Overall, the prevalence of hypothermia in hospital-based studies ranges from 32 to 85%, with the exception of a low outlier (8% in Guinea Bissau ). This wide range might in part be attributable to the varying case definition of hypothermia across studies, ranging from 35.0°C to 36.5°C, and in part to the climatic environment and its seasonal variations discussed below.
We caution that the direct comparability of prevalence data from these hospital-based studies are limited by selection bias, as study populations often represent a high-risk patient cohort and might not be representative of the local population. Data comparability is further limited by the heterogeneity in case definition of hypothermia, which ranges from as low as 35.0°C to the current WHO standard of 36.5°C body temperature, measured with homogeneously temperature measurement methods across studies. Furthermore, prevalence data are confounded by various covariates inconsistently present across studies, such as environmental temperatures and seasonality, the newborns' maturity and age, or maternal factors.
Risk factors for hypothermia
Various studies identified several risk factors for newborn hypothermia, which we categorize as follows.
Termed 'contextual' by other authors ; initially, an infant's body temperature is associated with maternal temperatures . Several studies have confirmed the intuitive association with environmental temperatures and with the cold seasons. The Gadchiroli trial, with an overall hypothermia prevalence of 17%, showed variations from summer (14.8%) to winter (21.5%) [21, 38]. Other studies from Haryana in Northern India recorded an overall hypothermia prevalence of 11%, ranging from 3% in the summer to 19% in winter . In Uttar Pradesh, hypothermia was detected in 14% (n = 148) and was found to strongly correlate with environmental temperature . Another study from the same state found a higher rate of 45%, which likewise was correlated with environmental temperatures and varied considerably over the seasons, ranging from 70% during winter to 20% during summer . Studies from Nepal suggest that the higher prevalence of hypothermia in hospitals during winter months can successfully be addressed through staff training of early drying, wrapping, and breastfeeding [40, 41]. The Sarlahi trial found that while even in the hottest season of the year almost one-fifth of infants were hypothermic , the risk of moderate-to-severe hypothermia further increased by 41% for each 5°C decrease in ambient temperature .
While newborns of all gestational ages are at risk of losing body heat after birth, premature and small babies are particularly vulnerable due to their physiologic disadvantages. A newborn's thermal regulatory mechanisms are highly sophisticated, but particularly in babies born prematurely easily overwhelmed . Neonatal anatomic characteristics add to the metabolic burden of increased energy requirements: term babies have a 2.7 times greater body surface and preterm babies an up to 4.0 times greater surface per weight than adults.
Several conditions of immature thermal regulation, such as LBW, prematurity, intrauterine growth restriction, and asphyxia (with heat loss due to lack of oxygenation and, where attempted, during reanimation efforts) during birth are significantly associated with an abnormal low body temperature [31, 32, 40, 42, 43]. Hypoglycemia is an important contributor to hypothermia , and vice versa: it maintains a vicious circle, which leads to feeding weakness, weight loss and finally increased mortality, which first was shown in studies in the 1950s and 1960s . Breastfeeding therefore treats hypothermia not only through bonding with and warming through the mother, but also by replenishing a newborn's glucose levels.
Early bathing contributes significantly to heat loss and increases the incidence of hypothermia in cold climates  and even in a warm environment  and should be postponed until at least after the first 6 h of life, and possibly longer. It is, however, a widespread practice even in high-risk environments [48–51].
Massage and oil applications to clean the child early after birth continue to be a widespread tradition . Evidence for the influence of massage and oil application on hypothermia is contradictory. While suggesting protection from hypothermia  and against nosocomial infections in preterm very low birth weight infants , it has also been shown in other studies to have detrimental effects on the skin as a protective barrier  and to lead to heat loss .
An infant's low body temperature is also associated with having a young and inexperienced mother, coming from a family with low socioeconomic status , or being born to a mother who already had multiple births .
