hospital records included 2 free-text fields: “provisional diagnosis” and “clinic diagnosis or reason for referral,” and checkboxes indicating maternal human immunodeficiency virus (HIV) status, prematurity, complications of labor and delivery, low birth weight, and malnutrition. Death certificates in Zambia categorize the cause of death hierarchically with 3 causes of death. All 3 of these fields were included in our analysis. These fields are as follows: “disease or condition directly leading to death,” “antecedent cause,” and “morbid conditions giving rise to the above cause.” We ran frequency tabulations of the data entered into these fields. The frequencies were then reviewed by the principal investigator who is a child survival expert and infectious disease specialist. He collapsed all of the underlying conditions into 1 of the 10 broad categories, as listed in Table 1. These categories were both hypothesis-generating, based on patterns seen in the data themselves, and based on historical precedent, utilizing prior knowledge of risk factors for infant mortality and respiratory disease. To confirm these categorizations, they were then reviewed by 2 additional infectious disease physicians, including a Zambian physician based in Lusaka. Often, the full-text field would list out the condition directly, but conditions were also described idiosyncratically and/or via a variety of different verbatim terms. For example, “low birth weight” was often listed using the acronym LBW, or VLBW (very low birth weight), or ELBW (extremely low birth weight), etc. Similarly, “congenital cardiac conditions” could appear as cyanotic heart disease, congenital heart disease, complex heart disease, or cardiomyopathy. It could also appear using various acronyms, or specifying certain syndromes directly (eg, Tetralogy of Fallot, often abbreviated as TOF). Infants were categorized as having prematurity in cases where the full words “prematurity” or “preterm delivery” were present. In other cases, conditions were described generically without a detailed anatomic explanation. The full list of verbatim terms and how these were collapsed into the 10 final conditions is included in Supplementary Table 1. Demographic Risk Factors, Medical Risk Factors, and Underlying Conditions by Presence of Respiratory Syncytial Virus Data are presented as n (%) unless otherwise indicated. Abbreviations: HIV, human immunodeficiency virus; RSV, respiratory syncytial virus. We then assigned underlying conditions to each infant based on the presence of keywords related to those 10 categories. Each field was analyzed independently of other fields for the same subject to avoid bias based on a full review of a particular infant’s health records. Infants could therefore be assigned 0, 1, or more underlying conditions. Demographic characteristics, such as place of birth (hospital or health facility vs other), parental employment and education, and number of people in the household, were collected via direct interview with the caretaker who accompanied the body to the morgue. Descriptive statistics were calculated to determine the prevalence of each of the above conditions, demographic characteristics, and medical risk factors, stratified by RSV status. Unadjusted log-binomial regression models were used to compute risk ratios (RRs) for each underlying condition and its association with RSV. Given the study design (prospective cohort study), we feel that we were able to capture all infants at risk during the period, and therefore RRs are an appropriate measure of the risk of RSV in the population. In order to determine which underlying factors and covariates were associated with RSV, we used a backward-selection approach. We ran 10 separate models: one for each of the underlying conditions, with the exception of conditions associated with prematurity, as there were zero RSV-positive infants with that condition, plus one for infants with none of the identified conditions. All demographic and medical risk factors shown in Table 1 were included in the original models, and the underlying condition was forced into the final model. We used a P-value threshold of .2 for inclusion in the final model. Last, we ran adjusted and unadjusted models for combinations of conditions which, based on prior medical knowledge, are often related. We created 1 merged condition such that infants with either 1 or both risk factors were classified as yes and infants who did not have either of the risk factors were classified as no. The following merged conditions were used: HIV exposure and/or malnutrition, prematurity and/or congenital cardiac conditions, and prematurity and/or conditions associated with prematurity.