Challenges in Classification and Assignment of Causes of Stillbirths in Low- and Lower Middle-Income Countries

Abstract

Stillbirths account for 2.6 million deaths annually. 98% occur in low- and lower middle-income countries. Accurate classification of stillbirths in low-resource settings is challenged by poor pregnancy dating and infrequent access to electronic heart rate monitoring for both the newborn and fetus. In these settings, liveborn infants may be misclassified as stillbirths, and stillbirths may be misclassified as miscarriages. Causation is available for only 3% of stillbirths globally due to the absence of registration systems. In low-resource settings where culture and autopsy are infrequently available, clinical course is used to assign cause of stillbirth. This method may miss rare or subtle causes, as well as those with non-specific clinical presentations. Verbal autopsy is another technique for assigning cause of stillbirth when objective medical data are limited. This method requires family engagement and physician attribution of cause. As interventions to reduce stillbirths in LMICs are increasingly implemented, attention to accurate classification and assignment of causes of stillbirth are critical to charting progress.

Background

Stillbirths account for 2.6 million deaths annually, making this a critical public health problem.¹ In 2015, 18.4 per 1000 births worldwide resulted in a third trimester stillbirth. 98% of these stillbirths occurred in low- and lower middle-income countries (LMICs).² For example, in southern Asia and sub-Saharan Africa, estimated stillbirth rates range from 21.3 to as high as 56.9 per 1000 births.³ Furthermore, these regions have shown the slowest declines in stillbirth rates over the past 15 years.⁴ By comparison, high-income countries (HICs) account for only a small fraction of the global burden of stillbirths, as evidenced by a low stillbirth rate of 3.4 per 1000 births.² These global and regional estimates underrepresent the burden of stillbirths as they quantify third trimester stillbirths only, which is the international unit of comparison recommended by the World Health Organization (WHO). In addition to these third trimester stillbirths, a similar number of stillbirths occur earlier in pregnancy from 20 to 28 weeks in HICs.^5–8 If we assume that a comparable ratio of early to third trimester stillbirths applies in LMICs, there may be double the rate of stillbirths occurring in LMICs each year.

Due to poor registration of stillbirths, these deaths have historically been an invisible problem. For example, in 2000 when the United Nations established the Millennium Development Goals defining social and economic targets for progress by 2015, stillbirths were not mentioned in this agenda.⁹ More recently, a stillbirth target of 12 per 1000 births in every country by 2030 is defined by the Every Newborn Action Plan, a World Health Assembly resolution that articulates specific milestones for the United Nations’ Sustainable Development Goals (SDGs).^10,11 In this era of SDGs, the burden of stillbirths is gaining global focus.¹²

While estimates of the incidence of stillbirths have raised awareness of the global burden of this problem, several limitations in our understanding of the outcome impact our ability to reduce the incidence. For example, in LMICs, liveborn infants may be misclassified as stillbirths, and stillbirths may be misclassified as miscarriages. In these settings, causes of stillbirth are poorly understood. Accurately classifying and assigning causes of stillbirth are vitally important to developing new interventions, charting progress, and fueling political willpower to reduce stillbirths. In this review, we address the challenges associated with accurate classification of stillbirths, as well as assignment of causes of stillbirth in LMICs.

Challenges in classification of stillbirths

The International Classification of Diseases 10^th revision (ICD-10) defines a stillbirth or fetal death as death prior to complete expulsion or extraction from its mother of a fetus that has reached a birth weight of 500g or gestational age of 22 weeks (Figure 1).¹³ Death is evidenced by the absence of signs of life at birth, including no heart rate or pulsation of the umbilical cord, no respirations, and no definite movement of voluntary muscles. Despite the gestational age parameter in this ICD-10 definition, many LMICs use 28 weeks as the lower gestational age limit for defining stillbirth; some HICs, including the United States, use 20 weeks. Given these practices, the WHO recommends distinguishing stillbirth as either early fetal death (500–999g or 22–27 weeks) or late fetal death (≥1000g or ≥28 weeks).¹⁴ In addition, stillbirths can be further classified based on their timing as antepartum (before labor) or intrapartum (during labor).

Fig. 1.

World Health Organization recommendations for classification of mortality of the fetus and newborn based on timing of demise.

Identifying stillbirths and classifying them as early or late and antepartum or intrapartum relies on the following key practices: techniques to identify signs of life at birth, including identification of a newborn’s heart rate; methods of pregnancy dating; and identification of fetal heart rate. Given poor pregnancy dating as well as infrequent access to electronic heart rate monitoring for both the newborn and fetus in LMICs, accurate identification and classification of stillbirths is challenging. Consider the cases below, which illustrate examples of misclassification of stillbirths in LMICs that occur as a result of these constraints.

