Neurological and developmental outcome of neonatal jaundice and sepsis in rural Kenya


  • Anne L. Gordon,

    1. Centre for Geographic Medicine Research – Coast, KEMRI/Wellcome Trust Research Laboratories, Kilifi, Kenya
    2. Neurosciences Unit, Institute of Child Health, University College London, UK
    Search for more papers by this author
  • Michael English,

    1. Centre for Geographic Medicine Research – Coast, KEMRI/Wellcome Trust Research Laboratories, Kilifi, Kenya
    2. Departments of Paediatrics, University of Oxford, John Radcliffe Hospital, Oxford, UK
    Search for more papers by this author
  • J Tumaini Dzombo,

    1. Centre for Geographic Medicine Research – Coast, KEMRI/Wellcome Trust Research Laboratories, Kilifi, Kenya
    Search for more papers by this author
  • Mary Karisa,

    1. Centre for Geographic Medicine Research – Coast, KEMRI/Wellcome Trust Research Laboratories, Kilifi, Kenya
    Search for more papers by this author
  • Charles R. J. C. Newton

    1. Centre for Geographic Medicine Research – Coast, KEMRI/Wellcome Trust Research Laboratories, Kilifi, Kenya
    2. Neurosciences Unit, Institute of Child Health, University College London, UK
    Search for more papers by this author

Anne L. Gordon, Michael English, J Tumaini Dzombo, Mary Karisa and Charles R. J. C. Newton (corresponding author), Centre for Geographic Medicine Research – Coast, KEMRI/Wellcome Trust Collaborative programme, P.O. Box 230, Kilifi, Kenya. Tel.:+254 44 522063; Fax:+254 44 522390; E-mail:;;;;


Neonatal jaundice (NJ) and sepsis are common causes of neonatal mortality in sub-Saharan Africa, but little is known about the long-term morbidity in this setting. This study aimed to describe the neurological and developmental sequelae of severe neonatal hyperbilirubinaemia and neonatal sepsis (NS) in a district hospital in rural Kenya. Twenty-three term infants with NJ [total serum bilirubin (TSB) >300 μmol/l] and 24 infants with a history of NS were identified from hospital records. These children were compared to 40 children from the community (CC) without neonatal problems. At ages 18–32 months, the children's neurological, motor and developmental status were assessed, and blood groups of the NJ and NS subjects and their mothers were determined. Ten (43%) of the NJ subjects were unable to sit and/or stand independently. The NJ subjects had significantly more neurological, motor and developmental difficulties and caused greater maternal concern than the CCs. Five (21%) of the NJ subjects had possible blood group incompatibility. The NS subjects had significantly more motor and eye-hand difficulties and maternal concerns expressed than the CCs. Severe NJ in term infants (of mainly non-haemolytic origin) was associated with a high prevalence of neurological and developmental sequelae at ages 18–32 months. The NS is also associated with neuro-developmental sequelae, but the pattern is different to those seen in NJ. Since NS is common in resource poor countries, this may be an important cause of neuro-developmental impairment in children living in this setting.


In sub-Saharan Africa, sepsis and jaundice occur in a significant proportion of neonatal hospital admissions. Neonatal jaundice (NJ) was the primary diagnosis in 17% and neonatal sepsis (NS) in 30% of over 1000 admissions during the first 7 days of life, to a rural Kenyan hospital, accounting for 24% and 25% of deaths (English et al. 2003). Little is known about the long-term sequelae of these conditions in this setting.

Neonatal hyperbilirubinaemia is a recognized cause of brain damage (Shapiro 2003), with unconjugated bilirubin causing kernicterus, which results in a dyskinetic movement disorder, gaze abnormalities and sensori-neuronal hearing loss as long-term sequelae (Volpe 2000). Much of the data is based on hyperbilirubinaemia brain damage caused by either Rhesus or ABO blood group incompatibilities, with 36% of neonates with a bilirubin >325 μmol/l developing kernicterus (Volpe 2000). In Africa, NJ is commonly associated with sepsis (Dawodu et al. 1984; Azubuike 1985; Ahmed et al. 1995) but it is unclear if this leads to neuro-developmental impairment. Although infection is associated with poor neuro-developmental outcome in extremely low birth weight neonates in North America (Stoll et al. 2004), there is no data on term neonates in developing countries. We investigated the frequency and nature of neuro-developmental sequelae amongst children admitted to a Kenyan district hospital with NJ or NS and surviving to age 1–3 years.



This study was undertaken in the Kilifi District, a rural community on the coast of Kenya. Assessment took place between August and November 2002. All participants were aged between 18 and 32 months at the time of assessment and lived in an area mapped and censussed as part of an ongoing demographic surveillance.

