Under-estimation of maternal and perinatal mortality revealed by an enhanced surveillance system: enumerating all births and deaths in Pakistan

Pakistan is among the ten countries estimated to account for 60% of global maternal deaths. Based on interview survey methods, the Maternal Mortality Ratio (MMR) in Pakistan is estimated to have declined from 430/10⁵ live births in 1990 to 180 in 2015 [1], and neonatal mortality is estimated to have declined from 64/10³ live births in 1990 to 46/10³ live births in 2015 [2]. Similar to several other countries, Pakistan did not achieve the Millennium Development Goals (MDGs) 4 and 5 which relate to these indices. In 2015, countries adopted the 17 Sustainable Development Goals with 169 targets to be achieved by 2030 [3]. Of the 13 health targets, the first two are to reduce the estimated global maternal mortality ratio to less than 70/10⁵ live births, and reduce neonatal mortality (0–27 days) to ≤12/10³ live births [3]. In order to monitor the progress on these targets, a renewed emphasis has been placed on the need for reliable and timely data involving counting all births and deaths, especially around the time of birth [4].

Most of the available estimates of maternal and neonatal mortality rates from lower and middle-income countries have been reported at the national level, with a wide variation among countries [1, 2]. Pakistan is the sixth most populous country in the world with an estimated 185 million people in 2012–13 [5]. There are wide variations among indicators including mortality rates among the six provinces [6]. For example, the estimated MMR for Punjab province is 227/10⁵ live births, compared to 785 in Baluchistan [7], and the estimated Neonatal Mortality Rate (NMR) for Khyber Pakhtunkhwa province is 41/10³ compared to 63 in Punjab and Baluchistan [5]. In Pakistan, these indices are derived from demographic and health interview sample surveys (DHS), to estimate mortality rates at a national or provincial level, but provide no information on district or sub-district variations because of small sample sizes [5, 7].

The current routine health information systems in Pakistan that report data on pregnancies, births, and deaths are inadequate in several aspects. Neonatal mortality may be under-reported by the District Health Information System (DHIS), since it collects data only from public health facilities [8], excluding the 34% of the births in private health facilities [5].

The Lady Health Worker (LHW) Program operates at the community level but covers only 70% of the population. The LHWs register pregnant women, collect birth and death data, and provide family planning, health education and referral services to pregnant women and families in their areas. Some community level data collected by LHWs are not linked with the DHIS [9, 10]. This contributes to inadequate data available to decision makers [11]. Moreover, maternal deaths may be under-reported since LHWs follow pregnant women for medical risks only until parturition, whereas maternal mortality can occur up to 42 days after delivery. Maternal Newborn and Child Health Program data are not incorporated into the health facility reports, nor is their Program data linked with the DHIS [12, 13]. Thus there is no national or provincial health information system that reconciles data from all sources, including the private sector and community areas not covered by LHWs, to provide accurate maternal, perinatal and neonatal mortality rates at a district or sub-district level. The aims of this study were to: 1). test the feasibility of establishing an enhanced surveillance system that captures data from all available health information systems, and extends surveillance to areas without any information systems; 2). estimate maternal, perinatal and neonatal mortality rates by more complete enumeration of all pregnancies, births, maternal, perinatal, and neonatal deaths and derive estimates of under-enumeration by comparison with previous routinely collected data; 3). compare mortality rates calculated by the surveillance system with the national and sub-national mortality rates estimated by Demographic Health Surveys and international agencies.

An enhanced surveillance system was established that endeavoured to capture all births and deaths using information from both public and private healthcare facilities, and extended community coverage, with an improved completeness of reporting and reconciliation of data.

Study population

This population-based prospective study was conducted over 1 year (June 2015–May 2016) in Tehsil Havelian (a sub-district) of the District of Abbottabad of Khyber Pakhtunkhwa province located in the North of Pakistan, approximately 110 km from Islamabad (the capital of Pakistan). The estimated population of the District Abbottabad in 2010 was 1.179 million, and the study area of Tehsil Havelian was 341,891 (29% of the district population) [14]. Approximately 80% of the population of Tehsil Havelian live in rural areas, and 54% have completed primary level education [14, 15]. Figure 1 shows the geographical location of District Abbottabad and the study area. The total population (both sexes and all ages) of the study area registered by the LHWs and CHWs at the start of this study was 368,454, enumerated by visiting each household at the commencement of the study in June 2015. The difference between the estimated populations in 2010, and actual population in 2015, may be attributed to population growth. However, other factors, including in-migration related to effects of natural disasters, also contribute to the population growth. Of the total registered population, 293,344 (80%) resided in the LHWs areas and 75,110 (20%) resided in the areas of CHWs. The study period for enhancement of surveillance was from 1 June 2015 to 31 May 2016. The study population consisted of 51,690 married women aged 18–49 years who were permanent residents of Tehsil Havelian.