Stillbirth or livebirth? The flaccid, term newborn at birth

A term newborn is delivered at a health center in the setting of maternal hemorrhage and probable placental abruption. The newborn is flaccid at delivery, with no evidence of respiratory effort. The provider cannot palpate a pulse through the umbilical cord, and thus concludes the newborn is stillborn.

Key Learning Point: All newborns who are still at birth are not stillborn. Palpation of the umbilical cord is an unreliable method to identify heart rate in a depressed newborn.

The distinction between a flaccid, still newborn without respiratory effort and a stillborn is subtle, and may come down to an accurate evaluation of heart rate. In many low-resource settings, providers commonly rely on auscultation of the heart or palpation of the umbilical cord to evaluate heart rate. These methods of heart rate detection are less sensitive than electronic heart rate monitoring, and may fail to detect a heart rate in a severely depressed newborn who is, in fact, alive.¹⁵ Early classification of this flaccid newborn as stillborn can be a self-fulfilling prophecy, and may lead to withholding resuscitation efforts under the presumption that the newborn is not alive. This misclassification of early neonatal death as stillbirth is evident from early studies of a newborn resuscitation program in LMICs. These studies demonstrated a reduction in intrapartum stillbirths.^16,17 Given the postnatal timing of the resuscitation intervention, the reduction in stillbirths is likely due to more accurate identification of liveborn infants. Neonatal mortality was either stable or decreased in these studies, suggesting that accurate identification of liveborn infants contributed to improved resuscitation. Additionally, low provider expectations of survival or quality of survival for flaccid, liveborn infants might prompt them not to resuscitate these newborns, and therefore contribute to misclassification of neonatal deaths as stillbirths.¹⁸

For settings in which tools to electronically measure newborn heart rate are not available, reporting perinatal mortality is an alternative. Perinatal mortality, defined as death occurring during the intrapartum or early neonatal period, is a frequently reported indicator that bypasses the distinction between stillbirth and livebirth. Another alternative is to measure the proportion of fetal and neonatal death occurring in-hospital, as represented by the intrapartum stillbirth and early neonatal death indicator.¹⁹ This indicator is accurately measured by determining fetal viability on admission using a fetal monitor. Both perinatal mortality and the intrapartum stillbirth and early neonatal death indicator can be helpful in evaluating the quality of intrapartum care at a health facility.

Miscarriage or early fetal death? The extremely premature fetus without signs of life at birth

An extremely premature fetus is delivered at a health center in a rural area with no respiratory effort, voluntary muscle movement or heart rate. The fetus has translucent skin and fused eyelids, and is of a size and maturity such that it would not typically be resuscitated in this environment. The provider classifies this fetus as a miscarriage.

Key Learning Point: This fetus has an exam that could be consistent with a fetus as old as 26 weeks gestation. Physical exam findings are an unreliable method to identify the gestational age of a newborn.

Considerable national and facility-level variation in the threshold of viability contributes to inconsistent classification of mortality as a miscarriage versus an early fetal death. ICD-10 defines a miscarriage as expulsion or extraction from the mother of a product of conception that is less than 500 grams or less than 22 weeks completed gestation (Figure 1). Despite this standard definition, providers may rely on the local threshold of viability to classify a fetus without signs of life as a miscarriage. For example, a fetus delivering between 22 0/7 weeks and 27 6/7 weeks would be considered pre-viable in many LMIC settings. In these settings, a fetus of this gestational age without signs of life at birth may be misclassified as a miscarriage rather than an early fetal death.

Given this variation in thresholds of viability, the WHO recommends inclusion of all fetuses weighing 500–999g in national statistics. These fetuses are to be classified as liveborn if they have signs of life at delivery; if signs of life are absent, they are to be classified as stillborn, specifically an early fetal death. Despite this recommendation, and given known variation in classification, the WHO recommends using late fetal death (i.e., stillbirth during the third trimester) as the international unit of comparison when considering stillbirths globally.

Antepartum or intrapartum stillbirth? The term fetus without signs of life at birth

A term newborn with shoulder dystocia is delivered at a referral hospital in an LMIC. The newborn is not breathing, does not have a heart rate and is not moving its extremities. The clinical exam is notable for skin breakdown and discoloration. The provider concludes that the newborn shows signs of maceration, and thus classifies the newborn as an antepartum stillbirth.

