**3. Results**

The database was accessed on 20 February 2016, and there were 5695 neonatal records on the database; 5386 records were for neonates born within the study period. There were 26 records with missing outcome data, four babies who had a birth weight below 400 g and 338 neonates who were admitted to the unit after 48 h. Thus, 5018 records were included in the review. The mean birth weight was 2148 g (SD 972) and the mean gestational age was 34.2 weeks (SD 4.8). The mean duration of stay was 13.75 days (SD 18.0).

**Figure 1.** Causes of neonatal deaths in Johannesburg, South Africa, between 01 January 2013 and 31 December 2015.

There were 724 deaths, giving an overall mortality rate of 14.4%, alternatively expressed as a percentage surviving to discharge of 85.6. Seventy-three percent (530/724) of neonates died in the early neonatal period, within seven days of birth. There were 147 (20.3%) deaths in the delivery room and seventy neonates (9.6%) died within the first 12 hours of admission to the neonatal ward. The various causes of neonatal death according to the PPIP classification are shown in **Figure 1**.

#### **3.1. Birth weight**

The mortality rate was strongly associated with birth weight. There were 3134 LBW neonates, with a mortality rate of 18.6% (586/3134). The majority of deaths in LBW neonates occurred in VLBW neonates (30.1% (479/1590)). Significantly more VLBW neonates died than babies >1500 g (30.1% vs. 7.1%; *p* < 0.001). The number of neonates and those who died in each birth weight category is shown in **Table 1**.


**Table 1.** Distribution of deaths by birth weight category for neonates at CMJAH between 2013 and 2015.

**Figure 2.** Percentage surviving by birth weight for neonates at CMJAH between 2013 and 2015.

The highly significant association between decreasing birth weight and increasing mortality is shown in **Figure 2** which depicts how the proportion surviving increases as birth weight increases. The percentage survival for a birth weight of 900 g is 52.8.

#### *3.1.1. Demographic and clinical characteristics in VLBW neonates compared to bigger babies*

There were 724 deaths, giving an overall mortality rate of 14.4%, alternatively expressed as a percentage surviving to discharge of 85.6. Seventy-three percent (530/724) of neonates died in the early neonatal period, within seven days of birth. There were 147 (20.3%) deaths in the delivery room and seventy neonates (9.6%) died within the first 12 hours of admission to the neonatal ward. The various causes of neonatal death according to the PPIP classification are

92 Epidemiology of Communicable and Non-Communicable Diseases - Attributes of Lifestyle and Nature on Humankind

The mortality rate was strongly associated with birth weight. There were 3134 LBW neonates, with a mortality rate of 18.6% (586/3134). The majority of deaths in LBW neonates occurred in VLBW neonates (30.1% (479/1590)). Significantly more VLBW neonates died than babies >1500 g (30.1% vs. 7.1%; *p* < 0.001). The number of neonates and those who died in each birth weight

**Birth weight (g) Number Died % Mortality**

**Table 1.** Distribution of deaths by birth weight category for neonates at CMJAH between 2013 and 2015.

**Figure 2.** Percentage surviving by birth weight for neonates at CMJAH between 2013 and 2015.

<1000 (ELBW) 524 315 60.1 1000–1499 1066 164 15.4 1500–2499 1544 107 6.4 2500–3999 (TAGA) 1730 130 7.5 >4000 (TLGA) 154 8 5.2

shown in **Figure 1**.

**3.1. Birth weight**

category is shown in **Table 1**.

Further results are reported for VLBW neonates compared to bigger babies. Demographic, maternal, and clinical characteristics are shown in **Table 2**. Certain conditions only occur in bigger babies and were thus not reported for VLBW neonates, namely meconium aspiration syndrome (MAS), persistent pulmonary hypertension of the neonate (PPHN), hypoxic ischemic encephalopathy (HIE), and cerebral cooling.



**Table 2.** Demographic, maternal and clinical characteristics by birth weight for neonates at CMJAH between 2013 and 2015.

