Objectives To investigate causes of and contributors to newborn deaths in eastern Uganda using a three delays audit approach.
Methods Data collected on 64 neonatal deaths from a demographic surveillance site were coded for causes of deaths using a hierarchical model and analysed using a modified three delays model to determine contributing delays. A survey was conducted in 16 health facilities to determine capacity for newborn care.
Results Of the newborn babies, 33% died in a hospital/health centre, 13% in a private clinic and 54% died away from a health facility. 47% of the deaths occurred on the day of birth and 78% in the first week. Major contributing delays to newborn death were caretaker delay in problem recognition or in deciding to seek care (50%, 32/64); delay to receive quality care at a health facility (30%; 19/64); and transport delay (20%; 13/64). The median time to seeking care outside the home was 3 days from onset of illness (IQR 1–6). The leading causes of death were sepsis or pneumonia (31%), birth asphyxia (30%) and preterm birth (25%). Health facilities did not have capacity for newborn care, and health workers had correct knowledge on only 31% of the survey questions related to newborn care.
Conclusions Household and health facility-related delays were the major contributors to newborn deaths, and efforts to improve newborn survival need to address both concurrently. Understanding why newborn babies die can be improved by using the three delays model, originally developed for understanding maternal death.
Utilisation du modèle des trois retards pour comprendre pourquoi les nouveaux-nés meurent dans l’est de l’Ouganda
Objectifs: Investiguer les causes et les contributeurs aux décès néonataux dans l’est de l’Ouganda en utilisant une approche de vérification des trois retards.
Méthodes: Les données recueillies sur 64 décès néonataux dans un site de surveillance démographique ont été codées pour les causes de décès en utilisant un modèle hiérarchique et analysées à l’aide du modèle modifié des trois retards afin de déterminer les retards contributeurs. Une enquête a été menée dans 16 établissements de santé pour déterminer leur capacité pour les soins néonataux.
Résultats: 33% des nouveau-nés sont morts dans un hôpital/centre de santé, 13% dans une clinique privée et 54% sont morts loin d’un établissement de santé. 47% des décès sont survenus le jour de la naissance et 78% dans la première semaine. Les retards les plus importants contribuant à la mort du nouveau-néétaient: le délai de reconnaissance du problème ou de prise de décision pour le recours aux soins par le soignant (50%, 32/64), le délai pour recevoir des soins de qualité dans un établissement de santé (30%; 19/64), et le délai dû au transport (20%; 13/64). Le délai médian de recours aux soins en dehors du foyer était de 3 jours à partir de l’apparition de la maladie (IQR: 1-6). Les principales causes de décès étaient: la septicémie ou la pneumonie (31%), l’asphyxie à la naissance (30%) et la prématurité (25%). Les établissements de santé n’avaient pas la capacité pour les soins du nouveau-né et les agents de santé avaient des connaissances correctes sur seulement 31% des questions de l’enquête sur les soins du nouveau-né.
Conclusions: Les retards des ménages et ceux liés aux établissements de santéétaient les principaux facteurs de mortalité néonatale. Les efforts pour améliorer la survie des nouveaux-nés doivent répondre aux deux aspects la fois. La compréhension des raisons des décès des nouveaux-nés peut être améliorée en utilisant le modèle des trois retards, initialement développé pour comprendre la mortalité maternelle.
Utilización del modelo de “tres retrasos” para entender porqué mueren los recién nacidos en Uganda del Este
Objetivos: Investigar las causas de y contribuciones a las muertes neonatales en Uganda del Este utilizando un enfoque basado en el modelo de “tres retrasos” (three delays model).
Métodos: Se recolectaron datos sobre 64 muertes neonatales de un lugar con vigilancia demográfica, se codificaron según causa de muerte utilizando un modelo jerárquico y se analizaron utilizando una modificación del modelo de tres retrasos para determinar aquellos retrasos que contribuían a la mortalidad. Se realizó una encuesta en 16 centros sanitarios para determinar su capacidad en el ámbito de los cuidados neonatales.