Key Learning Point: Skin breakdown and discoloration may occur due to reasons other than decomposition after fetal death—for example, trauma occurring during the delivery. Though commonly used in the absence of other information, maceration is not a very reliable indicator of the timing of stillbirth.

Defining the death-to-delivery interval is key to accurately identifying and counting the subset of stillbirths that are potentially preventable through intrapartum interventions. In the absence of fetal heart rate monitoring, the most frequent method to identify the death-to-delivery interval in LMICs is a postnatal clinical exam. Maceration on clinical exam, defined as skin and soft-tissue changes such as skin discoloration or darkening, redness, peeling and breakdown, suggests an antepartum event. In contrast, a fetus who died during the intrapartum period lacks such skin changes and is considered a fresh stillbirth. Despite the commonly used distinction of macerated and fresh stillbirths, these clinical features are subjective and may not be an accurate proxy for timing of death. For example, up to 18% of antepartum stillbirths may be defined as fresh, and up to 30% of intrapartum stillbirths may be defined as macerated.²⁰

The distinction between antepartum (macerated) and intrapartum (fresh) stillbirths is critical because of the relative frequency of the latter in LMICs compared to HICs, and the implications of timing on underlying cause. Intrapartum stillbirths account for 50–70% of stillbirths in LMICs, compared to only 10% of stillbirths in HICs.² In LMICs, intrapartum stillbirths are mostly term or late-term gestation, weigh more than 2000g, and are not anomalous.^19,21,22 In contrast, intrapartum stillbirths in HICs include a significant portion of fetuses who are pre-viable or peri-viable. The vast majority of intrapartum stillbirths in LMICs are caused by asphyxia from maternal or fetal conditions such as pre-eclampsia or eclampsia, prolonged and obstructed labor, multiple births, abnormal presentations, and cord accidents.^22,23 Stillbirths associated with these conditions are often preventable with improved intrapartum surveillance and access to cesarean section.

Adjuncts to clinical assessment of the stillborn infant are inconsistently available in low-resource settings, but can be useful in defining the death-to-delivery interval. Most commonly, tools to measure fetal heart rate when the mothers arrives at a birth facility can assist in determining whether the fetus was alive at presentation. For settings in which tools to measure fetal heart rate are not available, an alternative to defining the death-to-delivery interval is reporting the total number of stillbirths. Total stillbirths are defined by the combination of antepartum and intrapartum, or macerated and fresh, stillbirths. However, focusing on total stillbirths does not help us to understand the relative value of antepartum versus intrapartum obstetric or public health interventions.

Challenges in assignment of causes of stillbirth

Cause of stillbirth is available for only 3% of stillbirths globally because of an absence of reliable and comprehensive registration systems.²⁴ Furthermore, comparison and collation of available data on cause of stillbirth is difficult secondary to the numerous systems employed for determining cause. Recognizing the importance of consistent assignment of cause of death, the WHO recommends assignment of cause of stillbirth consistent with the ICD-10 system, known as the ICD-perinatal mortality (ICD-PM) system.²⁵ However, the lack of data for stillbirths has precluded its consistent use in LMICs.

More than 50 systems for assigning cause of stillbirth are described in the literature, ranging from clinically based systems to those that require extensive laboratory testing. Most require identification of fetal distress, birth weight, gestational age and anomalies, as well as an understanding of the timing of the death. Some systems assign a primary cause while others define underlying causes; some ascribe a degree of certainty about cause of death.²¹ Many also require resource-intensive diagnostics that make them more appropriate for HICs given limited access to these diagnostics in LMICs.^{19,21,26–28} Below, we explore the challenges of assigning cause of stillbirth in LMICs given the limited amount of readily available clinical and laboratory data in these settings.

Assigning cause of stillbirth by clinical course

Absent laboratory testing, histology and autopsy tools, clinical course is the predominant information available for assigning cause of stillbirth in LMIC health facilities. Few systems exist for assigning cause of stillbirth based on clinical course alone, and those that exist do not typically distinguish between stillbirth and neonatal death.^29–33 One recently published system proposed by the National Institute of Child Health and Human Development Global Network uses a hierarchical method to assign the cause of stillbirth with clinical data obtained by lay health providers.^27,34 This system first establishes whether the stillbirth was associated with maternal trauma (i.e., assault, suicide, or accident); if so, the cause of death is classified as trauma. If no trauma is identified and there is a visible congenital anomaly, this is attributed as the cause. In the setting of maternal or fetal infection, such as malaria, syphilis, or vaginal odor, with the absence of trauma or congenital anomaly, the stillbirth etiology is defined as infection. If none of these three causes are present, and any maternal or fetal condition associated with intrauterine asphyxia is implicated, asphyxia is determined as the cause of death. Finally, if all of the above four causes are excluded, and the infant is <32 weeks gestation and non-macerated, complications of prematurity is ascribed as the etiology of the stillbirth. Alternatively, for infants without clear evidence of trauma, anomaly, infection or asphyxia who are either <32 weeks gestation and macerated, or ≥ 32 weeks gestation, the cause of the stillbirth is assigned as unknown. This simplified system of assigning cause of stillbirth may miss rare or subtle causes.²⁷

