High levels of pathological jaundice in the first 24 hours and neonatal hyperbilirubinaemia in an epidemiological cohort study on the Thailand-Myanmar border

Population risks for neonatal hyperbilirubinaemia (NH) vary. Knowledge of local risks permits interventions that may reduce the proportion becoming severe. Between January 2015 and May 2016, in a resource-limited setting on the Thailand-Myanmar border, neonates from 28 weeks’ gestation were enrolled into a prospective birth cohort. Each neonate had total serum bilirubin measurements: scheduled (24, 48, 72 and 144 hours of life) and clinically indicated; and weekly follow up until 1 month of age. Risk factors for developing NH were evaluated using Cox proportional hazard mixed model. Of 1710 neonates, 22% (376) developed NH (83% preterm, 19% term). All neonates born <35 weeks, four in five born 35–37 weeks, and three in twenty born ≥38 weeks had NH, giving an overall incidence of 249 per 1000 livebirths [95%CI 225, 403]. Mortality from acute bilirubin encephalopathy was 10% (2/20) amongst the 5.3% (20/376) who reached the severe NH threshold. One-quarter (26.3%) of NH occurred within 24 hours. NH onset varied with gestational age: at a median [IQR] 24 hours [24, 30] for neonates born 37 weeks or prematurely vs 59 hours [48, 84] for neonates born ≥38 weeks. Risk factors for NH in the first week of life independent of gestational age were: neonatal G6PD deficiency, birth bruising, Sgaw Karen ethnicity, primigravidae, pre-eclampsia, and prolonged rupture of membranes. The genetic impact of G6PD deficiency on NH was partially interpreted by using the florescent spot test and further genotyping work is in progress. The risk of NH in Sgaw Karen refugees may be overlooked internationally as they are most likely regarded as Burmese in countries of resettlement. Given high levels of pathological jaundice in the first 24 hours and overall high NH burden, guidelines changes were implemented including preventive PT for all neonates <35 weeks and for those 35–37 weeks with risk factors.


Introduction
Neonatal hyperbilirubinemia (NH), characterized by elevated levels of total serum bilirubin (TSB) in the blood, is a common condition in neonates. An incorrect or delayed diagnosis may put neonates at risk of developing neurologic dysfunction such as acute bilirubin encephalopathy or a range of subtle to severe kernicterus spectrum disorders [1,2]. Resource-limited settings endure most of the burden worldwide as public awareness, access to care, and outpatient follow up are often limited: at least 75% of the affected infants reside in sub-Saharan Africa and South Asia [3].
The risk of developing NH depends not only on the physiologic status of the neonate at birth and maternal and environmental factors but also parents' care-seeking behaviour, health care service monitoring tools, and practice management [4]. Timely recognition and treatment maximize the likelihood of NH remaining uncomplicated. The challenge is to identify neonates at risk of NH as early as possible to plan optimal monitoring. The National Institute for Health and Care Excellence (NICE) NH guideline [5] has been effective in decreasing the number of neonates with severe NH (13% in 2009-2011) and NH mortality on the Thailand-Myanmar border [6]. This is a resource-limited setting with a high prevalence of glucose-6-phosphate dehydrogenase (G6PD) deficiency: 13.7% in adult males and 2-4% in adult females [7,8].
This study aims to describe the epidemiology of NH in the aforementioned setting in an actively screened birth cohort from 28 weeks gestation, including: incidence, mortality, time of NH onset by gestational age, and factors associated with NH in the first week of life. This information will identify if any adaptation of the current international guidelines is needed.

Study area and population
This prospective observational cohort study included live births between January 2015 and May 2016 in facilities run by the Shoklo Malaria Research Unit (SMRU). These facilities serve marginalized Karen and Burmese populations, comprising migrants and refugees. Free maternal and child healthcare is offered and includes pre-and postnatal care, and a birthing unit staffed by skilled birth attendants [9]. Low income and education, and poor health conditions prevail [10][11][12]. GA is estimated either by ultrasound (US) at first antenatal consultation (ANC) or by Dubowitz gestational assessment for late attenders [13,14]. Women needing caesarean section (no elective procedures) are referred to a local hospital. Following the WHO recommendations, delayed cord clamping is routine unless the neonate requires urgent resuscitation [15]. Most neonates are nursed on the post-natal ward but if an elevated level of neonatal care is required e.g. phototherapy, intravenous antibiotics, nasogastric feeding or oxygen, they are transferred to the Special Care Baby Unit (SCBU), another department within the SMRU facilities [16].