While some of these physiologic risk factors have been documented decades ago, awareness of the risks associated with hypothermia, as indicated in a multinational survey  and another one from India , indicating that healthcare professionals have limited knowledge of the diagnosis and management of newborn hypothermia. Facilities in resource-limited environments rarely have sufficient capacity to address thermal protection. In a recent study in Zambia, we found that health centers are not well prepared to provide thermal protection, with only very few equipped with heat control for the delivery room (7%) or a neonatal warmer (9%) .
Associations of hypothermia with newborn morbidity and mortality
Several studies investigated the association between neonatal hypothermia and associated mortality risks. In our review, case fatality rates (CFR) for newborn hypothermia globally range from 8.5% to 52% [21, 26, 34, 59, 60]. A study from India that included only hypothermic babies specifically investigated morbidities and mortalities and found CFRs that ranged from 39.3% for mild hypothermia to 80% for severe hypothermia. This study demonstrated a dramatic effect of comorbidities and confirmed that hypothermia has a much worse outcome when associated with other newborn problems. Fatality rates increased to 71.4% with hypoglycemia, 83.3% with hypoxia, and 90.9% with shock .
However, the above-cited studies do not sufficiently control for potential confounders of the effect of hypothermia on mortality and thus provide implausibly high CFRs in the context of high hypothermia prevalence particularly in community settings. Furthermore, these studies reflect the higher risk of hospital populations selected for these studies, and their CFRs are therefore not applicable to community settings. Yet, hypothermia has been shown to be associated with mortality in a community setting. A community-based study conducted in Sarlahi, Nepal found that mortality increased by approximately 80% for every degree Celsius decrease in first observed axillary temperature and that relative risk of death ranged from 2 to 30 times within the current WHO classification for moderate hypothermia, increasing with greater severity of hypothermia .
The evidence on the effect of thermal protection on morbidity and mortality is currently limited. Skin-to-skin care has been shown to substantially reduce neonatal mortality among preterm infants born in facilities , but its effectiveness for infants born at term and in communities is less clear. In Zambia, we recently showed that training traditional birth attendants in newborn care emphasizing simple thermal protection (wiping the newborn dry and wrapping the dried infant in a separate piece of cloth), along with resuscitation and early treatment of possible sepsis where indicated, reduced mortality rates at day 28 after birth by 45% .
Whether wrapping the newborn in plastic (polyethylene or vinyl) bags is sufficient for thermal protection in low birth weight and premature infants remains controversial. While one recent study failed to show an effect of wrapping on hypothermia rates , others found that this approach effectively raised body temperature [65, 66] and did so more quickly than radiant heaters [67, 68], however did not reduce mortality .
Rethinking and redefining neonatal hypothermia
Our and other reviews [36, 70] suggest that the burden of hypothermia is still highly prevalent including in tropical countries, practices contributing to heat loss in the newborn are still deeply rooted in many cultures and are difficult to change, technologies adapted to resource-poor environments are still at a developmental stage, simple thermoprotective interventions and behaviors are insufficiently practiced, providers and caretakers lack an understanding of the problem, and adverse health outcomes continue to take their toll in morbidity and mortality in newborns. There is thus a critical need for researchers and policy makers to take on the challenge of newborn hypothermia and address it within a larger framework of maternal and child health programs.
Qualitative studies from Africa and South Asia suggest that delivery and newborn care practices contributing to heat loss are still common globally. Various cultural and sometimes economic barriers often interfere with implementing simple steps to prevent hypothermia. Heating the birth place is costly for families in resource-poor countries , and drying and wrapping the baby is often not a priority when the mother needs attention after delivery . In Ghana, for example, the practice of bathing newborns immediately after delivery is sometimes rooted in concerns about 'ritual pollution'  or the belief of helping the baby sleep and feel clean, and reducing body odor in later life; attitudes that informants felt would be difficult to change  and which need to be taken into account when programming for behavior change.
