Introduction

The infant mortality rate (IMR) is one of the most sensitive indicators of a country’s health system.1 Trends in all countries, including India, have shown a decline in IMR. IMR in India declined from 89 per 1000 live births (1990) to 28 per 1000 live births (2020). A similar reduction was observed in Odisha during this period (96 per 1000 live births to 36 per 1000 live births).2 The reduction in IMR showed a steeper decline in India than the global reduction in IMR (65–28 per 1000 live births in India vs 53–29 per 1000 live births worldwide).2 The latest data show that Odisha’s IMR is higher (36/1000 live births) than the national level; Odisha’s IMR is the fifth highest among all states.2 India has to achieve a neonatal mortality rate of 12 per 1000 live births by 2030.

Studies have identified that inadequate coverage of institutional delivery, inadequate immunisation coverage, absence of care by trained healthcare workers, improper aseptic practices during delivery, inadequate postnatal care, etc, are the determinants of infant mortality.3–5 These determinants are addressed through various interventions in different health programmes in the maternal and child healthcare domains. Those interventions can be more effective after identifying specific causes and determinants of infant mortality.6 With an example of birth asphyxia, the interventions should be prioritised in the domains of care during delivery and immediate newborn care. Similarly, the type of healthcare received and the delay in receiving care are highly linked to infant mortality due to pneumonia. Thus, interventions such as home-based newborn care (HBNC) should focus more on early identification of danger signs.7 8

According to India's Civil Registration System (CRS), over 90% of deaths were reported; however, the Medical Certification of Cause of Death (MCCD) attributed causes to only 22.5% of these deaths at the national level.2 The verbal autopsy system was incorporated into the Sample Registration System (SRS) in 1998 to resolve this vast gap in reporting the cause of death.9 Identifying medical causes of death alone is insufficient to reduce mortality, as social factors and care-seeking patterns also play a significant role. 10The social factors lie distally in the web of causation, with medical causes in a proximal position with better visibility. Medical causes, such as neonatal tetanus, have different social factors, such as availing the services of trained birth attendants, selection of place of delivery, or number of antenatal care check-ups. Social autopsy tools can identify the pathway of seeking care. A social autopsy is an interview procedure designed to pinpoint the social, behavioural and health system factors that contribute to the deaths, which is based on the three-delay model. Level 1 delay is a delay in decision-making for seeking care, level 2 delay is a delay in reaching an appropriate healthcare facility, and level 3 delay is a delay in receiving adequate care after reaching an appropriate healthcare facility. The identification of determinants should be coupled with a verbal and social autopsy to find effective measures to reduce infant mortality. These will also help reallocate the resources to increase the effectiveness of the interventions. This study aimed to determine the medical causes of infant mortality by verbal autopsy and its determinants in two rural blocks of the Khordha district of Odisha. This study also assessed the pathway of care and delay in seeking care for the illness preceding infant death using the three-delay model.

Materials and methodsEthics statement

Written informed consent was obtained from all the study participants before data collection.

Study setting and population

This community-based matched case-control study was conducted in two rural blocks situated in the Khordha district of Odisha, an eastern state in India (online supplemental figures 1–4). The rural blocks are usually public health administration units at the subdistrict level, which are comprised of five to six primary health centres (PHCs).11 There were 277 000 people residing in those two rural blocks; two blocks were divided into eight sectors. Two community health centres (CHCs) are situated at the block level. There is one PHC situated in each sector. The PHCs provide services related to antenatal care check-ups along with normal vaginal delivery with essential newborn care. The CHCs provided services for lower segment caesarean section along with service provision in PHCs. Accredited Social Health Activists (ASHAs) work at the village level to mobilise the beneficiaries for maternal and child healthcare services. Auxiliary Nurse Midwives (ANMs) work at the subhealth centre level (below PHCs) to provide services such as pregnancy registration, immunisation and antenatal care check-ups. Both ANMs and ASHAs play a crucial role in the routine system of infant death reporting through timely notification.