#### **3.2. Risk factors for neonatal death**

Neonates who survived were born at a significantly more mature gestational age than those who died (34.8 weeks [SD 4.4] vs. 30.5 weeks [SD 5.5]; *p* < 0.001). Similarly, the birth weight of neonates who survived was significantly greater than those who died (2260 g [SD 932] vs. 1495 g [SD 940]; *p* < 0.001). Survivors stayed in hospital for a longer period of time than those neonates who died (14.8 days [SD 18.5] vs. 7.1 days [SD 13.3]; *p* < 0.001). Body temperature on admission was significantly higher in neonates who survived compared to those who died (36.3o C [SD8.0] vs. 35.6o C [SD 1.6]).

*Factor Cases % <1500 g >1500 g P-value*

94 Epidemiology of Communicable and Non-Communicable Diseases - Attributes of Lifestyle and Nature on Humankind

**PVL** 11 0.2 8 0.7 3 0.1 <0.001 **Died within 12 h of admission** 70 1.4 39 2.6 31 0.9 <0.001 **Pneumothorax** 36 0.7 10 0.7 26 0.8 0.444 **Pulmonary hemorrhage** 32 0.7 27 1.8 5 0.1 <0.001

**HMD** 2004 41.1 1347 89.6 657 19.5 <0.001 **NCPAP** 1565 32.5 1015 70.1 550 32.5 <0.001 **IPPV** 692 14.4 299 21 393 11.7 <0.001 **NCPAP without IPPV** 1228 24.5 795 78.9 433 79.4 0.816 **Surfactant therapy** 1580 33.1 1038 69.8 542 16.5 <0.001 **Steroids for CLD** 216 5.7 199 13.8 17 0.7 <0.001 **PDA** 245 5 152 10.2 93 2.8 <0.001 **NEC** 156 3.2 107 7.2 49 1.5 <0.001 **Other surgery** 136 2.9 29 2 107 3.3 0.014 **Packed cell transfusion** 674 13.4 527 35.9 147 4.4 <0.001 **Exchange transfusion** 24 1.5 9 1 15 2.2 0.039 **Hypoglycemia** 525 10.8 185 12.3 340 10 0.02 **Hyperglycemia** 375 7.7 287 19.1 88 2.6 <0.001 **Hypernatraemia** 169 3.5 148 9.8 21 0.6 <0.001 **Metabolic acidosis** 185 3.8 92 6.1 93 2.8 <0.001 **Late onset sepsis** 608 12.6 421 28.3 187 5.6 <0.001

**Table 2.** Demographic, maternal and clinical characteristics by birth weight for neonates at CMJAH between 2013 and

Neonates who survived were born at a significantly more mature gestational age than those who died (34.8 weeks [SD 4.4] vs. 30.5 weeks [SD 5.5]; *p* < 0.001). Similarly, the birth weight of neonates who survived was significantly greater than those who died (2260 g [SD 932] vs. 1495

2015.

**3.2. Risk factors for neonatal death**

**IVH 3/4** 61 1.2 N/A N/A

**HIE 2/3** 174 3.6 N/A N/A **Cerebral cooling** 103 38.1 N/A N/A **Meconium aspiration syndrome** 264 7.8 N/A N/A **PPHN** 54 1.6 N/A N/A

*n %n %*

Conditions significantly associated with death in all the neonates, including those who died in the delivery room, are shown in **Table 3**. Only data for babies who died are reported. The percentages refer to the number of babies who died with and without the various conditions. For example, 38.5% (102) of babies who had a major birth defect died and 12.8% (596) of babies without a major birth defect died. Percentages are reported per the total number of complete cases for each condition—missing data were excluded. All other conditions were not signifi‐ cantly associated with death in the whole group of neonates.



**Table 3.** Factors associated with death in all neonates who died (*n* = 724), including delivery room deaths.

The results of binary logistic regression, considering whether neonate survived to discharge as the outcome variable, are shown in **Table 4**. The chances of survival decreased with metabolic acidosis, hyperglycemia, mechanical ventilation, major birth defect and the need for resuscitation at birth, while increasing birth weight and gestational age and delivery by Caesarean section were associated with an increased chance of survival.