Inclusion criteria
Any liveborn neonate �28 weeks' gestation was included if seen within 48 hours (h) of birth, or if presenting with jaundice in the first week of life, as homebirth still occurs in this setting.

Follow up
Follow up in the first week of life included TSB, haematocrit (HCT), and weight; and it was deemed 'complete' if a minimum of 3 TSB measurements were available: I) one TSB before or at 30 h of life, II) one TSB measured at a maximum of 36 h after the first one, and III) one TSB between 120 and 168 h (5-7 day of life). An additional TSB was done if the second TSB was performed between 31 to 60 h to ensure that no more than 60 h spanned between measurements taken before 120 h of life. The neonate was considered NH-free between 120 h and 168 h if the TSB measured at 120 h of life was at least 50 μmol/l below the treatment threshold.
Neonates were then followed up (clinical exam, weight, visual assessment of jaundice, and TSB if clinically indicated) every week until one month of age [17]. Mothers were encouraged to bring unwell children to the clinics any time in-between scheduled appointments for examination and treatment.

Definitions
Neonatal hyperbilirubinemia. The NICE hour-specific treatment thresholds based on GA at birth were used to determine NH requiring phototherapy (PT) [5]. NH was defined as moderate when at least one TSB was on or exceeding the PT threshold and severe if at least one TSB was on or exceeding the exchange transfusion threshold. NH diagnosed within the first 48 h of life was categorised as 'early NH', while NH occurring between 48-168 h of life was defined as 'late NH'. The 48 hours cut-off was based on the median duration stay in the postnatal ward for a mother-child dyad with uncomplicated delivery in this setting.
Type, timing, and duration of PT were recorded. Light intensity was measured with a digital light meter (Lightmeter by Medical Technology Transfer and Services Ltd; measuring range 0 to 150 μW/cm 2 /nm; accuracy ± 2%) prior to starting phototherapy conditions were optimized to have the best light intensity possible depending on the severity of the NH [18]. TSB measurement frequency while receiving PT and treatment discontinuation followed the recommendations of the NICE guidelines [5]. The site medical doctor and the head of the department discussed care of each neonate requiring exchange transfusion prior to transfer to the local Thai Hospital in Mae Sot (60 km away) as the procedure was not available in the clinics. A diagnosis of acute bilirubin encephalopathy was considered in the presence of acute onset of neurological signs such as seizure or opisthotonos.
Cohort variables. Cohort characteristics were grouped into four categories: maternal, obstetric, neonatal characteristics and clinical events.
Available known risk factors associated with an increased risk of developing NH, as well as characteristics specific to the study site: migrant or refugee status; use of naphthalene balls; neonate G6PD status, and ethnicity; were included. The latter was based on both parents and grandparent's ethnicity and reported as "mixed" when parents' ethnicity differed. Of note, people of Islamic faith self-identified as "Burmese Muslim".
Maternal body mass index (BMI) before delivery was calculated using the last weight within 2 weeks of delivery [19]; a BMI �27.5 mg/kg 2 was categorized as overweight according to the WHO expert consultation on Asian populations [20]. Neonatal resuscitation was defined as the provision, at minimum, of positive pressure ventilation via bag and mask [21]; birthweight was valid if measured within the first 72 hours of life, and typically in the first 30 minutes [22]. Weight-for-gestational age adjusted for sex was calculated according to Interbio-21 international standards and categorized as small for GA (SGA) if lower than the 10 th percentile [23].
Clinical events other than neonatal hyperbilirubinemia were SCBU admission and three events occurring within the first week of life but prior to NH: excessive weight loss defined as �7% difference between birth weight and subsequent weights [24,25]; polycythaemia was defined by proxy as capillary HCT �70% as the only onsite measure available; and clinically diagnosed severe infections such as sepsis, meningitis, necrotizing enterocolitis or pneumonia, treated with a minimum of five days of intravenous antibiotics.

