Incidence of Mortality among Under-Five-Year-Old Children Born to Women Living with HIV and Those Born to Women Not Living with HIV in Botswana: A 5-Year Retrospective Study

Background Child mortality is a core indicator for child health and wellness. Botswana reported an under-five-year-old children (UFC) mortality rate of 48 deaths per 1000 live births in 2017 against 152 deaths per 1000 live births in 1971. This was a commendable accomplishment. However, given the current country situation whereby 23% of children are born to women living with HIV, the incidence of mortality among UFC born to women living with and not living with HIV and their survival are better health metrics to inform decision making. Nevertheless, such data are still very scarce in Botswana. The study's objective was to estimate the incidence of UFC mortality among children born to women living with and not living with HIV and to compare UFC survival between the two groups. Methods A retrospective cohort study of mortality among UFC was conducted in Botswana, including all UFC born between January 2014 and June 2018. Data were extracted from the National Under-Five Mortality Audit Committee (NUFMAC) database using a standardized data collection tool. The incidence rate of UFC death was estimated as a function of the duration from birth to death. Survival functions of UFC born to women living with and not living with HIV were plotted and compared using Kaplan–Meier survival analysis. Results The overall incidence of UFC death was 4.63/1000 child months (CM) (95% CI 4.36–4.90). The incidence of UFC death among children born to women living with HIV was 6.96/1000 CM (95% CI 6.47–7.45) and that of UFC born to women not living with HIV was 4.34/1000 CM (95% CI 4.03–4.65). The overall average and standard error (SE) time to event/death for UFC born to women living with and not living with HIV was 54.80 (0.18) months. The mean (SE) time to death for UFC born to women living with HIV was 52.79 (0.41) months and that of UFC born to women not living with HIV was 55.44 (0.19) months (log-rank X2 = 37.59, p < 0.001). Prematurity or low birth weight was the leading cause of UFC death in both groups; but, it was higher in UFC born to women not living with HIV subgroup than their counterparts. Four cases only or 0.5% of the 806 death cases reported by reporting physicians were attributable to HIV-related complications. Conclusion Despite the commendable efforts made in reducing UFC death, the incidence of UFC death among UFC born to women living with HIV in Botswana is still higher, and their survival is shorter compared to UFC born to women not living with HIV. Child survival interventions should prioritize UFC born to women living with HIV to improve their survival.