Three groups of participants were selected for this study, two of which were term infants admitted to Kilifi District Hospital (KDH) in their first week of life (English et al. 2003). The first group had a primary diagnosis of severe neonatal hyperbilirubinaemia with total serum bilirubin (TSB) level above 300 μmol/l documented at least once during admission (hereafter referred to as NJ). The second group were admitted with a primary diagnosis of NS, based upon the World Health Organisation's clinical criteria [i.e. in the first week of life had any of the following – abnormal temperature or respiratory rate, cough, poor feeding, abnormally sleepy or difficult to wake, convulsions or fever (The WHO Young Infants Study Group 1999a,b)] and with bilirubin levels below 300 μmol/l and no clinical features of kernicterus. Children with evidence of central nervous system infection (i.e. pleocytosis in the cerebro-spinal fluid) were excluded from this study. The third group consisted of children from the community (CC) identified through census records as born during the same period as the hospital admissions to act as controls. Children were excluded if age of gestation was <36 weeks. Approval for this study was obtained from the National Medical Research Ethics Committee.

Parental questionnaires

Questionnaires regarding birth history and socio-demographic data were used as previously described (Barlow et al. 2001) with the addition of an open-ended outcome question regarding parental concerns about the child's health and development.

Neurological and motor assessment

An assessment of neurological and motor impairment was devised for this study as no standardized assessment of motor function appropriate to the cultural context, age band or with sensitivity for anticipated movement abnormalities existed. The content was based on a number of standardized assessments including the Peabody Motor Scales (Folio & Fewell 2000), Hawaii Early Learning Profile (Parks 1996), Movement ABC (Henderson & Sugden 1992), Bruininks–Oseretsky Test of Motor Proficiency (Bruininks 1978) and Touwen's assessment of the child with minor neurological dysfunction (Touwen 1979). This assessment was divided into two sections, one assessing motor skills and the other a neurological examination. The motor skills section was divided into three sub-sections – posture and stability; locomotion; and eye-hand and manipulation skills. Items were scored dichotomously based on either objective attainment of a task or movement, or subjective rating of quality of movement. The second section included standard neurological assessment items, with particular emphasis on oculo-motor skills, and physical examination of tone, reflexes, power and movement quality. Diagnosis and description of a movement disorder was also included. For the purposes of this study, a movement disorder was defined as abnormalities of posture and movement, with or without alteration in muscular tone and excluding disorders of movement caused by pyramidal tract or peripheral nerve lesions (Fernandez-Alvarez & Aicardi 2001). ‘Poor’ outcome was defined as the inability to sit and/or stand independently. Assessment of hearing with brainstem auditory evoked responses could not be performed.

Blood grouping

In order to determine the possibility of Rh or ABO incompatibility-related haemolysis as the cause of neonatal hyperbilirubinaemia, the blood groups were determined in the children with NJ or NS and their biological mothers. Grouping was performed by standard methods (Dacie & Lewis 1984). Assay of Glucose-6-Phosphate Dehydrogenase (G6PD) was not possible.

Assessment of development

The developmental assessment, a locally developed tool previously described (Barlow et al. 2001) was administered. Scores are obtained through a combination of observation of capacity in structured activities and parental interview. An overall development score is obtained stratified for age, along with a summated score for each of the three domains (gross motor, eye-hand, speech, language and hearing skills). Higher scores indicate higher level of functioning. In the maternal assessment of development, mothers were asked to rate their level of concern (either ‘very concerned’, ‘some concern’ or ‘no concerns’) in four areas (overall health, language, co-ordination and other).


Field workers visited the homes of identified children, and families were informed of the study and invited to attend for assessment at KDH. All consent and assessment materials were presented in Kigiriama, the local dialect. If a child had died after discharge, parents or family members were interviewed using a ‘post-natal death form’ as described previously (Snow et al. 1992) to determine the date and probable cause of death. The mother and child then participated in a combined assessment session where a locally trained assessor performed the developmental assessment and another the neurological and motor assessment. Aspects of this assessment were videotaped, with consent, for later review. These two assessors conducted all the examinations and neither knew to which group the child belonged. Referrals to local therapy services were made as appropriate after the assessment. Measurements of height, weight and head circumference were taken of all subjects.

Statistical analysis

Data was double entered using FoxPro 2.1 and analysed using Statistical Package for Social Scientists (SPSS Version 11.0). Comparisons were made between groups. For data with a normal distribution an independent samples t-test was performed. Non-parametric data were investigated using Mann–Whitney U-test, or for differences in proportions a Fisher's Exact or Pearson's chi-square test. Percentages reported are based on available data.