**Table 4.** Results of binary logistic regression model for factors associated with survival in all neonates at CMJAH between 2013 and 2015.

#### *3.2.1. Binary logistic regression: VLBW neonates*

**Factor Condition present Condition absent** *P*

96 Epidemiology of Communicable and Non-Communicable Diseases - Attributes of Lifestyle and Nature on Humankind

**PVL** 4 36.4 424 9.5 0.003 **IVH grade 3/4** 29 47.5 32 52.5 <0.001 **HIE grade 2/3** 48 27.9 124 72.1 <0.001 **NEC** 65 41.7 508 10.9 <0.001 **Surgery (not NEC)** 38 27.7 524 11.4 <0.001 **Blood transfusion** 148 22.0 422 10.1 <0.001 **Hypoglycemia** 78 14.9 504 11.6 0.029 **Hyperglycemia** 165 44.0 417 9.3 <0.001 **Hypernatraemia** 64 37.9 518 11.0 <0.001 **Metabolic acidosis** 101 54.6 481 10.3 <0.001 **Late onset sepsis** 135 22.0 434 10.3 <0.001

**Table 3.** Factors associated with death in all neonates who died (*n* = 724), including delivery room deaths.

Caesarean section were associated with an increased chance of survival.

**Constant** 0.21

between 2013 and 2015.

**Condition Odds ratio 95% CI for OR**

**Metabolic acidosis** 0.135 0.09 0.204 **Hyperglycemia** 0.307 0.23 0.409 **Mechanical ventilation** 0.357 0.278 0.46 **Birth weight** 1.001 1 1.001 **Major birth defect** 0.118 0.079 0.175 **Gestational age** 1.109 1.054 1.167 **Caesarean section** 1.803 1.444 2.251 **Resuscitated at birth** 0.395 0.315 0.495

**Table 4.** Results of binary logistic regression model for factors associated with survival in all neonates at CMJAH

The results of binary logistic regression, considering whether neonate survived to discharge as the outcome variable, are shown in **Table 4**. The chances of survival decreased with metabolic acidosis, hyperglycemia, mechanical ventilation, major birth defect and the need for resuscitation at birth, while increasing birth weight and gestational age and delivery by

**Lower Upper**

**# Died % # Died %**

The results of binary logistic regression considering survival to discharge as the outcome variable were performed for VLBW neonates (see **Table 5**). The percentage survival increased with increasing birth weight, delivery by Caesarean section and the use of NCPAP without the need for mechanical ventilation. Maternal HIV, hyperglycemia, resuscitation at birth, pulmonary hemorrhage, NEC, and metabolic acidosis were associated with a reduced chance of survival.


**Table 5.** Binary logistic regression for factors associated with survival to discharge in VLBW neonates at CMJAH between 2013 and 2015.

#### *3.2.2. Binary logistic regression: bigger neonates*

The results of binary logistic regression considering survival to discharge as the outcome are shown in **Table 6**. Birth weight was not significantly different between survivors and nonsurvivors in this weight category. Decreasing gestational age, the need for resuscitation at birth, mechanical ventilation, metabolic acidosis, and hyperglycemia were all associated with a reduced chance of survival.


**Table 6.** Binary logistic regression for factors associated with survival to discharge in bigger neonates at CMJAH between 2013 and 2015.

#### *3.2.3. Delivery room deaths*

Neonates who died in the delivery room were less likely to have received antenatal steroids and be delivered to mothers with hypertension or HIV, compared to neonates who died in the neonatal wards. Delivery room deaths were associated with vaginal delivery and were more likely in neonates who had been resuscitated at birth (see **Table 7**). Neonates who died in the delivery room had a lower body temperature on admission than those who died in the neonatal wards (34.6o C [SD 2.8] compared to 35.8o C [SD 1.2]; *p* < 0.001). All other variables including birth weight and gestational age were not different between neonates who died in the delivery room compared to the neonatal wards.


**Table 7.** Maternal and delivery room factors compared between babies who died in the delivery room and those who died in the neonatal wards.