We enumerated infant deaths between 1 August 2021 and 30 March 2023. We contacted all ASHAs and ANMs in the study area every month to enumerate infant deaths. We verified these data with the data available in the health management information system at the block level and death registers of nearby hospitals. The infant deaths that happened during the period mentioned above with their parents usually residing in the study area were included in the study as cases. The controls were selected after matching with cases concerning gender, date of birth (DOB) and residence. Alive infants of the same sex residing in the same village as of the case were recruited after confirming a DOB similar to the cases (for neonatal deaths, DOB within 7 days of the DOB of issues, for postneonatal deaths within 1 month of DOB of the cases). The eligible control, if unavailable in the same village, was taken from the nearest village. In the case of multiple eligible controls in a single village, one control was recruited based on the lowest interval of DOB.

Sample size

The required sample size was calculated as 110 each in the case and control groups to achieve a minimum power of 80% with an alpha error of 5%. We have taken an anticipated OR of 2.5 for low birth weight for infant mortality.12

Data collection tool

We collected the data from the primary caregivers of cases using a semistructured interview schedule developed based on a SRS verbal autopsy questionnaire, the International Network for the Demographic Evaluation of Population and Their Health (INDEPTH) and household questionnaire of National Family Health Services (NFHS)-5.2 13 14 The data from the controls were collected using the semistructured interview schedule developed from the available literature and household questionnaire of NFHS-5. The interview schedules were divided into four parts: sociodemographic and socioeconomic variables, antenatal factors, intranatal factors and postnatal factors. We collected the necessary information, as mentioned in the SRS verbal autopsy tool, for assessing the cause of death (online supplemental table 1). The cause of death of infants was identified by two blinded trained physicians and one senior clinician adjudicated in case of any disagreement. This method of verbal autopsy involving physicians is already well established and being used in the cause of death estimation at the national level in India.9 This method is more commonly used in lower-middle-income countries like India than newer verbal autopsy methods based on algorithms such as InterVA.15 16 The usage of comparatively lower technologies coupled with an online coding platform made the physician-coded verbal autopsy method an integral part of SRS in India. Though newer tools are coming up, such as InterVA-5, they require validation in our study setting before further usage.17 18 Overall, no single coding method for verbal autopsies outperforms others across different studies and metrics. Currently, there is little evidence to support replacing physician-coded verbal autopsy with an algorithm-based method, especially since physician diagnosis continues to be the global standard for clinical diagnosis in living patients.19

We collected data on the availability of different types of care, including government and private formal healthcare facilities, along with informal healthcare, using the INDEPTH social autopsy tool. The INDEPTH tool also helped to assess the time spent deciding to seek care, the time spent reaching the appropriate healthcare facility and the time spent receiving healthcare services in the healthcare facility.13 20 The social autopsy thus conducted using the INDEPTH tool provided the information related to the delay and pathway of care of the infants’ deaths. The level 1 delay was identified as a delay in decision-making for seeking care, the level 2 delay was the delay in reaching an appropriate facility, and the level 3 delay was a delay in receiving adequate care after reaching a healthcare facility or timely referral whenever required.13 The English version of the study tools is validated in India in a similar sociocultural setting. We translated the tools into Odia by one native Odia speaker and then back-translated them into the English language by another person. We conducted personal interviews with primary caregivers within a period of 15–90 days after the death. The data were collected by one of the authors, who was trained in verbal autopsy and was familiar with the local language (Odia).

Statistical analysis

We reported the causes of death separately during the neonatal period, postneonatal period and infant period. We conducted conditional logistic regression to identify the determinants of infant mortality. The determinants that were significant, at p<0.20 in the bivariate analysis, were included in the multivariable model. The cut-off value was p<0.05 for identifying substantial determinants in the multivariable conditional logistic regression. We estimated the time concerning three levels, such as decision-making for care seeking, arriving at the appropriate healthcare facility, and receiving appropriate healthcare services or referrals, and reported the median time for those three levels. The delay was defined in three levels for cases when the time duration for either of the three levels was more than the median duration identified in this study for respective levels. The pathway of care for the illness preceding the death was presented as a framework identified from the existing literature.