From hospital records 38 term infants less than a week old were identified with NJ and 47 fulfilled the definition of NS (The WHO Young Infants Study Group 1999a,b) but did not have any evidence of a central nervous system infection. Of these cases, 38 (15 NJ and 23 NS) were not included in the study, because they had died after discharge (seven NJ and eight NS), lived outside the study area (six NJ and eight NS), refused to participate (two NJ and four NS) and three NS subjects were untraceable. Information regarding four of the seven NJ case deaths was obtained. All children died within the first year of life (range from 1 week to 11 months of age). Symptoms immediately prior to death varied from fever alone, to fever with seizures and diarrhoea, to breathing difficulties.

In total 87 children (48 males) were recruited to this study: 23 NJ subjects (12 males), 24 NS (15 males) and 40 CC subjects (21 males). Blood grouping revealed 5 (22%) NJ subjects had possible ABO incompatibility and 2 (8%) of NS subjects. There were no cases of possible Rhesus incompatibility.

Maternal report of pregnancy, delivery and socio-economic factors

There were no significant differences found between the groups on maternal report of problems during pregnancy or rate of home delivery. Significant differences were however found in the proportion of mothers who reported complications during delivery [seven mothers of NJ subjects (30% of available data), compared to 3 (13%) of NS subjects and 3 (8%) of CCs], and the infant's ability to breastfeed on the first day of life [17 NJ subjects (74%) compared to 13 (54%) NS subjects and 39 (98%) CCs (chi-square P < 0.05)]. No differences were found between groups for the socio-economic variables of maternal age, education, employment or occupation.

Neonatal admission data

Upon admission 22 (96%) of the NJ subjects and 5 (21%) of the NS subjects were clinically jaundiced. The median peak TSB measurement in the NJ subjects was 582 μmol/l (interquartile range 407–757). Age at peak TSB was median 5 days (range 2–9 days), and in most cases the peak was on the day of admission to hospital.

There was no difference between the NJ and NS subjects at the time of admission in age, weight and head circumference or blood glucose. Seven (29%) of the NS subjects had tachypnoea [respiratory rate >60 breaths per minute (McIntosh et al. 2003)], compared to none of the NJ subjects (Fisher's Exact Test P = 0.010). Three NJ subjects had overt features of kernicterus (opisthotonus) during the hospital stay.

Fifteen (65%) of the NJ subjects had phototherapy for a median of five days (range 2–13 days). Twelve (52%) of the NJ subjects also had one double-volume exchange transfusion. In those children who received an exchange transfusion, the median peak TSB was 690 μmol/l (range 352–781) and the median age at peak was 4 days of age (range 3–6). The median peak TSB of the remaining cases was 417 μmol/l (range 317–783). Guidelines for exchange transfusion were not always followed, apparently due to a lack of staffing and/or blood.

Assessment at age 18–32 months

The NJ and NS subjects were smaller than the community controls in weight, height and head circumference with the NJ subjects smallest in all three parameters (Table 1), although this group also had a younger median age. In the absence of Kenyan centile charts for head circumference, proportioning subjects within groups below the second centile on UK charts (Freeman et al. 1995) found that more NJ subjects had microcephaly than the community controls, suggestive of impaired brain growth in the children with NJ. There were also significant differences found between the NJ and CC groups in motor, neurological and developmental domains (Table 1). The NS subjects also had more difficulties across these domains than the CCs, but the frequency and distribution of motor and developmental difficulties differed from that of the NJ subjects.

Table 1.  Clinical features at time of assessment
Domain assessedNJ (n = 23)NS (n = 24)CC (n = 40)Comparison of NJ and CC groupsComparison of NS and CC groups
  1. N.B. Percentages calculated from numbers that could be assessed.

  2. SD, Standard deviation.

  3. † Pearson's chi-square.

  4. ‡ Mann–Whitney Test.

  5. § Percentiles based on UK reference charts (Freeman et al. 1995).