Infection, likely a common cause of stillbirth in LMICs, illustrates the challenge of assigning causation based on clinical course alone. In HICs, infection is the etiology for an estimated 10–25% of stillbirths; in LMICs, this number may be 50% or more.^35–37 At least 40 organisms have been implicated as a cause of fetal death, from bacteria, viruses, parasites, and fungi, to vector- and animal-borne infections.³⁶ Malaria and syphilis are commonly associated with stillbirth in endemic areas.^21,38 Confirmation of infection as an etiology for stillbirth may be especially variable in LMICs because of a disparity in access to diagnostic tools. Overall, the contribution of infection to stillbirth is likely underestimated, not only in LMICs where culture, histology and autopsy are rarely available, but also in HICs where they are not always routinely performed.

Identification of infection as the cause of a stillbirth based on clinical course alone can be challenging because infection causes a variety of signs and symptoms during pregnancy and labor that are common to other pathologic conditions. Maternal infection can result in systemic illness with death due to maternal respiratory distress or fever, and the resulting cytokine and chemokine release. The fetus may die without pathogens transmitted to the placenta or fetus. Alternatively, the placenta may be directly infected without transmitting organisms to the fetus. In these circumstances, decreased blood flow to the fetus results in reduced transmission of oxygen and nutrients, and can result in asphyxia. When the fetus is actually infected, damage to vital organs, or the development of an anomaly from an infection very early in gestation, can result in death.³⁹ Finally, infection can stimulate preterm labor that carries a higher mortality rate, with complications of prematurity confounding the assignment of cause of death.^38,40

An adjunct to clinical course for attribution of cause of stillbirth that is less resource-intense than autopsy is minimally invasive tissue sampling (MITS). MITS is currently being explored in a prospective cohort study known as PURPOSe (Project to Understand and Research Preterm pregnancy Outcomes and Stillbirths in South Asia). MITS is a relatively new technique in which placental and fetal organ tissue samples are obtained by needle biopsy then evaluated by histologic and PCR testing. PURPOSe, conducted in India and Pakistan, will obtain detailed information on causes of stillbirth and preterm neonatal death in low-resource areas using culture, MITS and autopsy.⁴¹ Participating stillborn infants undergo a standard autopsy and MITS, along with gross examination of the placenta and collection of maternal and cord blood. PURPOSe will provide valuable information about causes of stillbirth, including the prevalence of specific infections in low-resource settings. Furthermore, it will add to the understanding of the relative value of MITS as a technique for attributing cause of stillbirth. If MITS continues to show promise, further research will be required to understand facilitators and barriers to its scale-up in LMICs. Although less invasive and resource-intensive than autopsy, conducting MITS requires rigorous training and on-going access to cultures or PCR testing which are not readily available in most LMIC settings.

Assigning cause of stillbirth by verbal autopsy

Many stillbirths in LMICs occur at home and thus remain unregistered.^42,43 For the purposes of pregnancy-related health care planning, accurate data regarding the burden, timing, and causes of these stillbirths is needed.⁴⁴ One approach to assigning cause of stillbirth for these home births, and for births in other settings where objective medical data are limited, is the WHO’s verbal autopsy tool, revised in 2016 to include modules for stillbirth and neonatal death.⁴⁵ Verbal autopsy is an indirect method to determine cause of death from information about signs, symptoms, and circumstances preceding death. In the form of interviews with caretakers (such as trained or untrained birth attendants, nurses or nurse midwives) or family members of the deceased (usually mother of the stillborn), verbal autopsy is conducted with open and close-ended questions.^46–48

The WHO verbal autopsy tool has been utilized as a method for assigning cause of stillbirth in many community-based settings in LMICs. For example, a recent prospective cohort study was published in the Lancet describing population-based causes of stillbirth using verbal autopsy in eight countries in south Asia and sub-Saharan Africa.⁴⁹ This study demonstrated that the major causes of antepartum stillbirth in these settings were hypertensive disorders of pregnancy, infection and antepartum hemorrhage. For intrapartum stillbirths, complications of labor and delivery accounted for more than half, with other main causes similar to those of antepartum stillbirths. In another community-based study in the Democratic Republic of Congo, Guatemala, Pakistan and Zambia, verbal autopsy followed by assignment of cause of death by local physicians identified infection as the most common cause of stillbirth.³⁹