Blood sampling
G6PD fluorescent spot test, ABO & rhesus blood group determination, and direct Coombs test were measured on cord blood or venous blood at time of first visit for neonates born outside SMRU birthing unit. The ABO incompatibility definition was limited to neonates with A or B blood type born to type O mothers [26]. TSB and HCT measurements were done by capillary heel prick. This blood sample was centrifuged three minutes at 10 000 rotations per minute to separate red blood cells from plasma; HCT was first estimated using a Hawksley micro-haematocrit reader, then the sample was used to assess TSB using a micro-bilirubinometer (Pfaff Medical Bilimeter 2 and 3). G6PD fluorescent spot test was by double blind read at birth, repeated at one month, and the neonate final G6PD status amended according to the test result obtained at one month [27].
Moderate and severe NH incidence was calculated by survival analysis using data from neonates who had a complete regular follow up in the first week of life as defined in the Methods. One TSB per neonate with complete regular follow up measured at, or the nearest to 24 h (12-30 h), 48 h (31-60 h), 72 h (61-80 h) and 144 h (81-168 h) was used to describe TSB levels trend. Only TSB measurements prior to the initiation of PT were included. Time-point TSB median values were compared using Kruskal-Wallis Test.
Potential risk factors for developing NH were evaluated using a Cox proportional hazard mixed model, clustered by site. Variables with p-value <0.2 in univariable analysis, as well as risk-factors for NH described in the literature, were included in a multivariable model, and adjusted for confounding factors. Variables presenting obvious correlation (e.g., maternal age and primigravida, primigravida and history of jaundice in sibling, potential ABO incompatibility and Coombs test, absence of delayed cord clamping and the need of resuscitation or the place of delivery, gender and G6PD deficiency in the neonate, or refugee or migrant status and ethnicity) were not included simultaneously in the models. Analysis of potential risk factors for developing early and late NH excluded neonates who developed late and early NH, respectively.

Ethics approval and consent to participate
The study was approved by Oxford Tropical Research Ethics Committee, UK (OxTREC 41-144), the Mahidol University Faculty of Tropical Medicine Ethical Committee, Thailand (TMEC 14-012) and the Tak Province Border Community Ethics Advisory Board (TCAB-08-13). Informed consent was obtained from the parent or guardian of the neonate: written for those who were literate or by thumbprint in the presence of a literate witness for illiterate.

Results
There were 1710 neonates included of 2628 women attending the facility during the study period, and 95.3% (1630/1710) completed the one-month follow up (Fig 1).

NH incidence in the first week of life and by gestational age
Neonates born by caesarean section in referral hospitals and late presenters did not meet the definition for the sub-cohort of regular complete follow up in the first week of life as they missed the required windows of TSB measurements. Three-quarters (1283/1710) of the cohort including 84.8% of the neonates with NH (319/376) with regular, complete follow up were

TSB levels trend
Neonates born at a gestational age �34 weeks were excluded from this part of the analysis as all of them developed NH prior to or at the 24 h time-point measurement. Based on the prevalence and time of onset of NH, neonates were categorized as: born at a gestational age 35-37 weeks or �38 weeks.

Factors associated with an increased risk to develop NH
The general characteristics of neonates NH-free and those of neonates with early NH (within 48 h of life) or with late NH (48-168 h of life) are presented in S1

Time points in hours (min-max)
TSB a , μmol/L TSB a , μmol/L  Table 3. Maternal pre-eclampsia or eclampsia, prolonged rupture of membranes and birth bruising or haematoma were not associated with an increased risk of late NH (Table 4).