Age (months) median (IQR)23 (19, 27)24 (19, 29)27 (24, 30)0.040NS
Anthropometric data
 Weight (kg) mean (SD)9.6 (2.0)9.7 (1.6)10.6 (1.3)0.0160.011
 Height (cm) mean (SD)78.4 (5.8)79.5 (4.8)81.5 (4.2)0.011NS
 Head circumference (cm) mean (SD)46.1 (1.7)46.4 (2.5)47.8 (1.4)<0.0010.007
 Head circumference below second centile§ no. (%)19 (83)14 (58)16 (40)0.001†NS
Eye signs no. (%)
 Unable to elicit horizontal optokinetic nystagmus12 (52)1 (4)0 (0)<0.001†NS
 Smooth pursuit disturbed11 (48)2 (8)1 (3)<0.001†NS
 Saccades disturbed11 (48)2 (8)3 (8)0.001†NS
Neurological signs no. (%)
 Movement Disorder11 (48)0 (0)3 (2)0.001†NS
 Dystonia8 (35)0 (0)0 (0)<0.001†NS
Motor signs no. (%)
Posture and stability
 Unable to sit unsupported8 (35)0 (0)0 (0)<0.001†NS
 Unable to stand unsupported10 (43)4 (16)0 (0)<0.001†0.008†
 Unable to walk independently11 (48)4 (16)0 (0)<0.001†0.008†
Eye-hand and manipulation
 Unable to take lid off jar using two hands11 (48)5 (21)0 (0)<0.001†0.009†
 Unable to stack two blocks10 (43)5 (21)0 (0)<0.001†0.009†
 Unable to achieve pincer grasp with preferred hand9 (38)2 (9)0 (0)<0.001†NS
Developmental scores median (range)
 Total score37 (31, 42)35 (32, 39)42 (36, 48)<0.001‡0.002‡
 Motor11 (9, 13)11 (10, 12)12 (10, 14)NS‡NS‡
 Hearing, speech and language7 (5, 9)6 (4, 8)8 (5, 11)0.001‡0.005‡
 Eye-hand co-ordination20 (14, 26)20 (14, 26)23 (19, 27)0.043‡0.016‡

On neurological and motor assessment the NS subjects had significantly more motor difficulties than CCs – particularly in inability to stand (16%) or walk independently (16%), use two hands to remove a lid from a jar (21%) or stack two blocks (21%) compared to no difficulties in these areas in the CC subjects (Table 1). On developmental testing, the NS subjects scored lower than CCs in eye-hand co-ordination, speech and hearing and overall development score, but not motor development.

Although none of the NJ subjects had clinically apparent neurological sequelae at the time of discharge as neonates, when assessed 18–32 months later, they were more likely to have neurological sequelae, motor and developmental disturbances than CCs. Half of the NJ subjects [12 (52%)] had at least one eye-movement disorder including inability to elicit horizontal optokinetic nystagmus and/or disturbance of saccades and smooth pursuit. The presence of eye-movement disorder differed significantly from both the NS and CC subjects.

Eleven (48%) of the NJ children had a dyskinetic movement disorder compared to none of those with NS. The NJ children had generalised dystonia (n = 6), generalised dystonia with athetosis (n = 2) and chorea (n = 3). In terms of motor skills, the NJ subjects were significantly more likely to have difficulties in all three sub-sections of the motor assessment (Table 1). The NJ subjects had lower developmental scores (indicating greater difficulties), both overall and in each subsection than both the NS and CC subjects. The neuro-developmental difficulties detected in the NJ and NS children is reflected in the significantly raised concerns expressed by their mothers – with 44% of NJ mothers and 21% of NS mothers ‘very concerned’ about their child's health or development, compared to none of the CCs.

Neonatal hyperbilirubinaemia – good vs. poor outcome

Ten (43%) of the NJ subjects were classified as having poor outcome. The median TSB in the poor outcome group was 752 μmol/l [interquartile range (IQR) 651–768] compared with a median of 401 μmol/l in the good outcome group (IQR 331–563). The children in the poor outcome group accounted for the majority of neurological and motor difficulties seen overall in the NJ subjects. All of the children in the poor outcome group had at least one abnormal neurological sign, whereas 9 (69%) of the good outcome group had no abnormal neurological signs. In terms of motor abilities none of the children in the poor outcome group were able to walk independently [compared to 12 (92%) of the 13 children in the good outcome group], none were able to take a lid off a jar using two hands nor stack blocks (92% and 100% in good outcome group, respectively), and only one (10%) was able to achieve a pincer grasp (100% in good outcome group). Only one of the seven children with possible ABO incompatibility, an NJ case who had also had an exchange transfusion, was rated as having poor outcome (Fisher's Exact Test P = 0.018), with the presence of a dystonic movement disorder. Of the 12 NJ subjects that had an exchange transfusion, six had a good outcome and six had a poor outcome. There was no significant difference in outcome between those having an exchange transfusion and those who did not.