Patient and public involvement

We conducted training activities for ASHAs and ANMs on the reporting of infant deaths. We also conducted training activities for them on the care of sick infants after concluding the data collection. We also planned to discuss the findings with the public health administration.

ResultsBasic details of cases

We enumerated 100 infant deaths in the study area within the study period and recruited 100 live infants as controls. We were able to enumerate a more significant number of infant deaths through active surveillance mechanisms compared with existing mechanisms, which was double the existing reporting mechanism. The existing Health Management Information System reported 36 deaths, thus under-reporting 67.2% of deaths by the routine HMIS. We interviewed mostly mothers as primary caregivers of the cases (92%). We conducted the highest number of interviews during the third month (50, 50%) after death, completing all interviews within 3 months of the death. The distribution of variables related to sociodemographic determinants, antenatal determinants, intranatal determinants and postnatal determinants among cases and controls is given in table 1. Most infant deaths (70, 70%) happened during the first month of life with an equal number of male and female children. Almost half of neonatal deaths occurred during the first week of life, while deaths on day 5 contribute the highest to early neonatal deaths (see online supplemental figures 5–8).

Table 1

Distribution of variables related to sociodemographic determinants, antenatal determinants, intranatal determinants and postnatal determinants among cases and controls (n=200)

Cause of death

There was disagreement over the cause of death for 14 cases, which the senior clinician adjudicated. The three most common causes of neonatal deaths were birth asphyxia (29, 42%), prematurity and low birth weight (14, 20%), and pneumonia (8, 11%). The two most common causes of postneonatal deaths were pneumonia (10, 33%) and congenital birth disorder (9,30%). Overall, the three most common causes of infant deaths were birth asphyxia (30%), pneumonia (18%), prematurity and low birth weight (14%) (see figure 1, online supplemental table 2).

Figure 1

Distribution of cause of death in neonatal, postneonatal and infant deaths (n=100).

Determinants of infant mortality

The bivariate conditional logistic analysis concluded that the significant determinants of infant mortality were mother aged ≤25 years, father educated up to class 10, unskilled or unemployed fathers, more than five family members, ≤19 years of age at marriage, first Antenatal checkup (ANC) visit at >12 weeks of gestational age, and low birth weight. However, the multivariable model found that more than family members and fathers’ education up to class 10 were the only independent and significant determinants of infant mortality (see table 2).

Table 2

Determinants of infant mortality using bivariate and multivariable analyses (n=200)

Pathway of care and delay in receiving care

We obtained data on care received by 100 dead infants during the final illness preceding the death using the INDEPTH tool. Four infants were born at home, and two were born during transit to the hospital. One of those four infants born at home died at home only without receiving any care. Most infants (49, 98.0%) identified as sick after delivery were referred to higher healthcare facilities. Most (46, 93.9%) reached the hospital where they were referred, but three died during transport. Six infants died at a healthcare facility after being referred three times from different healthcare facilities (see figure 2). The main reason for those referrals was either the unavailability of beds or the requirement for intensive care. A similar pathway of care was observed among the infants for whom illness was identified at home. Two infants (4.1%) died at home due to trauma. Eight (17.0%) infants initially sought informal healthcare but ultimately sought care from a government/private healthcare facility. Despite the referral requirement, two infants didn't receive care at a higher healthcare facility. The remaining infants died in different healthcare facilities after being referred (see figure 3).

Figure 2

Pathway of care for cases that did not reach home after delivery.

Figure 3

Pathway of care for the cases of illness identified at home.

The period of stay in different healthcare facilities varied widely from 1 hour to 15 days (median: 5 days). The types of healthcare facilities visited during the referral period were private and government. The median time duration for decision-making related to healthcare seeking was 24 hours (level 1 delay). The median time duration for reaching any appropriate healthcare facility was one hour. Thirty-seven (75.5%) sick infants reached the healthcare facility at least 30 min after deciding to seek care. All cases received care at the healthcare facility or appropriate referral immediately after arriving there, leading to zero instances of level 3 delay. Further classification of cases concerning time in decision-making in seeking care and transport to the healthcare facility with the context of diseases is mentioned in online supplemental tables 3–5.