Discussion
This study conducted on the Thailand-Myanmar border, in a setting where neonatal care but not intensive neonatal care is available, reports a high NH incidence during the first week of life (249 per 1000 livebirths) in neonates born from 28 weeks' gestation. One in five neonates in this setting required phototherapy for NH at median [IQR] of 50 h . The increase in NH from 18.5% previously reported from this setting (2009 to 2014) [6] to 22% in this study, most likely results from the systematic TSB screening from 24 hours of life onwards including follow up to the end of the first month of life. Severe NH is 25 times higher than that observed in high-income countries (11 vs 0.4 per 1000 livebirths [29]), although systematic TSB screening is the most likely reason for its decline from 14.9% in 2011 [6] to 5.3% during this study.

PLOS ONE
Neonatal hyperbilirubinaemia and pathological jaundice (1st 24 hours): An epidemiological cohort study Systematic TSB screening and access to PT using the high-income country NICE international guidelines has been a positive effort towards reducing neonatal mortality as recommended under the global strategy of the Every Newborn Action Plan (ENAP) [30] but the study findings also indicate the guidelines are not optimal for this setting. UK, as well as the American guidelines propose that TSB should not be done routinely unless the neonate is visibly jaundiced in first 24 h of life or jaundice appears excessive for the age [5,31]. Jaundice can be visualized at TSB values between 85-100 μmol/L [32] but it may be more difficult in darker skin neonates, particularly preterms, and those <35 weeks would likely reach the PT threshold prior to having a TSB measurement done. The UK guidelines recommend NH cases within the first 24 h of life be referred to tertiary hospital for medical review, while neonates <38 weeks of gestation, those exclusively breastfed or with a previous sibling who had NH should remain in close monitoring for 48 hours-an impractical recommendation in a setting where nearly all neonates are exclusively breastfed from birth and access to referral care is extremely limited. Care for preterm neonates born <35 weeks of gestation, representing 5% of livebirths in this cohort, is not included in the American guidelines and although the PT thresholds are detailed for each gestational week from 23 until 37 in the NICE guidelines, there is no specific guidance for monitoring NH in those <35 weeks.
Several of the study findings are challenging to solve for a resource-limited setting and could be relevant to other health centres caring for born-too-soon and born-too-small neonates. As far as we are aware this study represents the first report of a high proportion (26%) of  [33]. Early screening is therefore paramount to avoid delaying the start of PT and thus reducing the risk of bilirubin-induced neurological dysfunction; as in this study, the value of the first TSB screening done at 24 h of life or earlier was already above the gestational-age adjusted PT threshold in nearly all preterm neonates. PT within the first 24 h has shown to be effective in reducing NH among preterms [34] and those born 28 to 34 weeks of gestation could probably benefit from low irradiance prophylactic PT [35]. Concerns of increased mortality in extremely low birth weight infants under PT [36] are not as relevant in this setting where mechanical ventilation is not available and survival below <28 weeks of gestation is unlikely [28]. However, it is impractical to offer prophylactic PT to all neonates born at 35 to 37 weeks' and tools to assist in identification of those with high risk to develop NH, such as the use of a non-invasive cord blood TSB measurement [37], are being tested at the Shoklo Malaria Research Unit clinics, as a possible method to detect rising TSB levels at the earliest stage. The 18.8% of NH among term neonates observed in this study represents a heavy burden: twice the estimates of 1 in 10 neonates requiring phototherapy annually worldwide [32]. This cohort was closely monitored, and this high prevalence is not explained because families presented late for TSB check as may be expected in a LMIC setting. This places a significant workload and intensity of TSB monitoring that is uncomfortable (multiple heel pricks because transcutaneous devices are beyond LMIC resources) for neonates and their families. Furthermore, NH occurs at a median [IQR] time of 59h [48-84h] in neonates born 38 weeks or more, and this has clinical repercussions. Otherwise healthy term neonates are usually discharged within 24-48 hours in a setting like the one on the Thailand-Myanmar border and due to transportation and access difficulties parents are not always able to return for timely followup.
Severe NH was observed in this setting after a first diagnosis of moderate NH. This highlights the need for constant vigilance once PT has started and for tools that improve detection of risk including among term infants.
Apart from prematurity and G6PD deficiency there were four independent risk factors specific to an increased risk of developing early NH, within the first 48 hours of life: maternal preeclampsia or eclampsia, prolonged rupture of membranes, both recognized in a systematic literature review on maternal risk factors for NH [38], a high neonatal haematocrit and the presence of birth bruising or haematoma. Of these factors, only birth bruising or haematoma is listed as major risk factor by the American Academy of Pediatrics guidelines for closer monitoring [18]. One additional context-specific factor was recognized: neonates of Sgaw Karen ethnicity showed an increased risk to develop NH in the first week of life compared to the other ethnic groups. A study in 105 neonates of different Asian ethnicities proposed differences in genes involved in bilirubin metabolism or environmental factors as a possibility for NH readmission [39]. In this setting, Sgaw Karen have a higher frequency of G6PD deficiency [7] and analysis of genetic factors involved in bilirubin metabolism is ongoing. Knowledge of an increased NH risk in Sgaw Karen ethnicity is valuable locally in Thailand and Myanmar, and internationally given that more than 100 000 Sgaw Karen refugees from the Thailand-Myanmar border, most likely regarded as Burmese, now reside in at least 14 countries, following the largest UNHCR relocation programme globally [40].
Based on the results from the current study, a decision to modify the current guidelines using a combination of TSB levels and risk factors was implemented, aiming to provide the earliest diagnosis and treatment to those the most at risk of NH while still providing adequate monitoring and follow up to those deemed less at risk. As 6 in 10 neonates <38 weeks developed NH within 24 h, preventive PT for neonates <35 weeks, and for those born 35 to 37 weeks with risk factors, was implemented using a simple checklist (S1 File). For those born 38 weeks or more the checklist aims to alert the staff to neonates with an increased risk to develop NH so they can plan for the follow-up.
The strength of the results presented relies on a large dataset of routinely collected clinical and laboratory variables and a low proportion of missing information. Notwithstanding, there are limitations: neonates with potentially high NH risk (e.g., born by caesarean section after a rupture of membranes �18 hours) were excluded and results likely underestimate NH incidence and proportion of severe jaundice. Exclusive breastfeeding was universal including in preterm neonates and its association with NH could not be measured. No score of acute bilirubin encephalopathy was used preventing international comparison [41]. G6PD deficiency was evaluated using a qualitative phenotypic method (FST test, [27]) and the effect of G6PD genotypes on risk of NH was not evaluated. The checklist developed requires access to simple laboratory services e.g., for blood grouping, HCT or TSB and would not be helpful to reduce the risk for severe jaundice in homeborn neonates, a major risk for exchange transfusion in rural Myanmar [42] where the government supports auxiliary midwives [43] and two in three births still take place at home.