Resultados: Un 33% de los neonatos murieron en el hospital/centro sanitario, 13% en clínicas privadas y 54% lejos de cualquier centro sanitario. Un 47% de las muertes ocurrió en el día del nacimiento y un 78% durante la primera semana. Las principales contribuciones a la muerte neonatal venían de los retrasos del cuidador a la hora de reconocer el problema o buscar ayuda (50%, 32/64); del retraso a la hora de recibir atención de calidad en el centro sanitario (30%; 19/64); y del retraso en el transporte (20%; 13/64). El tiempo medio para buscar cuidados fuera de casa era de 3 días desde el momento en que comenzaba la enfermedad (IQR 1-6). Las principales causas de muerte eran sepsis o neumonía (31%), asfixia en el parto (30%) y parto antes de llegar a término (25%). Los centros sanitarios no tenían la capacidad necesaria para brindar cuidados neonatales, y los trabajadores sanitarios tenían los conocimientos adecuados en solo un 31% de las preguntas de la encuesta relacionadas con cuidados neonatales.
Conclusiones: Los retrasos relacionados con el hogar y los centros sanitarios fueron los principales contribuyentes a las muertes neonatales, y los esfuerzos que se realicen con el fin de mejorar la supervivencia neonatal deben tener en cuenta ambos retrasos de forma concurrente. El entendimiento de la muerte neonatal puede mejorar con el uso del modelo de tres retrasos, originalmente desarrollado para entender la muerte materna.
Understanding the causes and contextual factors surrounding the 4 million newborn deaths occurring every year (Lawn et al. 2005a) is important for health programming and policy. Most newborn deaths occur in low-income countries (Rudan et al. 2005); the majority remain uncounted (Lawn et al. 2005a,b). This is due mainly to the fact that most births and deaths occur at home, and that causes of these deaths are not medically certified. Most available information on the causes of neonatal death is from health facility data, which are not representative of the general population. Furthermore, the documentation of births and deaths is often poor in health facilities; consequently, understanding the epidemiology of newborn deaths in developing countries is, to a great extent, based only on statistical estimations that present many uncertainties (Lawn et al. 2006; Sachdev 2006).
Attempts to obtain population-based data on newborn causes of death in low-income countries have relied on verbal autopsy. Verbal autopsy is a process employing post-mortem in-depth interviews with the primary caregiver, usually the mother of the newborn (Thatte et al. 2009). This approach assumes that each cause of death has a set of observable symptoms that can be recognised and recalled by the primary caregiver and that the characteristics of one cause of death can be distinguished from others. However, determining the specific causes of newborn deaths is difficult because of the non-specific signs and symptoms in sick newborns (Sachdev 2006). For perinatal deaths in hospitals, where clinicians and other facilities are available, the Wigglesworth method can used to classify causes of perinatal deaths (Wigglesworth 1980,1987; Keeling et al. 1989), but the method is not suitable for data collected at community level.
Although most studies estimate only the cause of death among newborns, for health programming it is equally important to understand the care-seeking processes and treatment actions that occurred before each death (Bojalil et al. 2007). Routine death investigation and verbal autopsy do not include questions on care-seeking before death. Thereby valuable information on any inadequacies or modifiable factors in the home, community, health facilities and referral mechanisms is missed, which could guide programming and policy. To include such questions in a death investigation is an approach known as social autopsy.
A number of studies using various data collection methods and analyses have been conducted to study care-seeking for fatal illnesses or to conduct social/mortality audits in children (Kalter et al. 2003; Krug et al. 2006; Bojalil et al. 2007; Kallander et al. 2008). However, these methods may be unsuitable for newborn deaths because most are related to maternal conditions and care-seeking. Thus, a key challenge for employing social autopsy to understanding newborn deaths in low-income countries is lack of a consistent method for analysing the care-seeking delays that contribute to newborn deaths.
For exploring maternal death, barriers to care-seeking can be characterised by the three delays model developed by Thadeus and Maine (1994). The model comprises delay in deciding to seek care (delay 1), delay in reaching the health facility (delay 2) and delay in receiving quality care once at the health facility (delay 3). Recently, researchers in Tanzania systematically analysed perinatal deaths in a hospital by using the three delays model and also found it useful for perinatal deaths (Mbaruku et al. 2009). Another limitation of this study is that our findings include neither the deaths of babies that occurred among the community nor deaths of older neonates (aged >1 week), leaving a clear knowledge gap. In our community-based study, we aimed to investigate the causes and care-seeking contributors to newborn deaths at <28 days using the three delays model in a defined geographic area in eastern Uganda.