Data sources and collection

Prior to the establishment of the enhanced surveillance system in the study area, data on pregnancies, births and deaths at the community (household) level were reported monthly by the LHWs to their program. Births and deaths that occurred at the community birth stations were reported by the Community Midwives. These monthly reports by LHWs and Community Midwives were entered at the District level and flow directly to their Provincial offices. Identified high-risk pregnancies by LHWs, were referred to appropriate hospitals but neither recorded, nor followed-up for the outcome. Births and deaths at public health facilities were reported to the District health office monthly and entered into the District Health Information System for onward submission to the Provincial DHIS office. These data, reported from the community or public health facilities, were not analyzed or used at District level for improving the maternal and child health. The DHIS lacks a mechanism to assign unique identifiers to the reported deaths.

Following the establishment of the enhanced surveillance system in the study area pregnancies were recorded at the household level and followed after birth for maternal, perinatal and neonatal mortality by the LHWs. Monthly reports from the LHWs were aggregated at the community level by LHW supervisors. Community Health Workers were employed to record similar data from areas not covered by LHWs. Birth and death data were also captured from the private health facilities in addition to the routine monthly reports from public health facilities. Data from all sources (community level and health facilities) were integrated at the District office. All deaths (maternal, neonatal, early neonatal and stillbirths) were recorded from the community level (households), and from all health facilities (public and private), and verified directly from the households, if the death met the inclusion criteria. Figure 2 shows the routine process of data collection on births and deaths, and the extensions made to the routine process by the enhanced surveillance system.

A total of 283 LHWs were engaged in data collection that covered 79% of the study population in Tehsil Havelian, Abbottabad District. Each LHW provides services to a population of 800–1000 in a defined geographical area. LHWs maintain a register of all married women aged 18–49 years for their assigned households and visit 7–10 houses per day to update the records and registers. LHWs register any pregnant women they find during their routine visit to the households, and prepare a mother and child health card for follow-up. In an endeavour to capture all pregnancies, births and deaths at the community level, an additional 73 Community Health Workers) were recruited to collect information from uncovered areas.

A list of all married women aged 18–49 years living areas was prepared by LHWs and CHWs and a unique code was assigned to each woman. A total of 51,690 women were recorded by the LHWs and CHWs, residing in the study area. Among the listed women, 40,952 (79%) were provided with services by LHWs, and 10,738 (21%) resided in the areas of CHWs. During household visits, LHWs and CHWs registered pregnant women, and followed them up to 42 days after delivery. This process of registration and follow-up continued for the entire study duration. Women delivering in last month of the project were followed until the end of the month.

A list of all public and private health facilities situated in the study area was obtained from the local District Health Authority. Nineteen public health facilities that fulfilled the inclusion criteria, i.e. providing antenatal, delivery, postnatal or newborn care, were included into the study. A copy of monthly reports of selected public health facilities was obtained from the Provincial cell of the District Health Information System. Ten private health facilities that provide maternity services (inpatient and outpatient), or pediatric services (both inpatient and outpatient) were included in the study. A focal person was nominated by the District Health Officer to collect monthly reports from the private hospitals which were submitted to the District Health Office.

Mother and Child Health cards were used to capture data on pregnancies, births and deaths from both LHW’s and CHW’s areas. These are currently used by the LHWs to capture data from pregnant women on all previous pregnancies, prenatal care, medical conditions, delivery and pregnancy outcomes including baby alive or dead, birth weight, sex, and newborn complications. Data from public and private health facilities were obtained through the DHIS Monthly Reporting Forms (DHIS-21 and 22). These forms are currently in use by the Health Department to collect data from the public health facilities and currently report data on services provided by the health facility. The reporting form for secondary health facilities differs from that of the Primary Health Care Facility Monthly Report form in that it has additional inpatient and laboratory data.

LHWs and CHWs were trained on the process of selecting eligible women ages 18 to 49 years, enlisting them, and taking informed written consents in their respective areas. CHWs recruited for the research project were provided with an additional 2 days training on correct completion of the Mother and Child Health Card, monthly report forms, and referral procedures. Of 73 CHWs selected for the research project, 61 (84%) attended the initial training session, and those who could not attend were trained in subsequent training sessions at their respective health facilities. One day training was provided to staff engaged for data collection from the public and private hospitals on correct completion of DHIS monthly reports, and collection of the reports from the selected private healthcare providers and hospitals, and selected public hospitals.

