Prevalence and associated factors of admission hypothermia among neonates admitted to the premature baby unit of a secondary care hospital in Sri Lanka: a cross-sectional analytical study

Abstract

Background

Hypothermia is defined as core body temperature being below 36.5°C. This study aimed to identify the prevalence, associated factors and outcomes of hypothermia among neonates admitted to the premature baby unit (PBU) of a secondary care hospital in Nawalapitiya, Sri Lanka.

Methods

In a cross-sectional analytical study, medical records of neonates admitted to the PBU were selected retrospectively from 31 March 2022 to 1 April 2021 using consecutive sampling. The recorded axillary temperatures on admission, sociodemographic and clinical data were extracted. Following bivariate analysis, multivariable logistic regression was performed.

Results

Among 407 neonates, the median admission age was 1 day. The majority were male (52.6%), were term (59%) and had a normal birth weight (52.5%). The number of babies with hypothermia was 157 with a prevalence of 38.6% (95% CI: 33.9 to 43.4). Maternal factors like teenage pregnancy, multiple pregnancy, hypertension during pregnancy, premature rupture of membranes and lower-segment caesarean sections; neonatal factors like age on admission being less than 24 hours, prematurity, corrected gestational age on admission being less than 37 weeks, low birth weight, weight on admission being less than 2.5 kg and having been resuscitated at birth had statistically significant associations with hypothermia on the bivariate analysis. Hypothermia showed no significant association with the month of admission. Following multivariable analysis, age on admission being less than 24 hours (adjusted OR (aOR): 3.3, 95% CI: 1.9 to 5.8), teenage pregnancy (aOR: 8.2, 95% CI: 1.8 to 37.2), multiple pregnancy (aOR: 2.8, 95% CI: 1.1 to 7.1) and hypertension in pregnancy (aOR: 2.3, 95% CI: 1.2 to 4.7) remained statistically significant. Neonates with hypothermia had 5.2 times (95% CI: 1.8 to 14.6) odds of mortality and 4.9 times (95% CI: 2.8 to 8.5) odds of receiving ventilatory support compared with normothermic neonates. Hypothermia also showed statistically significant associations with infant respiratory distress syndrome, metabolic acidosis and neonatal jaundice.

Conclusions

Nearly two out of five neonates admitted to the PBU were hypothermic. There were significant maternal and neonatal associations to be addressed. Hypothermia on admission may indicate serious neonatal morbidity and mortality.

WHAT IS ALREADY KNOWN ABOUT THIS TOPIC.

WHAT THIS STUDY ADDS

• This study adds evidence on the prevalence of admission hypothermia among newborns, and the maternal, neonatal and environmental factors independently associated with it. It also highlights the specific neonatal complications to which hypothermia is linked.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• These findings can be used to identify risk groups during neonatal care and inform advocacy and policy for thermoprotective interventions within hospitals.

Introduction

Hypothermia is defined as a core body temperature that is less than 36.5°C. Neonates—newborns who are less than 29 days of age—are at a higher risk for hypothermia than adults. Hypothermia manifests in neonates even without major exposure to cold temperatures and leads to greater morbidity and mortality. It is estimated that annually 17 million neonates become hypothermic in low- and middle-income countries, mostly in countries with a higher burden of neonatal mortality.¹ Neonatal hypothermia is not limited to countries with cold climates but is similarly prevalent in tropical regions.2,5 It has also been identified as a neglected challenge worldwide.⁶ These facts underscore the importance of addressing neonatal hypothermia in the global agenda for newborn care.

The vulnerability of neonates to hypothermia is related to physiological factors in the neonatal period. After birth, heat loss mechanisms including radiation, evaporation, convection and conduction take effect. The rate of heat loss may be as high as 100–200 calories/kg/min resulting in a temperature drop of 0.2–1°C/min.^{7 8} Neonates face greater heat loss in comparison with heat generation as they have a large surface area compared with body mass, thin and heat permeable skin, very little subcutaneous fat and poorly developed autonomic and chemical pathways including poor heat conservation mechanisms such as shivering.9,11 Hypothermia can occur during routine baby care following delivery, during resuscitation or during transportation. Therefore, studying the prevalence and associated factors of hypothermia among neonates that are admitted to neonatal care units, commonly referred to as ‘admission hypothermia’, has been given an important focus in the literature.