Study setting and data collection
The study was conducted at the Makerere University Iganga/Mayuge Health and Demographic Surveillance Site (HDSS) located in eastern Uganda. The HDSS is about 120 km east of the capital, Kampala, and is located in Iganga and Mayuge districts, which were selected to represent rural Uganda. Over 90% of the population live in rural areas and only one-third of health facilities in the study area provide institutional deliveries. The HDSS has a population comprising about 67 200 people in 65 villages, 18 parishes and 12 000 households. At the time of the study, there were no intervention studies in the HDSS. The neonatal and post-neonatal mortality rates in the HDSS are estimated at 22.3 and 55.2 per 1000 live births (Waiswa et al. 2010), which compares very well with estimates for the entire region (24 and 50 per 1000 live births) (Uganda Bureau of Statistics (UBOS) and MACRO ORC 2007). Forty locally recruited field assistants whose minimum education is 12 years routinely collect data from each household every 6 months. Village-based demographic scouts notify HDSS staff of all deaths and births in the area continuously as they occur. The area has relatively good physical access to health services, including 122 small drug shops and private clinics.
In the Iganga/Mayuge HDSS, deaths are routinely investigated using an InDepth Network (http://www.indepth-net.org) standard verbal autopsy questionnaire, similar in content and administration to those used in Asian studies (Awasthi & Pande 1998; Iriya et al. 2002; Marsh et al. 2003; Baqui et al. 2006). The verbal autopsy tool was merged with a social autopsy questionnaire (Kalter et al. 2004). Whenever a death is reported, the merged questionnaire is administered to a close caregiver of the deceased. At the time of the study, still births were excluded. Sixty-four newborn deaths were investigated covering the period January 2005–December 2008. We conducted a health facility survey in all 16 major public and private health facilities serving the HDSS. These comprised a general hospital, six admitting/delivery care facilities and nine outpatient-only health facilities. Data were collected on physical infrastructure, staff inventory, and on the presence of essential and desirable equipment for newborn care. Finally, knowledge about maternal and newborn care was assessed by a self-administered questionnaire adapted from one used for a similar study (Harvey et al. 2007). The assessment was conducted among 52 health providers selected proportionally to level of care. Briefly, we adapted the tool (available on request from the first author) to measure knowledge along the maternal and newborn continuum of care (MNC) on prevention, diagnosis and management of the three leading causes of maternal deaths, namely haemorrhage, pre-eclampsia and sepsis (Hill et al. 2001), as well as to measure the causes of neonatal deaths (Lawn et al. 2005a, 2006).
Ethical approval was given by the Institutional Review Board of Makerere University School of Public Health. Because verbal autopsy is culturally sensitive, interviews are conducted 4–6 weeks after death to allow a period of mourning in accordance with local customs.
Two practicing medical doctors independently reviewed each death and assigned cause of death using a hierarchical approach (Figure 1) (Baqui et al. 2006). Each doctor had undergone two trainings offered by the InDepth Network in coding newborn deaths. Whenever there was disagreement, they met to review the case, and if agreement was reached, the diagnosis was accepted as the definitive cause of death. However, if agreement was not reached, the cause of death was coded as undetermined. We defined delays as follows (Box 1): Delay 1, which is the delay to recognise illness and the need to seek medical care, included any newborn baby who died at home or where it took more than 12 h to seek outside care; Delay 2, the transport delay, included newborn babies whose caregivers expressed problems with getting transport; and Delay 3, the delay in receiving quality care, included delay in receiving or failure to receive quality care at a health facility (as judged by the audit doctor). Each death could be attributed to one or more delays.