Background
Child mortality is a core indicator of child health and wellbeing. According to the World Health Organization (WHO), the under-five-year-old children mortality rate (U5MR) has decreased by 58%, from an estimated rate of 93 deaths per 1000 live births in 1990 to 39 deaths per 1000 live births in 2017 globally [1]. However, in sub-Saharan Africa (SSA), child mortality still remains a significant public health problem with the U5MR at 76 deaths per 1,000 live births in 2017 [2]. is translates to 1 child in 13 dying before his/her fifth birthday, which is 14 times higher than the average ratio of 1 in 185 in high-income countries and 20 times higher than 1 in 263 in the region of Australia and New Zealand [2].
Human immunodeficiency virus (HIV) infection affects child mortality directly through vertical transmission and indirectly through HIV associated maternal death and/or ill health/psychosocial status [3]. In 2013, 2% of global mortality among under-five-year-old children (UFC) was attributed to HIV infection [4]. However, the successful roll-out of the program for the vertical transmission prevention of HIV has led to fewer infants and children becoming infected with the virus each year [4][5][6][7].
Botswana, a country in SSA, is characterized by a high U5MR of 48 per 1000 live births in 2017 [8]. Over the years, Botswana has made significant progress. According to the 2017 Botswana Demographic Survey (BDS), UFC mortality declined from 152 deaths per 1000 live births in 1971 to 48 deaths per 1000 live births in 2017 [8]. While infant mortality declined from 97 deaths per 1000 live births to 38 deaths per 1000 live births, respectively [8].
e decline in child mortality in the mid-2000s was partly due to the introduction of antiretroviral therapy (ART) and high coverage of vertical transmission prevention program [8]. e 2015 Botswana Millennium Goals Report indicates that the country attained the Millennium Development Goal 4 (MDG-4) [9]. However, the U5MR in most health facilities is still very high, and about 40-50% of this mortality is due to neonatal deaths [10]. In its efforts to alleviate the situation, the government of Botswana, in collaboration with its partners, has embarked on several programs to promote the child's health and survival. ese efforts include implementing the Integrated Management of Childhood Illnesses (IMCI) strategy in 1998, the introduction of the Accelerated Child Survival and Development Strategy in 2009, a robust routine immunization program, and periodic prevention campaigns (vitamin A supplementation and deworming campaign) as well as vertical transmission prevention program. ese initiatives have resulted in a tremendous decline in infant and child mortality and improved maternal health as observed by Statistics Botswana between 2008 and 2013 [11].
As a signatory to the Sustainable Development Goals (SGD), Botswana has a target to reduce U5MR by 25 per 1000 live births. Looking at health metrics from Statistics Botswana [8], there is no doubt that the country will meet the SGD goal. e time frame to attain the SGD goal is, however, not clear as the death and survival details of UFC born to women living with HIV are scanty. is situation may obscure differences or inequalities in the distribution of health outcomes among UFC subgroups and likely mislead policy and interventions. Metrics that provide information on the distribution of health outcomes among UFC subgroups are helpful in policy and decision making because 23% of UFC in Botswana are born to women living with HIV [12]. erefore, metrics such as the incidence of mortality among UFC born to women living with HIV, survival, and time-to-event (death) are critical when planning and implementing health programs; literature search yields very little on these metrics [8][9][10][11], yet this information is key to informing decision making. e current study aimed to (i) estimate the incidence of under-five mortality among UFC born to women living with and not living with HIV, (ii) compare the survival of UFC born to women living with HIV with the survival of those born to women not living with HIV, and (iii) assess the time to event/death in these subgroups.

Study Design and Area.
We undertook a retrospective cohort study of mortality among UFC in districts in Botswana between January 2014 and December 2018.
Botswana has 2 cities, namely, Gaborone, the capital city, and Francistown, and 5 towns, namely, Jwaneng, Orapa, Lobatse, Selibe-Phikwe, and Sowa [13]. e total population of Botswana is 2,024,904 [14] and that of children under five years (UFC) is 237,387 or 12% of the total population. e study population included all UFC born in Botswana between January 2014 and June 2018. We excluded from the study all UFC with unknown place of birth or born outside Botswana, multiple births, and those with undocumented cause of death.

Sample Size, Data Collection, and Assessment.
e sample size was exhaustive and included everyone as this was a nationwide study. Data was manually extracted from the National Under Five Mortality Audit Committee (NUF-MAC) database and exported into a Microsoft Excel spreadsheet for cleaning and screening.
Extracted data included the following variables: e medical record number of the UFC, sociodemographic and biomedical data: gender, date of birth (DOB), gestational age, birth weight, mode of delivery (spontaneous vaginal delivery (SVD) or caesarean section), maternal HIV status, whether HIV prophylaxis was given, feeding status (exclusive breastfeeding/exclusive formula feeding/mixed feeding), immunizations received by a child, mother's residential area, diagnosis or cause of death including UFC HIV status (as diagnosed by the reporting physician), and date of death (DOD). Aside from exclusion factors mentioned in previous paragraph, UFC missing any of these variables were excluded from the study. Children born in Botswana's health facilities are all scheduled for postnatal care for mothers and newborns based on WHO 2013 guidelines [9]. However, whether or not the numbers of follow-up visits between women not living or living with HIV were comparable was not recorded. Mortality was ascertained as by the reporting physician. UFC who died out of health facilities were excluded from the study.
Data analysis was performed using IBM SPSS version 21 (Chicago, IL). e frequency distribution of sociodemographic and biomedical characteristics known as potential risk factors of UFC death was assessed. ey included gender, maternal HIV status, and age at death as a continuous-time variable. Age groups were 0-0.91 months (neonate), 0.92-12.00 months (infant), and 12.01-59.99 months. Age at death was estimated by subtracting the DOB from the DOD of the UFC. Proportions of UFC born to women living with HIV and to women not living with HIV were computed. In addition, the proportions of UFC death by cause as diagnosed by the reporting physician were estimated.
Groups or subgroups were compared using chi-square analysis for categorical data and t-student test for continuous data with the level of statistical significance set at p < 0.05. e incidence rate of UFC death was estimated as a function of the duration from birth to death. To compute the incidence of deaths among UFC, the rate of new deaths was calculated as the number of new deaths divided by the total child months of follow-up (CM). e CM estimated the actual time-at-risk in the months that the UFC contributed to the study. UFC death being the event/status, survival was estimated as the time elapsed from DOB to the DOD of the UFC or the end of the study. Cases which were lost to follow-up were censored at the last time they were seen while those who stayed until the end of the study without dying from any conditions were censored at the end of the study. To investigate whether survival was the same between UFC born to women living with and not living with HIV, Kaplan-Meier survival analysis was performed. e average time-to-event/death was estimated for UFC born to women living with HIV and for UFC born to women not living with HIV and their 95% confidence intervals (CIs) were computed. e survival function was plotted for the UFC born to women living with HIV and for the UFC born to women not living with HIV and we were checked whether they were similar using the log-rank (Mantel-Cox) chi-square.