Most studies in the literature about admission hypothermia have been conducted in the African region. A meta-analysis by Beletew et al⁵ showed a pooled prevalence of 57.2% of neonatal hypothermia among hospital-based studies in East Africa.⁵ Phoya et al¹² detected 77% of neonates in a tertiary hospital in Malawi to be hypothermic on admission to the neonatal unit.¹² Evidence on the burden of neonatal hypothermia in healthcare settings is scarce in South Asia as most studies in the region have targeted community settings.⁶ Among limited available studies, a study in Nepal showed a hypothermia prevalence of 64% among babies in a newborn unit.¹³ Only one study in Sri Lanka had assessed the prevalence of hypothermia among admitted neonates previously. Madhvi et al¹⁴ (2014) conducted a study involving two hospitals in Western Sri Lanka which reported a prevalence of 63% of admission hypothermia.¹⁴

There are neonatal, maternal, institutional and environmental risk factors for neonatal hypothermia. The systematic review by Beletew et al⁵ identified low birth weight, prematurity and medical problems as neonatal risk factors; caesarean section delivery, obstetric complications during pregnancy and labour were observed as maternal risk factors.⁵ Several studies in African countries reported early bathing, absence of immediate skin-to-skin contact with the mother after birth, delay in initiation of breastfeeding, night-time delivery and admissions during the cold season as common institutional and environmental risk factors for hypothermia on admission to neonatal intensive care units.23 15,21

Though hypothermia is rarely identified as a direct cause of death among newborns, it contributes significantly towards neonatal morbidity and mortality by acting as a comorbidity.⁶ It is associated with the most common causes of neonatal deaths worldwide: neonatal infections, prematurity and birth asphyxia.⁶ Neonatal hypothermia leads to increased energy consumption, hypoxia, reduced peripheral circulation causing tissue ischaemia, renal failure, necrotising enterocolitis, coagulation defects and sepsis.²² Nayeri et al²² reported a significant association between hypothermia at birth and respiratory distress in the first 6 hours of life, jaundice, hypoglycaemia and metabolic acidosis among newborns in Iran, regardless of birth weight and gestational age.²²

Considering mortality, a study conducted in a health facility in Nepal showed an 80% rise in mortality for every 1°C decrease in neonatal body temperature.²³ Johanson et al¹³ reported that 15.6% of neonates that were hypothermic on admission to a neonatal unit in Kathmandu, Nepal, died indicating a significant association between hypothermia on admission and mortality.¹³ Neonates with hypothermia had 2–3 times higher mortality rates than normothermic babies on admission to neonatal intensive care units.^{3 4}

Given the serious morbidity and mortality associated with neonatal hypothermia, generating epidemiological evidence on its prevalence, associated factors and consequences will be imperative in designing health interventions for early detection and prevention, especially in low-resource settings. As stated before, there is also a critical gap in knowledge concerning the burden and associations of admission hypothermia in Sri Lankan and South Asian healthcare settings. Our findings will thus support policy actions and advocacy for newborn care in the region. Therefore, the aims of this study were primarily to determine the prevalence and secondarily to determine the associated factors and outcomes of hypothermia among neonates admitted to the premature baby unit (PBU) of a secondary care hospital in Sri Lanka.

Methods

Study design and setting

This was a retrospective cross-sectional analytical study conducted at the PBU of the District General Hospital (DGH) Nawalapitiya. DGH Nawalapitiya is a secondary care hospital that is located in the Kandy district of Central Province, Sri Lanka. There are about 200 deliveries that take place at DGH Nawalapitiya per month and approximately 40 neonates are admitted to the PBU monthly. It must be noted that contrary to the name of the unit, its care is not limited to premature babies. PBU of DGH Nawalapitiya provides services to any neonate—either term or preterm—who requires specialised neonatal care. It consists of facilities standard to any neonatal care unit in Sri Lanka including incubators, warmers and ventilators. This study was conducted at the PBU from April 2022 to November 2023. Data collection was carried out retrospectively starting from 31 March 2022 to 1 April 2021.

Patient and public involvement

This study did not involve co-designing with patients or the public.

Study population and sampling

All neonates (infants who are aged 1–28 days) admitted to the PBU of DGH Nawalapitiya with a measurement of axillary temperature using a digital thermometer were included in the study. Re-admissions of the same neonate were excluded.

The minimum sample size was calculated using the formula to estimate a population proportion with specified absolute precision.²⁴ For a 95% confidence interval (CI) level, 5% precision and an expected proportion of 0.5 to estimate the largest sample size for an unknown proportion of hypothermia among neonates admitted to the PBU, the minimum sample size was computed as 384. This was inflated by 10% to account for possible missing information in bed head tickets (BHT) related to variables other than the recorded temperature, rendering a sample size of 422. A consecutive sampling technique was followed starting with the last patient admitted on March 2022 and proceeding retrospectively till the required sample size was achieved.