Major gaps in our understanding of neonatal hypothermia
Since in most parts of the world temperature is not measured and recorded in most newborns immediately after birth, the epidemiological picture of hypothermia and its clinical consequences is yet incomplete. Hypothermia is believed to be a common problem not only in developing countries, but also in formerly socialist countries . There are few studies from current lower-middle-income countries.
To the best of our knowledge, no population risk attributable to hypothermia has been published yet. There is no consistent definition of normal newborn body temperature, and consequently data that would allow for pooled risk estimates for newborn hypothermia are still incomplete. Temperatures have been shown to vary widely in healthy newborns , and standard medical textbooks disagree on the lower normal limit, ranging from 35.5 to 36.5°C, as well as the normal upper level, citing values from 37.0 to 37.9°C .
Standard randomized controlled trials to define temperature thresholds associated with adverse health effects (morbidity as well as mortality) and to quantify the contribution of hypothermia to neonatal mortality as specific cause of death have not been conducted. In fact, those studies might be methodologically impossible to undertake, or at least ethically problematic, because the detection of hypothermia prompts therapeutic intervention and thus artificially reduces the associated mortality risk.
Given these limitations, standard measurements of body temperature could be included in newborn studies to complement available epidemiologic data. Newborn trials conducted for other reasons that include data on newborn body temperature could facilitate further investigation of the association of newborn hypothermia with morbidities and disease-specific mortalities. Axillary measurements with standard digital thermometers are inexpensive and can easily be incorporated into most newborn care clinical guidelines and study protocols.
Currently, the best available data come from trials with hypothermia-unrelated interventions providing temperature data, such as the Sarlahi trial in Nepal . Further studies on the mortality and morbidity risks posed by newborn hypothermia are warranted, particularly in sub-Saharan Africa, to refine the current WHO classification scheme for hypothermia that, as has been suggested, might have to revised into narrower categories to more appropriately reflect the overall mortality-hypothermia risk relationship [18, 70].
Further research is needed to understand the magnitude and perception of the problem as well as the feasibility and effectiveness of thermoprotective interventions for newborn morbidity and mortality. Methodologically sound hospital-based and community-based studies are required to understand the problem in sub-Saharan Africa specifically. These studies will ideally include potential confounders and mediating factors that have largely not been adequately addressed so far, such as maternal temperature, environmental conditions, and sociocultural contexts and their association with newborn hypothermia.
Various international neonatal advocacy alliances have included thermoprotection strategies in their newborn health programming, including the Healthy Newborn Partnership (Save the Children), Partnership for Safe Motherhood and Newborn Health (WHO), and the Child Survival Partnership (UNICEF). Nevertheless, hypothermia remains a major challenge for newborn survival. Globally, progress has been particularly slow in improving survival in infants less than 7 days old . This might be attributable to the delivery gap in developing countries for interventions addressing early causes of death such as preterm birth and asphyxia, both of which have worse outcomes in the presence of hypothermia.
More than a quarter of a century ago, a study from Senegal reported: 'Deaths [from hypothermia] seem easy to avoid. Purchasing blankets, putting the newborn babies with their mother and not in cots, prohibiting unnecessary washing of the babies, and supplying maternity hospitals with solar water heaters are all easy steps which could greatly reduce this problem. These improvements are gradually being introduced in the maternity hospital we surveyed though it is proving very difficult to persuade elderly auxiliary midwives not to remove the vernix caseosa by thorough washing, this being part of a strong tradition. Health workers in dispensaries should also be aware of the problem of neonatal hypothermia. We frequently see at our clinic mothers with young infants whose growth seems unsatisfactory. They usually complain about not having enough milk. If the infant is lightly clothed and his rectal temperature is well below normal, he may be expending too much energy on thermogenesis. Advising such mothers to use warmer clothing for their babies is usually enough to make them put on weight. Artificial feeding would be wholly inappropriate in such conditions and could even prove fatal' . As our review suggests, most of these statements and easy remedies still hold true today in most of the world.