Results
Overall, 12,798 UFC were investigated. Data presented in Table 1 show sociodemographic and biomedical characteristics of UFC born to women living with and not living with HIV between January 2014 and December 2018. Of the 12,798 UFC investigated, 7,264 (56.8%) were male. e median age (IQR) of participants was 8.5 (0.6-22.1) months; that of UFC born to mothers living with HIV was 7.4 (0.7-11.1) months and for UFC born to mothers not living with HIV was 9.2 (0.5-23.2) months. e youngest UFC was 0.1 months while the oldest was 59.74 months. ere were 5453 UFC (42.6%) in the age group of 12.01-59.9 months. UFC born to women living with HIV numbered 3300 or 25.8% of the study population, including 10 (0.3%) who were infected through vertical transmission. Under-five-year-old children born to women not living with HIV totalled 9,498; none of them was HIV positive. e overall follow-up time was 174,165 child months (CM), the total number of events or deaths was 806 (including 4 HIV-related deaths), corresponding to an incidence of 4.63/1,000 CM (95% CI 4.36-4.90). e total CM follow-up among UFC born to women living with HIV was 40,382.73 and the total number of events or deaths was 281, corresponding to an incidence of 6.96 CM (95% CI 6.47-7.45).
e total CM follow-up among UFC born to women not living with HIV was 133,782.32 and the total number of events/deaths was 525, corresponding to an incidence of 4.34/1,000 CM (95% CI 4.03-4.65).
Leading causes of death as by the reporting physician among UFC born to women living with and not living with HIV are given in Table 2. Of the 806 death cases recorded among UFC, 359 or 44.5% were caused by prematurity/low birth weight. Other causes were birth asphyxia, diarrhoea, septicaemia, pneumonia, and malnutrition. ere were 4 (0.5%) cases of HIV-related UFC death.
Significant (p < 0.05) differences between UFC born to women living with and not living with HIV were noted in mortality, mainly attributable to prematurity/low birth weight, pneumonia, and malnutrition. e cause of death was unknown for 12.8% of UFC born women living with HIV and 20.0% of UFC born to women not living with HIV (p < 0.001). UFC deaths were comparable between the two groups for the rest of the leading causes of mortality.
Results presented in Table 3 show the distribution of deaths by age among UFC born to women living with and not living with HIV. Seventy percent (70%) of deaths occurred among UFC aged between 0.0 and 91.00 months, 24% among UFC age between 0.92 months and 12.00 months while 9.00% of death occurred among UFC aged more than 12 months. e average time-to-event for UFC born to women living with and not living with HIV is presented in Table 4 Kaplan-Meier survival curve in Figure 1 showed a significant (p < 0.001) difference between UFC death for UFC born to women living with HIV and UFC death for UFC � under-five children, LBW � low birth weight, and * p < 0.05; * * p < 0.001; different letters between rows reflect significant differences (p < 0.05).