Study instruments and data collection

The data source was the BHT pertaining to the neonate admitted to the PBU. The BHT is a medico-legal document that is used for record-keeping of inward patients in Sri Lanka from the point of admission to a ward or any other care unit, filled by attending consultants, medical and nursing staff. A structured data collection sheet was used to collect data on patient demographic data, clinical information and measurements of axillary temperature available on the BHT. At the PBU of DGH Nawalapitiya, all axillary temperature measurements were recorded on admission by the admitting nursing officer using a digital thermometer. The nursing officers at the PBU have been trained to adhere to standard techniques as outlined by the National Guidelines for Newborn Care published by the Ministry of Health²⁵ (figure 1). However, no instrument-related standardisation techniques were used for this study.

Figure 1. Admission temperature recording.

A pretest was carried out using a few sample BHTs to ensure the comprehensibility of the structured data collection sheet and to identify other possible issues that may be encountered during the process of data extraction and entering. Neonates fulfilling the inclusion criteria were recruited, and the structured data collection sheet was used to extract data from their BHTs. Data collection was performed by a research assistant who was given prior training in secondary data collection using medical records. Extracted data was cross-examined by a co-investigator for accuracy and reporting errors. Ambiguity in data that arose due to the style and legibility of writing in BHTs was resolved through a consensus among co-investigators, or in severe cases, it was classified as missing data.

Variables

Admission hypothermia (outcome variable) was defined as an axillary temperature that is less than 36.5°C, measured on admission to the PBU using a digital thermometer.¹⁰ Operationalisation of all neonatal and maternal exposure variables including birth weight and prematurity categories followed standard definitions in paediatric practice and are defined in tables1 2.^{9 11} Additionally, infant respiratory distress syndrome (IRDS) also known as hyaline membrane disease was specified as breathing difficulty that occurs due to surfactant deficiency, commonly seen in preterm neonates, characterised by one or more signs including tachypnoea, chest recessions, grunting, nasal flaring, apnoea and gasping.²⁵ Neonatal hypoglycaemia was quantified as a blood sugar level <45 mg/dL.²⁶ Metabolic acidosis was quantified as a capillary or venous blood gas revealing a pH value <7.3 or a base deficit >−10.²⁷ Neonatal jaundice was identified as elevated total serum bilirubin values above the treatment thresholds in the gestation-specific graphs of the national guidelines, characterised clinically by yellowish discolouration of the skin, sclera and mucous membranes.²⁵ Pulmonary haemorrhage was defined as an acute, catastrophic discharge of blood from the upper respiratory tract or endotracheal tube as a result of fulminant lung oedema with leakage of red blood cells and capillary filtrate into the lungs.²⁸ Neonatal sepsis was operationalised as a suspected or proven systemic infection associated with systemic inflammatory response that occurs in newborns up to 28 days of age, including two or more of either increased or decreased total white cell count for age or >10% immature neutrophils, abnormal temperature of >38.5°C or <36°C, increased respiratory rate for age and increased heart rate for age. Both early and late sepsis were included in this category.²⁵ Corrected gestational age on admission was identified as the gestational maturity of the baby by the day of admission to the PBU. For example, if a neonate born at 37 weeks gestation was admitted to the PBU on day 2 of life, the corrected gestational age on admission would be 37 weeks and 2 days written as 37+2.^{11 25}

Table 1. Neonatal characteristics.