Discussion

Our present study described the causes and determinants of infant mortality along with the pathway of care for the illness preceding the death and the delay in seeking care. We enumerated a more significant number of infant deaths through active surveillance mechanisms compared with existing mechanisms. Almost two-thirds of infant deaths happened in the neonatal period, and half of the neonatal deaths occurred in the first week of life. The most common cause of neonatal death was birth asphyxia, and the most common cause of postneonatal death was pneumonia. The most common cause of all infant deaths was birth asphyxia. The multivariable conditional logistic regression concluded that only the father’s education (education status till class compared with class 12 or higher) and family size of more than five were significant risk factors for infant mortality. Among 49 infants for whom illness was recognised at home, 2 died without receiving care, and the rest received care either of formal or non-formal type. The median delay of 24 hours was observed in level 1, with no delay in level 3. The median delay of level 2 was 30 min.

The latest SRS data revealed that almost 71% of infant deaths happen during the neonatal period, similar to our study.2 The causes of postneonatal deaths depend on the care provided during postnatal periods. Different effective interventions, such as the expanded programme of immunisation, HBNC, etc, caused a certain reduction in postneonatal deaths.21 Those interventions not only involve healthcare delivery at the hospital level but also involve community-based approaches to improving infant mortality. However, neonatal deaths depend on the antenatal care services and care during delivery. In India, those interventions related to antenatal and delivery care practices are in place, but neonatal deaths are still higher than postneonatal deaths. It is expected that with the improvement of healthcare facilities, the proportion of postneonatal deaths will be higher than neonatal deaths, with major causes of death as different congenital defects. With the advent of a more advanced healthcare delivery system, more premature infants can survive. Simultaneously, the early identification of danger signs among infants is also a crucial part of the reduction in infant mortality. The involvement of community health workers in the early identification of danger signs and prompt referral of sick infants were found to be effective strategies in low-middle-income countries. Also, more comprehensive training among healthcare workers related to essential newborn care with the availability of facilities like intubation can reduce deaths from birth asphyxia. The interventions thus already in place need to be strengthened by means of human resource training, evaluation of programme activities and in-depth gap analysis. The proportion of early neonatal deaths was different in the other studies compared with our study. Earlier studies conducted in India suggest that a proportion of early neonatal mortality varied between 60% and 80% in all neonatal deaths.2 22 23 Our study reported a proportion of early neonatal mortality of 50%, which suggests that as a result of prematurity, low birth weight and birth asphyxia, infants survived beyond the first week of life due to the availability of facilities like special newborn care units (SNCU). The deaths during the first day of life were highest in other studies, whereas our study reported the highest number of deaths during the fifth day of life. The proportion of early neonatal deaths during the first day of life was below compared with Indian studies but lies in between the result reported by Tesfay et al and Ma et al.24 25 This suggests the availability of facilities for improving the survival of sick infants.

Our study finds that birth asphyxia is the most common cause of infant and neonatal mortality, which differs from the SRS report. The report suggested that the most common cause of neonatal mortality was prematurity and low birth weight. But the result of our study corroborates with the findings of the study conducted by Shah et al, Bapat et al and Deshmukh et al23 26 27

Similarly, the distribution of causes of infant deaths was different from the SRS report. But the results are similar to the study conducted by Rai et al.22 Rai et al reported the three most common causes of infant mortality as birth asphyxia (19.1%), pneumonia (17.1%), and low birth weight and prematurity (16.3%).22 Birth asphyxia usually happens due to adverse events during the peripartum period with the unavailability of a trained health workforce to facilitate intubation during the time of delivery. The latest newborn resuscitation protocols suggest early intubation when required with referral to higher centres. The deaths due to prematurity can be prevented by the availability of trained healthcare workers with beds in SNCUs. Our study suggested that there is a lack of neonatal resuscitation facilities, which causes birth asphyxia among neonates.