Three hospital-based studies compared assignment of cause using the WHO verbal autopsy tool for stillbirth against hospital standard diagnoses that included chart review supplemented by any available laboratory tests and autopsy results.^44,50,51 These studies reported variable performance of verbal autopsy with sensitivities of 13–85% and specificities of 50–100% across a variety of causes of stillbirth including congenital malformations, pregnancy-induced hypertension, other maternal illnesses, hemorrhage and obstetric complications. Generalizability of these studies to community-based settings is limited by the use of obstetricians to analyze verbal autopsy results (two studies) as well as changes in risk, exposures and interventions from hospital to community settings. However, given the lack of standard diagnosis of cause of stillbirth with chart review, laboratory tests and autopsy in community-based settings in LMICs, validation studies of verbal autopsy in these settings have not been performed.

The results of the WHO verbal autopsy tool may vary based on such factors as the type of respondent and interviewer, the recall period and the training of the person interpreting the verbal autopsy information.⁵² For example, in cases where the interview is optimally performed with a family member of the deceased, there may be respondent reluctance to engage in lengthy conversation regarding the details surrounding the fetus’s death in close proximity to its occurrence. Non-physicians trained in analysis of verbal autopsy may assign differing causes of death than physicians trained in the same analysis; this is evidenced by a community-based study where non-physician assignment of cause of death demonstrated only 50% agreement with physician assignment.⁵³ Ultimately, the complexity of the verbal autopsy tool and its administration has limited its uptake in routine health programs. Further simplification of the tool is needed to facilitate its widespread implementation.

Conclusion

Stillbirth is a critical, global public health problem that disproportionately affects LMICs. However, there is reasonable expectation that this burden can be reduced to meet the SDG stillbirth goal in some countries. Globally, stillbirth rates in HICs have fallen nearly ten-fold over the past 70 years, likely due to improved prenatal care, including prevention, screening and treatment of maternal syphilis, Rh disease, diabetes, and preeclampsia.²¹ Some congenital anomaly-associated stillbirths have been decreased by folic acid food fortification. Early detection of anomalies through ultrasound and other types of screening has prompted some mothers to elect for pregnancy termination before 20 weeks in affected fetuses that would otherwise have had an increased risk of stillbirth.^54,55 Monitoring before labor for poor fetal growth and conditions related to fetal asphyxia has prompted earlier delivery of at risk infants. Improved surveillance during the intrapartum period, including fetal heart rate monitoring, has led to expedited delivery (often by cesarean section) for conditions such as preeclampsia, eclampsia, fetal distress, abruption, and obstructed labor.²² Implementation of some of these interventions in LMICs would not require additional costly resources and might significantly reduce stillbirth rates.

Accurate classification of stillbirth in LMICs remains challenging secondary to poor pregnancy dating as well as infrequent access to electronic heart rate monitoring for both the newborn and fetus. Given these limitations, evaluations of interventions to reduce the rate of stillbirth or the rate of early neonatal death should use perinatal mortality as the primary outcome unless the differentiation between stillbirth and livebirth can be made with certainty. All fetuses who are 500g or above at birth should be included in national statistics, including those who are classified as a miscarriage, stillbirth or early neonatal death. However, third trimester stillbirth rate remains the optimal unit for global comparisons given current variation in thresholds of viability. Health care systems in LMICs should invest in accessible and affordable methods to detect fetal heart rate so that the death-to-delivery interval can be more precisely assessed.

Accurate assignment of cause of stillbirth in LMICs remains challenging secondary to varied systems to assign cause with low stillbirth registration rates overall, as well as poor access to culture, histology and autopsy. Analysis of the clinical course preceding a stillbirth remains the primary method to estimate cause of death in these settings.⁴⁴ This approach can miss rare or subtle causes of stillbirth, as well as causes with non-specific clinical presentations such as infection. For stillbirths occurring in settings where objective medical data are limited, verbal autopsy is a method for ascertaining cause of death. However, this requires engagement of the family and a trained physician to analyze the data and ascribe a cause. As public health interventions to reduce stillbirths in LMICs are increasingly implemented, attention to the challenges of accurate classification and assignment of causes of stillbirth is critical to charting progress.