Variable	Categories	Frequency	Percentage
Sex	Male	214	52.6%
	Female	192	47.2%
	Ambiguous	1	0.2%
Age on admission	≤24 hours	235	57.7%
	>24 hours	172	42.3%
Gestational age	Post term (≥42 weeks)	0	0%
	Term (37–41+6 weeks)	240	59.0%
	Late preterm (34–36+6)	97	23.8%
	Moderate preterm (32–33+6)	27	6.6%
	Very preterm (28–31+6)	37	9.1%
	Extremely preterm (<28 weeks)	6	1.5%
Corrected gestational age on admission	≥42 weeks	0	0%
	37–41+6 weeks	259	63.6%
	34–36+6	82	20.2%
	32–33+6	29	7.1%
	28–31+6	32	7.9%
	<28 weeks	5	1.2%
Birth weight	Overweight (>3.5 kg)	18	4.4%
	Normal birth weight (3.5–2.5 kg)	213	52.5%
	Low birth weight (2.499–1.5 kg)	119	29.3%
	Very low birth weight (1.499–1 kg)	43	10.6%
	Extremely low birth weight (<1 kg)	13	3.2%
Weight on admission	>3.5 kg	21	5.2%
	3.5–2.5 kg	185	45.6%
	2.499–1.5 kg	144	35.5%
	1.499–1 kg	44	10.8%
	<1 kg	12	3.0%
Resuscitated at birth	Yes	59	14.5%
	No	347	85.5%
Admitted from	Operation theatre	130	31.9%
	Outpatient department	98	24.1%
	Postnatal ward	97	23.8%
	Labour room	66	16.2%
	Transferred from other hospitals	10	2.5%
	Emergency treatment unit	4	1.0%
	COVID isolation room	2	0.5%
Infant respiratory distress syndrome	No IRDS	321	79.3%
	Mild IRDS	72	17.8%
	Moderate IRDS	6	1.5%
	Severe IRDS	6	1.5%
Hypoglycaemia	No hypoglycaemia	366	90.4%
	Hypoglycaemia	39	9.6%
Metabolic acidosis	No metabolic acidosis	370	91.4%
	Metabolic acidosis	35	8.6%
Neonatal jaundice	No neonatal jaundice	260	64.2%
	Neonatal jaundice	145	35.8%
Pulmonary haemorrhage	No pulmonary haemorrhage	403	99.8%
	Pulmonary haemorrhage	1	0.2%
Sepsis	No sepsis	353	87.2%
	Sepsis	39	9.6%
	Meningitis	12	3.2%
Ventilatory support	No ventilatory support required	331	82.0%
	Continuous positive airway pressure	43	10.6%
	Mechanical ventilation	30	7.4%
Outcome	Alive and discharged	383	94.3%
	Deceased	20	4.9%
	Alive and transferred	3	0.8%

The total number in each category may vary due to missing data.

IRDSInfant respiratory distress syndrome

Table 2. Maternal characteristics.

Variable	Categories	Frequency	Percentage
Maternal age	<20 years	13	3.2%
	20–35 years	346	85.4%
	>35 years	46	11.4%
Parity	1	182	44.7%
	2	116	28.5%
	≥3	109	26.8%
Type of pregnancy	Singleton	366	90.1%
	Twin	35	8.6%
	Triplet	5	1.3%
Diabetes	No diabetes	345	85.2%
	Gestational diabetes mellitus	43	10.6%
	Type 2 diabetes mellitus	17	4.2%
Hypertension	No hypertension	342	84.5%
	Pregnancy-induced hypertension	52	12.8%
	Pre-eclampsia	6	1.5%
	Chronic hypertension	4	1.0%
	Young hypertension	1	0.2%
Infections	Yes	24	5.9%
	No	380	94.1%
Premature rupture of membranes	No PROM	353	87.4%
	<18 hours	24	5.9%
	18–24 hours	8	2.0%
	>24 hours	19	4.7%
Mode of delivery	Normal vaginal delivery	152	37.4%
	Emergency lower segment caesarean section	143	35.2%
	Elective lower segment caesarean section	93	23.0%
	Vacuum-assisted delivery	18	4.4%

The total number in each category may vary due to missing data.

PROMPremature rupture of membranes

Data analysis

Data were entered into Microsoft Office Excel and were checked for duplicates and missing records. A complete case analysis was carried out. After the coding of data, the dataset was added to Statistical package for the social sciences (SPSS) V.25.0 for analysis.

Descriptive statistics were reported using both numerical and graphical methods. Bivariate analysis of categorical variables was performed using χ² tests. Categorical variables with multiple groups were recoded into dichotomous variables when required. Relevant crude odds ratios (ORs) were calculated with their 95% CI. A probability value (p value) of <0.05 was considered to be statistically significant.

Multivariable logistic regression was carried out in line with the standards developed by Bagley et al²⁹ for using and reporting logistic regression in the medical literature.²⁹ Only associations that were found to be significant in bivariate analysis were included in the regression model. Each predictor variable was coded in binary groups similar to the bivariate analysis. The number of events per predictor variable was assessed. Multicollinearity between the predictors was evaluated by running collinearity diagnostics using multiple linear regression with the outcome of hypothermia taken as a continuous variable. Variance inflation factor (VIF) over 5 was considered the cut-off. Since we aimed to develop an explanatory model for neonatal hypothermia in hospital settings rather than a predictive model, all the eligible predictors were fitted simultaneously into the model, without following automated stepwise procedures. The overall model significance and the goodness of fit were determined using the Omnibus test and Hosmer and Lemeshow test, respectively. Classification accuracy of the model and pseudo R² values were reported. Adjusted ORs (aORs) were calculated with their 95% CI. Testing for interactions between the predictors was not performed. The STROBE cross-sectional checklist was used when writing our report.³⁰

Ethics

Ethical clearance was obtained from the Ethical Review Committee of the National Hospital Kandy (ERC No. NHK/ERC/19/2022). Necessary administrative permissions were taken from DGH Nawalapitiya.