Table Box 1. Delay model for newborn deaths (modified from the maternal three delays model (Thadeus & Maine 1994)
Delay 1: Delay in deciding to seek care
Death at home or,
Delay of more than 12 h to seek care
Delay 2: Delay in reaching a health facility
Transport delay to get to a formal health facility
Delay 3: Delay in receiving quality care
Delay in receiving quality care
Data were entered using FoxPro, cleaned, linked with the HDSS database and then transferred to STATA version 10 for analysis. The study was not powered for statistical significance testing, and data were therefore subjected to standard descriptive analysis. The mean knowledge assessment score was calculated by marking each participant out of 100 and then dividing the total by the number of health workers assessed.
Descriptive characteristics of the study
Of the 64 newborn deaths investigated, 53% (34/64) were boys. Of the newborns, 37% (24/64) had been born in a hospital or a health centre, 23% (15/64) in a private clinic and 39% (25/64) away from a health facility [using traditional birth attendants (TBA), at home or on the way to hospital] (Table 1). Of these deaths, 47% (30/64) occurred within the first 24 h after birth and 78% in the first week, and only 22% occurred in the remaining 3 weeks of the neonatal period (Figure 2). The median age at death was 2 days (IQR 1–4). During the same period, most births were reportedly attended by a trained health worker (58%, 37/64). Twenty deaths (33%) occurred either in a hospital or a health centre and 8 (13%) in a clinic; thus, the majority (54%, 32/60,) died away from a health facility (in care of a TBA, at home or on the way to hospital). Socio-economic status data were available on 86% of the deaths investigated. As expected, there were more deaths among the poorest, with the lower two quintiles alone contributing to 51% (30/59) of the deaths compared to 49% in the upper three quintiles.
Table 1. Descriptive statistics of the study population, N = 64 neonatal deaths
TBA, traditional birth attendants.
Place of birth (n = 64)
On route to hospital & health facility
Who assisted in birth (n = 64)
Age at death (days) (n = 64)
Place of death (n = 60)
On route to health facility/Other
Death by economic status (n = 59)
Quintile 1 (Poorest)
Quintile 5 (Least poor)
Causes of deaths
Sepsis or pneumonia was the leading causes of death, accounting for 31% (20/64) of all newborn deaths (Figure 3), followed by birth asphyxia (30%, 19/64), preterm birth (25%, 16/64) and tetanus (3%, 2/64). In 6 cases (9%), a cause of death could not be determined. Among the 30 deaths which occurred at a health facility, birth asphyxia was the leading cause of death (37%, 11/30), followed by preterm birth (30%, 9/30) and sepsis/pneumonia (27%, 8/30); whereas for death which occurred at home, sepsis/pneumonia was the major cause (35%, 12/34) followed by birth asphyxia (24%, 8/34) and preterm birth (21%, 7/34). Among the 30 deaths which occurred on the day of birth, birth asphyxia was the leading cause of death (40%, 12 cases), followed by preterm births (27%, 8 cases) and sepsis or pneumonia (20%, 6 cases) (Table 2). Among deaths which occurred on day 2 or later, sepsis or pneumonia was the major cause of death (14/26) followed by preterm births (8/26).
Table 2. Causes of newborn deaths by day of death
Day of death
Sepsis or pneumonia
Cause not identified
Delays contributing to newborn deaths
Delay in problem recognition/deciding to seek care outside the home (Delay 1) was the greatest contributor to deaths (50%, 32/64). Most newborn babies who died had started being unwell during or immediately after birth (57%, 36/64) and were unwell for a short period, with the median duration of illness being 2 days (IQR 1–6). Care-seeking was generally delayed, with the median duration to seeking care from outside the home being 3 days from illness onset (IQR 1–6 days). The second major contributor to newborn deaths was delay in receiving quality care at the health facility (Delay 3) (30% 19/64). Of the newborns taken outside the home for care 53% (9/17) reportedly made contact with a qualified health worker, but five caretakers went to drug shops and one to a spiritual leader. Surprisingly, the transport delay to a health facility (Delay 2) was the least impacting contributor to deaths (20%, 13/64). Another, second delay was identified as being a contributor to 22% of the newborn deaths investigated. The major causes of death by main contributing delay were as follows (Figure 4): Delay 1–Sepsis or pneumonia (32%) followed by birth asphyxia (22%); Delay 2–Birth asphyxia (46%) followed by sepsis or pneumonia (31%); Delay 3–Preterm births (37%) followed by birth asphyxia (32%). The distribution of delays for neonatal deaths at health facility was Delay 1, 10 (26%); Delay 2, 11 (28%); and Delay 3, 18 (46%); and for death at home, Delay 1 was the main contributor 22 (88%).