Neonatal sepsis and neonatal jaundice are common conditions in Africa, where they account for a large proportion of infant mortality. This study found that both of these conditions are associated with neuro-developmental sequelae, although the nature and severity differed between the children admitted with NJ compared with those with NS. Although the neuro-developmental impairment in the NS subjects was less widespread and severe than in NJ, this is important in sub-Saharan Africa, since NS occurs more than twice as frequently as NJ (English et al. 2003).

Of children admitted with severe NJ, 18% died after discharge. At age 18–32 months, 22 (96%) of the NJ survivors had motor or neurological impairment and/or developmental difficulties. These children frequently had multiple disabilities: 43% were unable to either sit and/or stand independently, 48% had a movement disorder and 56% had an eye-movement disorder. Developmental disturbance and parental concerns were more common than in children from the community. This adverse outcome in terms of disability represents a significant personal and socio-economic burden to families in this region, and is considerably higher than that reported in Western countries (Newman & Klebanoff 1993; Grimmer et al. 1999; Harris et al. 2001; Newman et al. 2003).

This study suggests that NS is also an important cause of morbidity in sub-Saharan Africa with nearly one-sixth of children unable to walk, and one-fifth unable to use their hands together in functional activities. The morbidity in children with a primary diagnosis of NS cannot be explained by the possible co-existence of NJ, since the patterns of sequelae are different. Neuro-developmental impairment has been documented in extreme low birth weight American neonates who developed sepsis (Stoll et al. 2004), but our study suggests that term infants are at risk in tropical areas. Functional disability for NS survivors in terms of difficulties in standing, locomotion and eye-hand co-ordination in the absence of eye- or limb-movement disorders will have long-term consequences for the individual and the family.

In sub-Saharan Africa, ABO incompatibility has been described as a significant cause of haemolytic disease of the newborn (Khan 1965; Waterston & Kuchena 1983; Owa et al. 1991), particularly in those infants with severe hyperbilirubinaemia (>300 μmol/l) (Vos et al. 1981). In this retrospective study it was not possible for us to define aetiology of neonatal hyperbilirubinaemia, however, none had Rhesus incompatibility and less than one quarter of our NJ subjects had the possibility of ABO haemolytic disease. It has been suggested that the correlation between TSB concentration and kernicterus is poor in term infants without haemolysis (Wolf et al. 1997; Hansen 2002) and further that the risk for neurological sequelae is lower in these infants than in those with haemolytic disease (Wolf et al. 1999; Volpe 2000; Dennery et al. 2001; Hanko et al. 2001). However, our results do not support this suggestion, as at least half of the poor outcome patients have the typical motor and eye-movement impairments of kernicterus.

In this retrospective study, it was difficult to obtain specific data regarding perinatal events since the majority of children were born at home, thus we cannot exclude the possibility of undetected factors, for example, birth trauma influencing outcome. A study undertaken in Nigeria suggested that jaundice and kernicterus are more severe in children born at home than in those delivered in hospital (Ahmed et al. 1995). Low birth weight and delay in reaching hospital may influence the outcome. In addition, the amount of data on early bilirubin levels, such as the duration and peak of bilirubin exposure was incomplete from this district hospital in Kenya. Coombs tests on admission or G6PD status were not determined. The identification of the cause of NS is difficult within this context, but bacteria were isolated from 9% of the neonates who fulfilled this definition in this hospital (English et al. 2003). We could not confidently exclude sepsis in the NJ group. Furthermore the differences between the groups in the developmental tests may have been influenced by the ages of the groups. However, the pattern of sequelae was markedly different and the significant differences between the NJ and CC groups in head circumference were based upon differences in the age-related proportions on head circumference charts.

This study shows that sepsis and severe jaundice, both of which are common in neonates admitted to district hospitals in sub-Saharan Africa, are important causes of long-term morbidity. In this setting, the scarcity of health service resources severely restricts access to early detection and acute intervention for children with NS and NJ. Although NS causes less severe impairment, its impact may be greater since it is more common than severe NJ. For NJ, guidelines for management need to be established for sub-Saharan Africa, since it appears that the causes may be different, in particular non-haemolytic causes are more common.


We thank the children and their families who participated in this study. This study is published with the permission of the Director of the Kenya Medical Research Institute. Also we thank Dr Penny Holding for the developmental assessment, Victor M'ungala Odera and Douglas Konde for locating children, Maimuna Mohammed for assessing the subjects as neonates, Rachel Odhiambo for data entry, Silas Haro for co-ordinating the blood sampling and Professors Brian Neville and Robert Surtees for their helpful comments on earlier drafts of this paper. Charles Newton holds a Wellcome Trust Career Post in Clinical Tropical Medicine (No 050533). This study was funded by the Wellcome Trust/Kenya Medical Research Institute Collaborative Programme.