Results

Study participants

Out of 434 records, 27 were excluded due to missing data (6.2%) and 407 neonates were included in the final analysis (figure 2).

Figure 2. Study participants.

Baseline maternal and neonatal characteristics

Characteristics of the neonates

Table 1 shows the characteristics of the study population. The majority (52.6%) were males. One baby was categorised as having ambiguous genitalia. Most neonates (57.7%) were admitted within the first 24 hours of life. The median admission age was 1 day (IQR: 4). The majority (59%) were term babies. Among the preterm, more than half (58%) were late preterm. 38% of the neonates had a corrected gestational age that was less than 37 on admission. Most neonates (52.5%) had a normal birth weight. The mean birth weight was 2.46 kg (SD: 0.76). 49.3% of the neonates weighed less than 2.5 kg on admission. One out of seven of the babies (14.5%) were resuscitated at birth. The most common places from where the neonates were admitted were the operation theatre (31.9%), the postnatal ward (23.8%) and the outpatient department (24.1%).

The main morbidities experienced by neonates during PBU stay are also summarised in table 1. 18% of the sample required ventilatory support. There were 20 deaths (4.9%) reported during the period of admission.

Characteristics of the mothers

The mean maternal age was 29 years (SD: 5.3). 3.2% were teenage mothers (less than 20 years). The median parity of mothers was two pregnancies (IQR: 2). Over three-fourths (90.1%) had singleton pregnancies. Nearly half (44.7%) were primi mothers. Table 2 summarises the main morbidities experienced by mothers during their pregnancies. The majority (58.2%) were delivered via lower-segment caesarean section (LSCS).

Prevalence of neonatal admission hypothermia

The prevalence of hypothermia among neonates on admission to the PBU was 38.6% (95% CI:33.9 to 43.4) (n=157). As seen in figure 3, of the 157 neonates that were found to be hypothermic, the majority had moderate hypothermia (n=101). There were no severely hypothermic babies.

Figure 3. Grades of hypothermia among neonates admitted to the PBU.

Associations of admission hypothermia

Neonatal and maternal factors

As shown in table 3, in the bivariate analysis, neonatal factors that were significantly associated with admission hypothermia were admission age being less than 24 hours, prematurity, corrected gestational age on admission being less than 37 weeks, low birth weight, admission weight being less than 2.5 kg and being resuscitated at birth. The largest measure of association with admission hypothermia was observed among neonates whose corrected gestational age was less than 37 weeks (OR: 8.7, 95% CI: 5.5 to 13.7). Being a female baby was identified as a protective factor in the sample (OR: 0.8, 95% CI: 0.8 to 1.3) although it was not statistically significant.

Table 3. Bivariate and multivariable analysis of associated factors of admission hypothermia.

Variable	Hypothermic(157)	Non hypothermic(250)	OR (95% CI)	P value	aOR(95% CI)	P value
	n (%)	n (%)
Neonatal factors
Sex of neonate
Female	77 (40.1%)	115 (59.9%)	0.8(0.6 to 1.3)	0.600
Male	80 (37.6%)	134 (62.4%)
Age on admission (hours)
≤24	129 (54.9%)	106 (45.1%)	6.3(3.8 to 10.1)	<0.001	3.3(1.9 to 5.8)	<0.001
>24	28 (16.3%)	144 (83.7%)
Gestational age (weeks)
<37	106 (63.5%)	61 (36.5%)	6.4(4.1 to 10.0)	<0.001	1.5(0.8 to 2.9)	0.234
≥37	51 (21.2%)	189 (78.8%)
Corrected gestational age on admission (weeks)
<37	103 (69.6%)	45 (30.4%)	8.7	<0.001
≥37	54 (20.8%)	205 (79.2%)	(5.5 to 13.7)
Birth weight (kg)
<2.5	110 (62.9%)	65 (37.1%)	6.6(4.3 to 10.3)	<0.001	3.2(0.9 to 11.0)	0.058
≥2.5	47 (20.3%)	184 (79.7%)
Weight on admission (kg)
<2.5	112 (56%)	88 (44%)	4.5(2.9 to 7.0)	<0.001	0.8(0.2 to 2.6)	0.724
≥2.5	45 (21.8%)	161 (78.2%)
Resuscitated at birth
Yes	37 (62.7%)	22 (37.3%)	3.2(1.8 to 5.6)	<0.001	1.8(0.9 to 3.4)	0.104
No	120 (34.6%)	227 (65.4%)
Maternal factors
Teenage pregnancy
Yes	10 (76.9%)	3 (23.1%)	5.5(1.5 to 20.5)	0.004	8.2(1.8 to 37.2)	0.006
No	147 (37.5%)	245 (62.5%)
Parity
Primi	74 (40.7%)	108 (59.3%)	1.2(0.7 to 1.7)	0.437
Not primi	83 (36.9%)	142 (63.1%)
Multiple pregnancy
Yes	31 (77.5%)	9 (22.5%)	6.5(3.0 to 14.2)	<0.001	2.8(1.1 to 7.1)	0.025
No	126 (34.4%)	240 (65.6%)
Diabetes in pregnancy
Yes	23 (38.3%)	37 (61.7%)	0.9(0.5 to 1.7)	0.941
No	134 (38.8%)	211 (61.2%)
Hypertension in pregnancy
Yes	40 (63.5%)	23 (36.5%)	3.3(1.9 to 5.8)	<0.001	2.3(1.2 to 4.7)	0.016
No	117 (34.2%)	225 (65.8%)
Maternal infections
Yes	11 (45.8%)	13 (54.2%)	1.36 (0.59 to 3.1)	0.47
No	146 (38.4%)	234 (61.6%)
Premature rupture of membranes
Yes	32 (62.7%)	19 (37.3%)	3.0(1.6 to 5.6)	<0.001	1.5(0.7 to 3.4)	0.288
No	125 (35.4%)	228 (64.6%)
Mode of delivery
LSCS	101 (42.8%)	135 (57.2%)	1.5(1.0 to 2.3)	0.044	1.1(0.6 to 1.8)	0.796
Not LSCS	56 (32.9%)	114 (67.1%)