The multivariable conditional logistic regression concluded that the number of family members more than five (adjusted OR (aOR) 6.87, CI 2.07 to 22.77) and education of the father till class 10 (aOR 3.58, CI 1.33 to 9.65, p=0.012) were the significant risk factors of infant mortality. Fathers’ education plays a crucial role in health-seeking behaviour, specifically in a patriarchal society, as in our study area. The higher education of fathers is expected to be associated with a positive attitude towards availing of healthcare facilities during pregnancy and delivery.28 Lower fathers’ education can be attributed to poor health-seeking behaviour and poor socioeconomic conditions.29 30 Those risk factors contributed to infant mortality directly or indirectly. The higher number of family members is a surrogate marker of lesser usage of contraception, which is an essential determinant of infant mortality.31 It is already established that poor acceptance of contraceptive practices is already a known risk factor for infant mortality.

Our study didn’t find any level 3 delay, which was different from other studies.22 32 Those studies conducted in lower-middle-income countries reported a significant delay in receiving appropriate care after reaching the healthcare facility. Study participants reported no delay in receiving care or appropriate referral after reaching the healthcare facility. The reporting of level 3 delay may be due to social desirability bias as the interviewer also belonged to the healthcare facilities where care was sought. The median level 1 and level 2 delay observed in our tudy were similar to the evidence from other studies.32–34 The proportion of cases without delay was much higher than in other studies.22 32–34 The proportion of infants who sought care for illness was similar to that in studies conducted by Rai et al.22 This suggests that the care-seeking behaviour during illness is better than in other studies. A less number of infants with Level 2 delay can be attributed to the availability of free ambulance facilities or the condition of roads for transport.

Our study used an active surveillance method coupled with data triangulation to enumerate infant deaths. We could enumerate almost double the number of infant deaths (n=100) compared with the existing Health Management Information System (HMIS). The reporting of infant deaths in HMIS depends on the incentive-based reporting of infant deaths by community health workers. Though reporting of a death is incentive-based, it might be under-reported due to various operational reasons. Thus, the findings of this study have an important implication for improving routine infant death surveillance and demanding appropriate intervention in the health system. It might be possible that infant deaths in disadvantaged communities are under-reported to a larger extent, as suggested in earlier studies.35

This is the first study conducted in Odisha that reported on the determinants of infant mortality. This is also the first study in Odisha that reported a pathway of care with delay. We conducted all of the interviews within 3 months of deaths, excluding recall bias to a great extent. We used an active surveillance method with data triangulation to enumerate deaths in our study. We used verbal autopsy as a method of cause-of-death identification, a well-accepted method in countries with a low percentage of medical cause-of-death identification, specifically in low-income and lower-middle-income countries.

We got fewer deaths than our estimates even after active surveillance and data triangulation. We enumerated a total of 100 infant deaths in place of an estimated sample size of 110 cases. However, we were able to achieve adequate power and discordant pairs. This under-reporting can’t be generalised due to the limited study area. However, the higher CIs of the determinants in the multivariable model conclude that the number of variables was higher than the sample size in the multivariable model. Due to feasibility, we also do not include the newer verbal autopsy methods, such as algorithm-based methods.

Conclusion

In our study conducted in rural Odisha, birth asphyxia, pneumonia, prematurity and low birth weight were found to be the common causes of infant mortality. Birth asphyxia and pneumonia were the most common causes of death during the neonatal period and postneonatal period, respectively. Determinants of infant mortality were mostly socioeconomic, with lower education status of fathers and large family size as the independent factors associated with infant mortality. Pathway of care analysis revealed a significant delay at level 1, that is, a delay in decision-making to seek care. Study findings highlight the importance of accentuating cause-specific facility-level interventions, particularly focused on the perinatal period. Social factors remain an important attribute for infant mortality as well as delay in seeking care. Thus, it must be addressed to maximise the maximum benefit of health facility related interventions. We also recommend improving the infant death reporting system through existing mechanisms complemented with an active surveillance mechanism.