Capacity of health facilities for newborn care
Health facilities had just about half of the minimum qualified health workers recommended by the Ministry of Health, and almost all lacked basic newborn equipment, drugs, supplies and an effective referral system. For instance, only 44% (7/16) of health facilities had a delivery kit, 44% (7/16) had a neonatal weighing scale and only 6% (1/16) had a neonatal resuscitation kit. Participants correctly answered only 58% of the questions about the MNC continuum. Medical assistants/clinical officers had the best mean score (63%), followed by registered midwives (61%), enroled midwives (56.5%) and enroled nurses (50%). Participants were correct mostly for questions on antenatal care (65%), followed by intrapartum care (52%); the fewest correct answers were on newborn/postnatal care (31%).
To our knowledge, this study is the first in which the three delays model originally developed for maternal deaths (Thadeus & Maine 1994) has been employed to explore newborn death using social autopsy data collected at the community level as part of verbal autopsy in a low-income setting. A recent study in Tanzania employed a three delays approach to investigate perinatal deaths in one regional hospital (Mbaruku et al. 2009) but did not include any death occurring at home, yet these deaths are often the majority (Lawn et al. 2004).
In our study, we found that household (Delay 1) and health facility-related delays (Delay 3) were the major contributors to newborn deaths. Together, these two delays contributed to two-thirds of the newborn babies who died. The finding that both health facility and home-related delays are the main contributors to newborn deaths supplements calls for strengthening both health facility and community programmes if newborn care is to be improved in low-income countries (Darmstadt et al. 2005; Haines et al. 2007; Kerber et al. 2007; Baqui et al. 2008).
Our results on the contribution of delays to newborn death differ from those in the Tanzanian study, where most newborns were reported to have died because of Delay 3 (Mbaruku et al. 2009). We attribute this difference to the fact that the Mbaruku study, unlike ours, only collected data from the hospital and did not include older neonates (>1 week). In addition, our study was conducted in an area with reasonably good physical access to health facilities, meaning that transport or Delay 2 would not be a major limiting factor. Our results are consistent with earlier studies reporting that delayed care-seeking and treatment contribute significantly to childhood deaths (Bojalil et al. 2007; Kallander et al. 2008) although these studies did not investigate newborn deaths.
To reduce newborn death, addressing Delay 1 or delay in problem recognition and in deciding to seek outside care will be critical, as it was a major contributor to half of the deaths investigated in this study. We further found that care-seeking for newborn babies was poor and even when caregivers sought care at health facilities it seems the care is substandard. Over half of the newborn deaths had been born or died outside a health facility setting. Furthermore, most newborn babies who did not die on the day of birth were sick for at least 3 days before care was sought outside the home. Given the fact that newborn babies are very vulnerable, it means that care was sought when the newborn was very sick and could not be helped by the already weak health care facilities. A recent ethnographic study in Ghana found that mothers may not be able to recognise serious illness in their newborns, and also they often do not seek care outside the home even when they do realise that their child is seriously ill (Bazzano et al. 2008).
Such findings suggest that besides problem recognition, referral for sick newborns also needs to be improved, both from the community and from first level health facilities, as findings show that most caretakers of sick newborns do not comply with referral advice. Studies in older children conducted in the same setting (Rutebemberwa et al. 2009a,b) and elsewhere in Uganda (Peterson et al. 2004) have shown similar challenges in care-seeking and referral.
We found that health facility-related delays (Delay 3) were the second major contributor to newborn deaths, far more than the transport delay in this setting. Most newborns died on the day of birth or within the first week, and that sepsis or pneumonia, preterm birth and birth asphyxia were the leading causes of death, as has been described before (Lawn et al. 2005a, 2006). However, the contribution of birth asphyxia in our study is higher than in other reports (Lawn et al. 2005a, 2006), implying that the intrapartum quality of care even in health facilities is very poor. Our findings from the health facility survey confirm the lack of capacity for newborn care in this setting. The proportion of health workers who had correct knowledge on maternal and newborn care was very low, especially for intrapartum and postnatal care, and health facilities lacked equipment and drugs for newborn care. Similarly, a recent study of Kenyan hospitals also found that these did not have capacity to manage sick newborn babies (Opondo et al. 2009).