LSCSLower-segment caesarean section

Considering maternal factors, teenage pregnancy, multiple pregnancy, hypertension during pregnancy, premature rupture of membranes and undergoing LSCS were found to have statistically significant associations with neonatal hypothermia in the bivariate analysis as seen in table 3. Associations with being a primi mother, having diabetes or infections during pregnancy were not statistically significant.

Environmental factors

The month of admission was the only external environmental factor that was assessed in this study. The highest mean temperature on admission was observed during April (36.57°C) and the lowest mean temperature was observed in January (36.07°C). Figure 4 illustrates the monthly changes in the percentage of neonates with hypothermia. Admission hypothermia was highest during January when over half of the neonates (52.5%) were found to be hypothermic. It was the lowest during April (25.7%). Bivariate analysis showed no significant association with the month of admission (χ²: 14.3, p value: 0.215).

Figure 4. Monthly change in the percentage of neonates with admission hypothermia.

Logistic regression analysis

In the multivariable logistic regression model, the number of events per predictor variable was equal to the recommended 10:1 ratio. In collinearity diagnostics, prematurity and corrected gestational age on admission variables were seen to have VIF over 5. Therefore, out of the two, only prematurity was selected for the final model. The model showed satisfactory statistical significance (Omnibus test, χ²: 140.1, p value: <0.001) and goodness of fit (Hosmer and Lemeshow test, χ²: 13.64, p value: 0.09). It had an overall classification accuracy of 76.5%, and Nagelkerke pseudo R was 39.8%. In the multivariable analysis, as shown in table 3, only admission age being less than 24 hours remained statistically significant out of the neonatal factors (aOR: 3.3, 95% CI: 1.9 to 5.8). Among maternal factors, only teenage pregnancy, multiple pregnancy and hypertension in pregnancy maintained statistically significant associations with neonatal hypothermia in the final model with all the predictors. The maternal factor with the largest measure of effect with hypothermia was teenage pregnancy (aOR: 8.2, 95% CI: 1.8 to 37.2).

Morbidity and mortality associated with neonatal admission hypothermia

On analysis of the complications associated with admission hypothermia shown in table 4, infant respiratory distress syndrome, metabolic acidosis, neonatal jaundice and requiring ventilatory support were found to be statistically significant. Neonates with hypothermia had 4.9 times (95% CI: 2.8 to 8.5) odds of receiving ventilatory support and 5.2 times (95% CI: 1.8 to 14.6) odds of mortality compared with normothermic neonates.

Table 4. Analysis of neonatal complications associated with admission hypothermia.