### **4. Discussion**

The ongoing audit of neonatal mortality and neonatal care to determine risk factors for poor outcome is essential so that correct interventions can be implemented. The MDG 2015 report states that better readily available data is urgently needed to guide health policies [1]. There is a slogan in the report that says "together we can measure what we treasure". The so-called "Plan Do Study Act [PDS] cycle is a tool for quality improvement projects [10]. Ongoing clinical audit is fundamental to quality improvement projects, both for planning the intervention and then measuring the benefit of the intervention [11, 12]. It is also essential to have appropriate local data available; different NICUs and neonatal populations have different problems and need tailored solutions. For example, maternal HIV is an important issue in the current study, but would not apply in a European setting.

The best example of clinical audit and quality improvement in neonatal care is the Vermont Oxford Network [VON] (www.vtoxford.org). The VON is a multinational multicenter collaboration of neonatal units established in 1989 with the aim of improving quality and effectiveness of neonatal care by research, education and quality improvement projects [13]. There are currently more than 1000 neonatal units from around the world that participate in the VON. Collaborative multi-disciplinary quality improvement projects [NIC/Q] are conducted annually [14].

The present study was an audit of neonatal survival and risk factors for poor outcome in Johannesburg, South Africa. The overall neonatal survival rate in the present study was 85.6%. Birth weight greatly influenced survival with 69.1% of VLBW surviving compared to 92.1% of neonates above 1500 g birth weight. The VLBW survival in our unit was significantly less than that reported in the VON [www.vtoxford.org] for the same period (69.1% vs. 85.6%). Neonatal mortality rates among different neonatal units are highly variable, but the rates reported in the present study are within the reported range for developing nations [3]. The current neonatal survival rates are better than those reported from NICUs in The Gambia [4] and Ethiopia [15], but worse than those reported from a NICU in Thailand [16]. It must be noted that different mortality rates will be reported depending on which neonates are included in the audit—the present study included neonates from 400 g birth weight and those who died within the delivery room—omission of these would improve the results.

*3.2.3. Delivery room deaths*

C [SD 2.8] compared to 35.8o

room compared to the neonatal wards.

but would not apply in a European setting.

wards (34.6o

died in the neonatal wards.

ducted annually [14].

**4. Discussion**

Neonates who died in the delivery room were less likely to have received antenatal steroids and be delivered to mothers with hypertension or HIV, compared to neonates who died in the neonatal wards. Delivery room deaths were associated with vaginal delivery and were more likely in neonates who had been resuscitated at birth (see **Table 7**). Neonates who died in the delivery room had a lower body temperature on admission than those who died in the neonatal

98 Epidemiology of Communicable and Non-Communicable Diseases - Attributes of Lifestyle and Nature on Humankind

birth weight and gestational age were not different between neonates who died in the delivery

**Condition present Delivery room death Percentage Neonatal ward death Percentage** *P***-value Antenatal steroids** 15 9.4 144 32.3 0.001 **Maternal hypertension** 12 12.2 86 17.3 0.032 **Maternal HIV** 25 21.6 184 35.2 0.004 **Caesarean section** 51 35.4 252 44.8 0.042 **Resuscitated at birth** 95 64.6 297 53.2 0.013

**Table 7.** Maternal and delivery room factors compared between babies who died in the delivery room and those who

The ongoing audit of neonatal mortality and neonatal care to determine risk factors for poor outcome is essential so that correct interventions can be implemented. The MDG 2015 report states that better readily available data is urgently needed to guide health policies [1]. There is a slogan in the report that says "together we can measure what we treasure". The so-called "Plan Do Study Act [PDS] cycle is a tool for quality improvement projects [10]. Ongoing clinical audit is fundamental to quality improvement projects, both for planning the intervention and then measuring the benefit of the intervention [11, 12]. It is also essential to have appropriate local data available; different NICUs and neonatal populations have different problems and need tailored solutions. For example, maternal HIV is an important issue in the current study,

The best example of clinical audit and quality improvement in neonatal care is the Vermont Oxford Network [VON] (www.vtoxford.org). The VON is a multinational multicenter collaboration of neonatal units established in 1989 with the aim of improving quality and effectiveness of neonatal care by research, education and quality improvement projects [13]. There are currently more than 1000 neonatal units from around the world that participate in the VON. Collaborative multi-disciplinary quality improvement projects [NIC/Q] are con-

The present study was an audit of neonatal survival and risk factors for poor outcome in Johannesburg, South Africa. The overall neonatal survival rate in the present study was 85.6%.