Higher maternal mortality, perinatal mortality and neonatal mortality rates were observed through enhanced surveillance compared to mortality rates estimated by the routine monitoring system. Integration of data from various sources to identify maternal, perinatal and neonatal deaths and extending coverage to previously uncovered areas, improved the enumeration of births and deaths and provided accurate mortality rates in the study area. A small proportion (4%) of the study population living in very remote areas could not be completely covered by the enhanced surveillance system for the entire duration of the study.

Sex difference

Rates of stillbirth (38 versus 25/10³ births), early neonatal mortality (36 versus 23/10³ live births, perinatal mortality (73 versus 48/10³ births) and neonatal mortality (42 versus 38/10³ live births) were all higher in males than females in 2015/16. This difference was statistically significant for all rates except neonatal mortality (Table 6 in Appendix).

A national MMR of 178/10⁵ live births was reported by the international agencies in 2015, whereas the only available MMR of 275/10⁵ live births in the province was estimated by a DHS conducted in 2006-07 (Fig. 3).

A national NMR of 46/10³ live births for 2015 was reported by the international agencies compared to a NMR of 40/10³ live births calculated by the enhanced surveillance system. The DHS for 2012–13 reported national and provincial NMR of 55/10³ live births and 41/10³ live births, respectively (Fig. 3).

This study demonstrates the under-reporting of maternal and neonatal mortality rates in the study area compared to the previous 1.5 years; maternal mortality was under-estimated by 27% and neonatal mortality by 50%. The mortality rates in the study area calculated from the enhanced surveillance system were higher than those estimated from the previous routine LHWs data. This study provides accurate maternal, perinatal and neonatal mortality rates by establishing an enhanced surveillance system that captured births and deaths from 96% of the study population, through multiple data sources, including data collection from the public and private health facilities and extended community coverage, in a rural area of Khyber Pakhtunkhwa province, Pakistan. During enhanced surveillance 2015/16 higher maternal, neonatal and perinatal rates were found in CHWs areas subject to extended coverage than in the LHWs areas reflecting the more rural and remote character of the previously uncovered areas. The inclusion of the CHW areas increased coverage of women by 20%, but this could not be maintained for 4% because of difficulties associated with remoteness. Comparison of the LHWs data 2015/16 with the previous LHWs data shows that enumeration by LHWs improved with enhancement of surveillance, quite apart from expansion of coverage to new areas. Improvement in maternal, neonatal and perinatal mortality was due to improved completeness as a result of data collection on births and deaths from all sources including community, public and private health facilities, and increased coverage to CHW areas with higher maternal and neonatal mortality.

The surveillance system successfully integrated birth and death data from available routine health information system and extended the surveillance system to the areas and health facilities from where previously no birth or death data were reported. The robust enhanced surveillance system provided sufficient evidence of underestimation of mortality rates in the study area through before and after comparison of mortality rates separated by 1.5 years. The comparison of the study area with itself is closely related in time (1 year), with no change in socioeconomic status and health system, and no disaster, or epidemic diseases or civil disturbance over the comparison period. Although a higher maternal mortality was calculated using the enhanced surveillance system data in the study area compared to routine health information systems, the number of maternal deaths were not sufficient to demonstrate a statistically significant difference in maternal mortality rates. This is a consequence of the short duration of the study, and the small number of maternal deaths.

This is the first study in Pakistan to include private health facilities in a maternal and perinatal surveillance system. The proportion of deliveries reported by private health facilities was 11% of the total births in the study area. This proportion is less than the 2012/13 DHS Provincial statistics of 24% of births [5]. Possible reasons for the discrepancy may include lower affordability for private hospitals in Tehsil Havelian than that for the Province, or there may be under-reporting of births by the private hospitals for various reasons. Nevertheless, the likelihood of missing any birth or death that occurs at private health facilities is much less, because 96% of the population in the study area was covered either by the LHWs or by the CHWs during the enhanced surveillance in 2015/16.