Variable	Hypothermic	Nonhypothermic	OR (95% CI)	P value
	n (%)	n (%)
Infant respiratory distress syndrome
Yes	54 (34.4%)	30 (12.1%)	3.8 (2.3 to 6.3)	<0.001
No	103 (65.6%)	218 (87.9%)
Hypoglycaemia
Yes	19 (12.2%)	20 (8%)	1.5 (0.8 to 3.1)	0.169
No	137 (87.8%)	229 (92%)
Metabolic acidosis
Yes	20 (12.8%)	15 (6%)	2.3 (1.1 to 4.6)	0.018
No	136 (87.2%)	234 (94%)
Neonatal jaundice
Yes	77 (49.4%)	68 (27.3%)	2.6 (1.7 to 3.9)	<0.001
No	79 (50.6%)	181 (72.7%)
Pulmonary haemorrhage
Yes	1 (0.6%)	0 (0%)	1.0 (0.9 to 1.0)	0.204
No	154 (99.4%)	249 (100%)
Neonatal sepsis
Yes	20 (12.8%)	31 (12.5%)	1.0 (0.5 to 1.8)	0.925
No	136 (87.2%)	217 (87.5%)
Ventilatory support
Yes	51 (32.5%)	22 (8.9%)	4.9 (2.8 to 8.5)	<0.001
No	106 (67.5%)	225 (91.1%)
Outcome
Deceased	15 (9.6%)	5 (2%)	5.2 (1.8 to 14.6)	0.001
Not deceased	141 (90.4%)	245 (98%)

Discussion

This study assessed the prevalence, factors and outcomes associated with neonatal hypothermia in a secondary care setting in Sri Lanka. Our findings showed that nearly two out of every five neonates had admission hypothermia. In the multiple logistic regression model, admission age being less than 24 hours was the only neonatal factor associated with hypothermia; teenage pregnancy, multiple pregnancy and hypertension in pregnancy were the identified maternal factors associated. Infant respiratory distress syndrome, metabolic acidosis, neonatal jaundice and the requirement of ventilatory support were the complications associated with neonatal hypothermia. Hypothermic neonates were five times more likely to die compared with normothermic neonates.

The prevalence of neonatal hypothermia seen in our study was lower than the prevalence seen in healthcare settings around the world. It was lower than the pooled prevalence of hypothermia (57.2%) as reported by a hospital-based systematic review done in Africa and also the 64% prevalence observed in a healthcare facility in Nepal, where the rectal temperature was measured on admission.^{5 13} Our findings were also lower than previously observed prevalence rates in Sri Lanka (63%) as reported by Madhvi et al.¹⁴ This disparity in prevalence could be due to varying availability of resources in individual healthcare facilities and geographical differences. East African studies have identified possible causes as a lack of appropriate towels and other laundry items for the thermal control of neonates. The lower-than-expected rates of hypothermia in our study may be attributed to protective factors identified in the literature such as the availability of warm towels, plastic bags for delivery of preterm neonates, radiant warmers in the delivery rooms and transport incubators at PBUs in the Sri Lankan setting.^{3 13 31} Measurement of rectal temperature in the previous Sri Lankan study could have also given falsely high values compared with the axillary temperature measurement that is recommended by guidelines.³² Nevertheless, more investigations are needed into regional differences in neonatal hypothermia in Sri Lanka.

The only statistically significant neonatal factor identified was the admission age being less than 24 hours, which was similar to the findings of several studies done in hospitals of East and West Africa which showed that such neonates had up to 3.6 times higher odds of developing admission hypothermia.2,4 This could be due to poorly developed thermoregulatory mechanisms in neonates which is even more pronounced in the early hours of life and lack of adipose tissue, thus putting newborns at a higher risk of hypothermia. In the analysis of maternal factors, our findings showed that teenage pregnancy was significantly associated with neonatal hypothermia. A study done in a health facility in Uganda also found that newborns of adolescent mothers were at a significantly higher risk of hypothermia while another study done in a hospital in Nigeria found that maternal age being more than 20 years was a protective factor against neonatal hypothermia on admission.^{3 21} This may be due to vulnerabilities in the bio-psycho-social environment surrounding teenage pregnancies that challenge the effective provision of warmth to neonates. The second maternal factor found to be significant was multiple pregnancy. This was consistent with a study done in a healthcare setting in Iran which showed 1.6 times the odds of neonatal hypothermia in multiple pregnancy compared with singletons.³³ This could be due to inadequate facilities and health staff in the delivery rooms to effectively maintain the warm chain for multiple babies delivered at once. The third significant maternal factor was hypertension during the current pregnancy. A multivariable analytical study done in a hospital in Canada and a study done in a neonatal intensive care unit in Brazil found that neonates of mothers with hypertension during pregnancy and pre-eclampsia, which is an entity of hypertension in pregnancy, had a significantly higher chance of developing hypothermia.^{34 35} Three similar hospital-based studies done in Ethiopia have found maternal obstetric complications to be significantly associated with neonatal hypothermia on admission.^{15 16 36} This could be attributed to the mothers being too ill to adequately care for a newborn and to maintain close contact in providing kangaroo care after birth. Though low birth weight, prematurity, being resuscitated at birth and delivery via LSCS were found to be significantly associated with neonatal hypothermia on bivariate analysis, they were not significant in multivariable analysis. This contradicts the findings of several previous studies done in healthcare facilities in African, American and Middle Eastern countries where these risk factors were found to be significant.2,516 The discrepancy in these findings is likely due to the heterogeneity of the study population, study methods and settings.