Conclusions
In this resource-limited setting with a dedicated special baby care unit but not intensive care nor timely access to exchange transfusion, this epidemiological study identified a high NH prevalence, 1 in 4 developing within 24 hours of life. Active screening likely lowered severe case incidence but mortality due to encephalopathy was still 1 in 10, and this is probably an underestimate. Guidelines were modified towards early NH identification and treatment and adequate follow up of those at risk of late NH. Sgaw Karen have the highest risk of NH according to ethnicity but are internationally known as Burmese refugees. The increased risk of NH in neonates identified as G6PD deficient by fluorescent spot test most likely underestimate the risk in female neonates and further research is currently evaluating this. This study demonstrates that mortality and potentially crippling morbidity from severe NH can be managed in a resource limited setting that can provide screening for TSB and offer phototherapy based on guidelines.  Table. Uni-and multivariable analysis of potential risk factors for developing NH in the first week of life (within 168 hours) using Cox proportional hazard mixed model clustering by site, n = 1283. Data are presented as n (%). a Adjusted for Primigravida, Pre-eclampsia or eclampsia, Rupture of membrane � 18 h, Delayed cord clamping, Gestational age <38 weeks, Birth bruising or haematoma, Sgaw Karen ethnicity, G6PD deficiency (by FST), ABO incompatibility, Severe infection at 0-24 h, Weight loss �7% at 24 h, and Polycythaemia at 24 h. Gender was not included because it was highly correlated with G6PD deficiency (by FST). b Rupture of membrane �18 h n = 1258, Birth bruising or haematoma n = 1282, Ethnicity