Estimates of MMR in Pakistan suggest a decline from 431/10⁵ live births in 1990 to 178/10⁵ live births in 2015. Although a substantially higher maternal mortality was observed by the enhanced surveillance system in the study area of Tehsil Havelian (247/10⁵ live births), compared to the national MMR of 178/10⁵ estimated by the international agencies (World Bank, WHO, UNICEF, UNFPA) for the year 2015, the Provincial MMR of Khyber Pakhtunkhwa estimated by 2006/07 DHS was consistent with the enhanced surveillance system at 275/10⁵ live births. A study by Sathar reported an estimated national MMR of 220/10⁵ live births in 2012. The study also estimated Khyber Pakhtunkhwa Province MMR of 423/10⁵ live births in 2001, 275/10⁵ live births in 2006, and 206/10⁵ live births in 2012 [17]. A population-based prospective study on active surveillance of pregnancies and their outcomes conducted in six countries, including Pakistan, during 2010-2012 reported a MMR of 313/10⁵ live births, in a rural District of Sindh province [18], similar to MMR reported by DHS 2006/07. A retrospective study of facility-based maternal mortality which collected data for 10 years (2002–2012) in a tertiary care hospital of Khyber Pakhtunkhwa province (Pakistan) reported a MMR of 772/10⁵ live births [19], but this may be affected by referral of complicated cases. Under-enumeration of maternal and neonatal deaths by the LHWs was noted in a study in Lahore (Punjab Province) in 2010, that reported underreporting of maternal and infant deaths by LHW due to fear of not maintaining adequate performance indicators [9]. Verification of LHWs reports showed 92.5% correctly reported maternal death, while 5% underreported and 2.5% over reported maternal deaths [9].

The neonatal mortality in Pakistan changed little during the past two decades according to survey data from the DHS, and the enhanced surveillance neonatal mortality rate is consistent with national and provincial estimates around 2015, but much higher that the LHW data (Fig. 3). The neonatal mortality rate reported by Pakistan DHS 1990/91 was 51/10³ live births (1986–1990), Pakistan DHS 2006/07 (2002–2006) was 54/10³ live births and Pakistan DHS 2012/13(2008–2012) was 55/10³ live births [5]. A non-significant decrease in neonatal mortality was reported by DHS 2012/13 in Khyber Pakhtunkhwa, from 48/10³ live births in 1990 to 41/10³ live births in 2012 (Fig. 3) [5]. A population-based prospective study reported a neonatal mortality of 50/10³ live births, in rural District of Sindh province during 2010–2012 [18].

The stillbirth morality rate estimated from the previous 1.5 years routine LHWs data was higher (32/10³ births) compared with the stillbirth rate estimated from the LHWs data collected by the enhanced surveillance system (29/10³ births). The over-enumeration of stillbirths by LHWs could be due to their under-enumeration of early neonatal mortality. An international review in 2006 found that a live birth may be recorded as a stillbirth if the baby died immediately after birth, because of various reasons including inadequate knowledge, avoidance of blame, fear of extra work, or poor assessment for the signs of life [20]. A study conducted in 2011/13 [21] reported stillbirth rates of 50/10³ births in District Thatta (rural district), Sindh Province of Pakistan. Another prospective study in Sindh Province reported similar higher rates of stillbirths (66/10³ births) in 2003 [22]. These rates were higher than calculated from our enhanced surveillance system, as well as that estimated by Pakistan DHS 2012/13.

It is worth noting that neonatal mortality (40/10³ live births), early neonatal mortality (30/10³ live births), stillbirths (32/10³ births) and perinatal mortality (60/10³ births) in the study area in 2015, is similar to Provincial neonatal mortality (41/10³ live births), early neonatal mortality (33/10³ live births), stillbirths (31/10³ births) and perinatal mortality (63/10³ births) estimated by the DHS in 2012/2013 [5]. However, interpretation needs to take account the differences in methods and time period of 8 years between our study in 2015/16 and those estimated by the DHS 2012/13, and the differences between the Tehsil Havelian and the entire province.

Adolescent women are considered high risk for adverse pregnancy outcomes. A recent population-based prospective study conducted from 2010 to 2013 in six low and middle-income countries including Pakistan, reported a higher rate of maternal, neonatal and perinatal mortality among women aged 15–19 years compared to women age ≥ 20 years [23]. This study does not include pregnant women aged < 18 years as the sample size would be insufficient for subgroup analyses, and extended consent would be required from parents and additional approval from research ethics committees.

Higher rates of stillbirth, early neonatal, neonatal and perinatal mortality among males than females are consistent with the Pakistan DHS 2012/13 that reported higher neonatal mortality in males compared with females in Pakistan. An analysis using data from the Pakistan Demographic Health Survey 2006/07 reported a statistically significant hazard ratio of 1.57 for neonatal mortality in males compared to female neonates [24]. This is also consistent with international statistics where neonatal and infant mortality in males are reported to be higher than females which provide further validation for the study [25]. Reasons for this higher mortality in males are explained by biological factors, including a higher risk of respiratory syndrome (related to late maturity), infectious diseases, congenital malformations of the urogenital system in males, and fetal growth retardation [26–28]. Population characteristics of the study population were similar to the provincial population in terms of total fertility rate, general fertility rates, body mass index and sex ratio at birth (Male/Female) [5]. Hence the results of the study likely reflect the Provincial population.