Considering environmental risk factors, the monthly changes in the percentage of neonates with hypothermia tallied with the temperature changes experienced in the Kandy district. Admission hypothermia was highest during the colder climate from November to January, and it was lowest during the hotter climate from March to April.³⁸ Previous studies in hospitals in Ethiopia and Nigeria have found 1.7 and 1.8 times odds of developing neonatal hypothermia in admissions during the cold season, respectively.^{3 15} However, the association of neonatal hypothermia with the month of delivery was not significant in our study which could have been likely due to a Type II error in research. Meanwhile, institutional factors could not be investigated in our study due to the retrospective collection of secondary data. Among them, early bathing is not practiced in the Sri Lankan set-up, either in the community or in hospitals. Sri Lanka has also successfully implemented the Baby Friendly Hospital Initiative since 1992.³⁹ Nevertheless, more updated evidence is required on the status of skin-to-skin contact and early initiation of breastfeeding in Sri Lankan hospitals and their association with neonatal hypothermia.

Our findings showed that neonates with admission hypothermia are at a significantly higher risk for multiple complications. These findings are compatible with a study done in a health facility in Iran which found a significant rise in the development of respiratory distress, neonatal jaundice and metabolic acidosis in hypothermic neonates.²² A study done in a hospital in Malawi and a meta-analysis done in East Africa also found infant respiratory distress syndrome and neonatal health problems to be significantly associated with hypothermia, respectively.^{5 12} In persistent hypothermia, high oxygen consumption by brown adipose tissue for non-shivering thermogenesis results in lactic acid production. This lactic acid is poorly metabolised by the immature liver which results in metabolic acidosis. Neonates are already at risk of developing jaundice due to poorly developed metabolic functions. Hypothermic neonates are at an even greater risk for jaundice due to a hypothermia-induced reduction in metabolic rate which hinders bilirubin metabolism. Furthermore, hypothermia causes high oxygen consumption, reduced surfactant release and pulmonary vasoconstriction, leading to increased respiratory distress. This also results in a higher chance of requiring ventilatory support such as continuous positive airway pressure ventilation.⁹ In our study, neonates with hypothermia showed 5.2 times the odds of mortality compared with normothermic neonates. This finding is consistent with the regional literature. Two studies done in hospitals in Iran assessing hypothermic neonates at birth showed 3.1 and 3.6 times higher odds of mortality.^{22 33}

The strengths of our study included calculating the prevalence of admission hypothermia in a large sample of neonates using secondary data and analysing the associated factors and outcomes using a multivariable analysis, thereby minimising the confounding effects. There are several limitations of our study. First, since the study was based on secondary data, its internal validity may be subjected to the errors in medical record-keeping process. However, we attempted to minimise that information bias by using a single data extractor, using explicit operationalisation and standard review methods. Second, we were unable to verify whether the nursing officers adhered to the same techniques in measuring temperature or the inter-instrument reliability of the digital thermometers. Third, during multivariable analysis, our model did not account for possible interactions between the predictor variables. Fourth, the cross-sectional nature of this study can only establish associations and not causal risk factors of neonatal hypothermia. Fifth, since this study was carried out in a single centre in the Central Province of Sri Lanka, the generalisability of the prevalence findings to other regions may be limited.

This study generates recommendations for both practice and future research. For practitioners, strengthening antenatal counselling on neonatal hypothermia and its adverse effects for expectant parents with a special focus on teenage mothers, mothers with multiple pregnancies and mothers with hypertension in pregnancy is recommended. Special attention to maintaining the warm chain among day 1 neonates that are being admitted is proposed as standard operating procedures for neonatal intensive care units. Raising awareness among healthcare staff on the serious morbidity and mortality of neonatal hypothermia highlighted by these findings can be recommended to promote thermoprotective interventions. Since we have linked hypothermia with both specific associated factors and adverse outcomes, future research could focus on studying the mediation effect of hypothermia in manifesting the said outcomes from possible risk factors. Since this study failed to ascertain the relationship between specific institutional and environmental factors and admission hypothermia, future studies in the South Asian region could investigate those, preferably with prospective study designs.

Data availability statement

Data are available upon reasonable request.