Based on a number of small efficacy studies, almost all from Asia (Darmstadt et al. 2006; Baqui et al. 2008, 2009; Kumar et al. 2008), efforts to scale up newborn care in low-income countries through community-based interventions are gathering pace. Community-based interventions can address delays in problem recognition of sick newborns (Delay 1) by promoting supervised deliveries, birth preparedness and raising awareness on maternal and newborn danger signs. These are some of the practices we have found to be deficient in this setting (Waiswa et al. 2008, Waiswa et al. 2010). However, unless community interventions include treatment and care at home, their success will mainly depend on the strengthening of health facilities so that women in labour and sick newborn babies receive quality care. It has been suggested that introducing maternal and perinatal audit in health facilities (Pattinson et al. 2009), and improving neonatal resuscitation skills among health workers (Wall et al. 2009) are effective strategies to address quality of care issues for newborn babies. However, operationalising this in low-income countries remains a challenge because recent reviews have shown that the understanding of how to reduce health facility-based newborn deaths is still limited (Reyes et al. 1997; Bhutta et al. 2009).
This study adds to our understanding of why newborn babies die by utilising social autopsy data together with causes of death (verbal autopsy) data. However, social autopsy modules to assess care-seeking or other cultural influences are uncommon in routine death surveillance (Thatte et al. 2009). We call for integration of the social autopsy module in all studies investigating maternal, newborn and other childhood deaths to increase and simplify the evidence base for decision making, especially at the implementation level.
This study has investigated a small number of newborn deaths from one region of Uganda. Although we included all the newborn deaths identified in the study area during the study period, there was some under-reporting due mainly to the fact that the study was conducted while the HDSS was being set up, and death-reporting phased in. Furthermore, complete reporting of newborn deaths is known to be problematic and sometimes there is misclassification with stillbirths (Lawn et al. 2004, 2009) which were not recorded at the time. However, while we acknowledge likely missing neonatal deaths, we think the number of newborn deaths assessed here was satisfactory to meet our aim, which was to demonstrate how social autopsy data could be used to further elucidate care-seeking delays for newborn deaths. The bias possibly introduced by missing deaths and excluding stillbirths would likely have understated the importance of the first and second delays. Stillbirth data will therefore be included in forthcoming information collection in the HDSS.
While we depended on the community-based death reporting of the HDSS to conduct the study the location of the HDSS, with relatively good physical access to health facilities, might have understated the relative importance of the transport delay and might therefore limits the generalisability of this finding.
The three delays model, which was originally developed for maternal death, can be adapted to verbal and social autopsy to further understanding why newborn babies die in low resource settings, thereby providing useful information for programmes and policy making. In this rural setting especially, newborn babies died of preventable causes which were contributed to mainly by household and health facility-related delays to care and treatment. To improve newborn survival, interventions need to address both supply and demand-side practices and care. Methods for quantifying newborn care-seeking and treatment delays leading up to death need to be improved and validated to inform better programming and policies.
We thank the study participants, research assistants, Iganga/Mayuge Demographic Surveillance Site, and the Study Policy Advisory Group (Iganga and Mayuge districts, WHO/Uganda, UNICEF/Uganda, SNL/Uganda and MOH). We thank the InDepth Network for supporting the HDSS, Pauline Binder for language editing, and Geoffrey Namara for statistical advice. This study was supported by the Sida/SAREC – Makerere University – Karolinska Institutet research collaboration as well as by funds provided by Save the Children USA, through a grant from the Bill & Melinda Gates Foundation, for a newborn intervention study (UNEST: Uganda Newborn Study ISRCTN50321130). Its contents are solely the responsibility of the authors and do not necessarily reflect the views of Sida/SAREC nor Save the Children USA nor the Bill & Melinda Gates Foundation nor the authors’ institutions of affiliation.