The Pakistan DHSs estimated maternal, perinatal and neonatal mortality only at the national and provincial level, which may mask the district or sub-district variations in mortality rates. The need to access district and sub-district data is also reported by a South African study in 2016 [29], which emphasizes the need to use disaggregated data at the sub-district level for equitable resource allocation and targeting the areas in need. A study on a vital events surveillance system in India estimated causes of maternal and neonatal deaths in 2012 using CHWs and supported the application of targeted community-based interventions that resulted in a significant reduction in neonatal mortality [30].

The LHW Program, having 70% coverage of national population, provides an opportunity to measure accurate mortality rates at the sub-district level if coverage is enhanced to capture the entire population. The enhanced surveillance system demonstrated that the birth and death data reported by LHWs, CHWs (for enhanced coverage), community midwives, health care facilities (public and private), and the routine health information system (DHIS), can be reconciled to provide accurate and timely mortality rates at a district and sub-district level. This could be used to strengthen the healthcare delivery system through the application of area-specific and cause-specific targeted healthcare interventions and improving the coverage of current health care program in Pakistan. This surveillance system can enable health managers to utilize resources more efficiently and target them to the area most in need, thus have a maximum impact of the targeted intervention in the reduction of mortality rates.

Following decentralization of services in Pakistan (18th Amendment of the Constitution) in 2010, it is imperative to have accurate maternal, perinatal and neonatal mortality rates at the district and sub-district level. In addition, the local government ordinance highlights the need to empower local governments and improve the governance by decentralized decision-making [8, 31].

Accurate and timely data on mortality is required to monitor progress, implement health interventions and to evaluate health programs at national and sub-national levels [32, 33]. A global assessment of civil registration and vital statistics reported most South Asian countries, including Pakistan, have weak vital civil registration with inadequate coverage and poor quality data on deaths and causes of deaths [33]. In these circumstances, maternal, perinatal and neonatal mortality data are obtained from household censuses, Demographic Health Surveys [34], Multiple Indicator Cluster Surveys [35] and reproductive age mortality surveys [36], employing direct death inquiry of household members over a retrospective period, and/or indirect methods such as children ever-born and children surviving, and orphanhood and widowhood [37, 38] questions in national and/or sub-national surveys [39]. However, these sources have various limitations, including underestimation of maternal deaths and requirement of large sample sizes [40–42]. Although a population census may be a better approach to measuring mortality rates than surveys, there are issues with data quality and omission of up to 50% of deaths in population censuses has been reported [42].

A small proportion (4%) of the study population living in very remote areas could not be completely covered by the surveillance system for the entire duration of the study. Eighteen CHWs recruited for these areas registered 2599 women of reproductive age (18–49 years), ten left during the first month, and eight afterwards, mostly due to the arduous nature of the work. However, 37 births and no deaths were recorded by the CHWs prior to their resignation. These data were not included in analyses.

Further research is needed to demonstrate the feasibility of using this enhanced surveillance system that integrates births and deaths data from all possible sources for application of area-specific and cause-specific interventions with measurement of the impact of the reduction in mortality rates. This is particularly required in districts with low community coverage by LHWs and Midwives. Opportunities should also be explored to link births and deaths captured by routine health information systems with civil registration authorities to strengthening civil registration and vital statistics. Research is also needed to measure the effects of adequate surveillance on Mother and Child Health Programs and expected reductions in MMR and NMR.

A surveillance system that triangulates birth and death data from all health information sources, extends coverage and follows all pregnancies for the outcome, is needed to generate accurate mortality estimates at a district and sub-district level. A decline in maternal mortality in Pakistan was reported by the international agencies based on DHS. However, when maternal mortality was measured in a district through near complete enumeration of births and deaths by following all pregnancies in a defined population, a higher maternal mortality was observed. Similarly, perinatal, early neonatal and neonatal mortality in the study area was significantly higher than that estimated from the available data. There is no surveillance system in Pakistan that provides accurate and timely maternal, perinatal and neonatal mortality data at the district and sub-district level. This makes the problem less visible to policymakers and program managers.

A robust surveillance system that is capable of providing district and sub-district mortality rates in order to target areas with higher mortality rates is essential, and thus can lead Pakistan towards achieving the Sustainable Development Goals.