Discussion
is study found out that only 0.3% of UFC born to women living with HIV were HIV positive by the reporting physician's diagnosis. e incidence of UFC death was significantly higher in this subgroup compared to UFC born to women not living with HIV. eir survival and average time to death were also shorter compared to their counterparts born to women not living with HIV. e finding suggests that mortality of under-five-year-old children born to women living with HIV was not associated with the UFC serological status but with factors related to their mother serological status.
Studies supporting these findings are in the literature [3,[15][16][17][18][19][20][21]. Of particular interest is the study conducted by Zash and coworkers in Botswana in 2016. ese investigators (Zash et al.) recruited mothers and their newborns directly from postpartum wards in 5 different Botswana communities and followed them up for 24 months. ey estimated the cumulative incidence of UFC death per 1000 child-years and compared outcomes between UFC born to women living with HIV and UFC born to women not living with HIV. eir results showed profiles very comparable to the findings presented herein. UFC born to women living with HIV had a higher incidence of mortality and shorter survival and time-to-event than their counterparts.
Similar results have also been reported in neighbouring Zimbabwe where noninfected UFC born to women living with HIV had a death incidence of 150 per 1000 personyears compared to their counterparts born to women not living with HIV who had a death incidence of only 47 per 1000 person-years [3]. Outside the subregion, comparable results have also been reported [3,15,22,23]. What comes out of all these studies is that there are factors associated with mothers that independently affect UFC death and that strategies currently seeking improvement of UFC death are not fully addressing. Intervention programs are needed to get UFC death comparable between the 2 subgroups. e use of secondary data in this study is a limiting factor, since we were unable to analyse variables other than those collected by the primary users. Also, the extensive exclusion of potential participants due to missing data and loss to follow-up bias to which retrospective studies are prone are other limitations of this study. Further studies   using primary data may help to establish better facts brought up by this study. However, using a nationwide dataset of 12,798 persons and 5 years' follow-up gives strength and inferentiality to the outcomes. While it is difficult to imagine how the investigator or participant could be influenced by awareness of an outcome that has not yet occurred in prospective cohort studies, in retrospective cohort studies participants generally know both their exposure status and the outcome.
us, these studies are prone to selection bias. Subjects with both the exposure and the condition under study are more likely to be attracted by the study than those without, especially if questions under study have some potential liability or monetary consequences tied to the outcome. is may of course cause differential retention of subjects in relation to the exposure and the outcome, which will lead to an overestimation or underestimation of the association or loss to follow-up bias. is might add up to limiting factors of this work.
However, in this particular study, the likelihood of this bias was very low as no contact was made with subjects. Variables and questions addressed herein are those routinely collected and reported by reporting physicians during postnatal follow-up care. No potential liability or monetary consequences are tied to the outcome of the follow-up. Also driven phenomenon rather than anything else. After delivery every mother is excited to know where her child is keeping up with the recommended WHO growth curve [9] and, wants to make sure that her child receives all recommended vaccines in time.
is is perhaps what best explains the number of their contacts with the health system rather than their serological status.
In this study, all the women living with HIV were on ART and enjoyed good health status. Only 0.3% of their offspring had HIV infection through vertical transmission; it is therefore difficult that ascertainment of death among children born to women living with HIV be explained by the number of their contacts with the health system. e study brings in new inputs that may foster strategies aiming to improve UFC health and wellness. Differences observed between UFC subgroups may call for subgroup tailor-made public health actions to improve UFC health and wellness in general.

Data Availability
Data underlying the findings in this study are not publicly available to maintain patient confidentiality. e data include potentially identified demographic and clinical care information. However, the data can be requested from the corresponding author, who must first get permission from the Ministry of Health and Wellness and the National Under-Five Mortality Audit Committee before sharing.

Conflicts of Interest
e authors declare that there are no conflicts of interest.