C [SD 1.2]; *p* < 0.001). All other variables including

The most important causes of neonatal death in the present study were complications of prematurity, perinatal asphyxia, infection, and birth defects. These findings are similar to other studies evaluating risk factors for neonatal mortality [1, 17], although the contribution of prematurity to neonatal death is considerably higher than that reported in United Nations Millennium Development Goal Report 2015 [1] (42.3% vs. 35%). Birth weight is closely linked to gestational age in LBW neonates; the higher mortality with decreasing birth weight in the present study corresponds to increasingly premature neonates. It is interesting to note that in bigger babies, gestational age, rather than birth weight, was associated with survival. Almost 15% of deaths in the present study were due to congenital abnormalities; this reflects the fact that the unit was a referral centre for pediatric surgery; so many neonates with major congenital abnormalities were referred in for surgery.

The present results are also similar to a report from a private healthcare group in South Africa, who found that birth weight, Apgar score, and mode of delivery were all associated with neonatal mortality [18]. This is interesting, as the majority of patients in the private health care group were of White and Indian ethnicity, whereas those in the current report were almost exclusively Black African.

Most of the neonatal deaths in the current study occurred in VLBW neonates; therefore resources need to be focused on this group of neonates in order to reduce childhood mortality. Decreasing birth weight, maternal HIV, the need for resuscitation at birth, pulmonary hemorrhage, NEC, hyperglycemia, and metabolic acidosis were all associated with a decreased chance of survival in VLBW neonates, while delivery by Caesarean section and the use of NCPAP without the need for mechanical ventilation significantly increased survival. These findings are similar to reports from the same unit [5, 6]. Interventions need to be devised to address these specific risk factors, such as ensuring prevention of mother to child transmission of HIV, providing proper prompt neonatal resuscitation, maintaining normoglycemia, and promoting breastfeeding. All preterm neonates, irrespective of birth weight, should be provided with NCPAP. The use of surfactant and mechanical ventilation may not be available in all NICUs in LMICS due to resource limitations. If necessary, surfactant and mechanical ventilation can be rationed using prognostic criteria. The association of better survival with Caesarean section is a more difficult one – it is possible that neonates delivered by Caesarean section are the "better babies." These mothers may have attended antenatal care, been admitted earlier in labor, and received antenatal steroids. It is therefore possible that Caesarean section is a confounding variable. It is certainly not feasible to suggest that all preterm neonates in LMICS be delivered by Caesarean section. Other factors such as antenatal care, antenatal steroid use, and neonatal infection were not significantly predictive of survival in the present study. This does not mean, however, that regular antenatal care attendance, the use of antenatal steroids, and infection control should be omitted from interventions to improve VLBW survival.

The factors associated with poor survival in bigger neonates included decreasing gestational age, the need for resuscitation at birth, mechanical ventilation, metabolic acidosis, and hyperglycemia. This emphasizes the need for all birth attendants to be skilled in neonatal resuscitation. It is possible that mechanical ventilation will not be available in many NICUs in LIMICs, but bigger preterm infants can be successfully managed with surfactant therapy and NCPAP [9].

A recent report from Burundi showed that the neonatal survival rates were significantly improved in a low resourced district hospital, without specialist care [19]. This was achieved by integrating neonatal and obstetric services, with an emphasis on prompt referral and transfer of mothers in preterm labor, the ongoing on-site training of staff with clear protocols for case management, provision of essential equipment, and providing complementary kangaroo mother care and NICU facilities.

In conclusion, ongoing clinical audit is integral to the process of quality improvement, to develop appropriate health care policies and to monitor the impact of these policies. Focus on neonatal care and especially that of VLBW neonates is essential if we are to achieve the SDG goal of reducing neonatal mortality to 12 per 1000 births.
