Assessing health and economic outcomes of interventions to reduce pregnancy-related mortality in Nigeria

Country- and region-specific data were synthesized using a computer-based model that simulates the natural history of pregnancy and childbirth. Separate models were adapted to Southwest and Northeast zones using survey-based data and information about recognition of the need for referral, access to transport, and appropriate facilities [8, 9, 13]. Model outcomes include clinical events (e.g., pregnancies, live births, maternal complications), measures of maternal mortality (e.g., MMR, proportionate mortality ratio [i.e., proportion of deaths among women aged 15–45 that are pregnancy-related], and lifetime risk of maternal death), population outcomes (e.g., life-expectancy), and costs.

Strategies consisted of increasing the coverage of effective interventions, and could include improved logistics. Following standard recommendations for economic evaluation, we calculated incremental cost-effectiveness ratios, defined as the additional cost of a specific strategy divided by its additional clinical benefit, compared with the next least expensive strategy [14]. We considered interventions with cost-effectiveness ratios of less than the per capita GDP ($1,170) to be very cost-effective [2, 15]. Sensitivity analyses are conducted to assess the impact of parameter uncertainty, and Monte Carlo simulation was used to generate the number of per woman events such as pregnancies, live births, and facility-based births.

The Global Maternal Health Policy Model is a previously published computer-based model that simulates the natural history of pregnancy and pregnancy-related complications over a woman’s lifetime, and aggregates outcomes to a population level [16]. Factors modeled at the individual level include the probability of pregnancy (conditional on age, contraceptive use, and clinical history), the probability of spontaneous or induced abortion, and the risk of direct pregnancy-related complications such as hypertensive disorders, obstructed labor, hemorrhage, and sepsis. The case fatality rates of these complications are conditional on the type, severity and underlying comorbidity. Nonfatal secondary complications considered include neurological sequelae, obstetric fistula, severe anemia, and infertility. In addition to pregnancy-related mortality risk, women face an annual risk of death from age-specific all-cause mortality. The model is described in more detail in the Additional file 1.

Strategies to reduce maternal deaths consist of improving coverage of effective interventions, either individually or packaged as integrated services. These include: reducing the unmet need for contraception; increased accessibility to safe abortion and post-abortion care; prevention and treatment of anemia (including intermittent prevention and treatment of malaria in pregnancy); and increased availability of intrapartum and postpartum care. Recognizing that the investments in infrastructure required to assure high-quality intrapartum care will need to happen in phases, stepwise improvements are modeled over time.

The overall impact of interventions results from a reduction in the incidence and/or case fatality rate of a complication. Both mechanisms depend, in part, on access to specific services, trained personnel, and quality of the facilities delivering these services. Therefore, the model explicitly considers the location of delivery, type of assistance, access to basic or comprehensive obstetrical care, and the ability to overcome barriers around the timing of delivery (See Figure 1). Delivery setting is differentiated by provider (e.g., family member, traditional birth attendant [TBA], SBA or no one) and by site (e.g., home versus facility). Facilities providing basic EmOC (bEmOC) are assumed to be capable of administering injectable antibiotics, oxytocic drugs, and sedatives or anti-convulsants, and also conducting assisted vaginal delivery, removal of placenta and retained products. Facilities capable of comprehensive EmOC (cEmOC) are able to provide blood transfusion services, cesarean delivery, and management of advanced shock in addition to all the aforementioned bEmOC services [17].

Health facilities in Nigeria are classified as primary, secondary and tertiary [7]. Primary care facilities (e.g. Alausa Primary Health Care center, Lagos State) are the most abundant, but SBAs aren’t always available, and only few can provide all the services that constitute bEmOC [18, 19]. Conversely, all tertiary facilities (e.g. University College Hospital Ibadan, Oyo State), and most secondary facilities (e.g. General Hospital Calabar, Cross River State) can provide comprehensive emergency obstetric care, but they are fewer, and are predominantly in urban areas [19]. In adapting the model to the Nigerian context, we map the current facility structure on to the model structure described above.

The model is programmed using TreeAge Pro 2009 (TreeAge Software Inc., Williamstown MA) and analyses conducted using Microsoft Excel 2010 and Visual Basic for Applications 7 (Microsoft Corp., Redmond WA).

Selected model inputs and assumptions are shown in Table 1 and Table 2. In general, estimates of incidence and case fatality rates associated with pregnancy-related complications were obtained from published data, and a plausible range for sensitivity analysis was based on systematic review of the literature. Data on facility births, skilled birth attendants, family planning, and antenatal care were from country-specific surveys and government reports (more details in the Additional file 1).

For women delivering outside an EmOC facility, the probability of a successful referral depended on overcoming three categories of delays: delay in recognizing need for referral and being willing to go; delay in transport to referral facility; and delay in receiving appropriate care at appropriate EmOC facility. Assumptions about barriers to successful referral were based on country reports, published and grey literature, as well as in-country visits (between March and December 2010) to elicit expert local opinion [19, 45‐48]. We conducted an in-country survey of 121 healthcare facilities and 700 women aged 15–45 (see Additional file 1) to provide insight into the range of values for sensitivity analysis [49]. After using the best available data to parameterize the national and sub-national models (for Southwest and Northeast zones), key outcomes were generated and compared to independent data. Selected results of this exercise are shown in Table 3, and the procedure used is described in the Additional file 1.

Selected costs are shown in Table 1. Details of the costing methodology are included in the Additional file 1. Costs of delivering interventions and treating maternal complications were estimated from the United Nations Population Fund’s (UNPF) Reproductive Health Costing Tools (RHCT)[21]. Costs associated with personnel (salaries) were from public access databases [52, 53]. Drugs and supply costs were from the United Nations Children Fund’s (UNICEF) Supply Catalogue [54] and Management Sciences for Health (MSH) International Drug Price Indicator Guide [54, 55]. To estimate the costs of improving transport and scaling up facilities we used methods previously described [16] and assumptions from in-country experts. All costs were converted to 2008 U.S. dollars.

Modern contraceptive prevalence rates range from 3.5% in the Northeast zone (17.6% unmet need) to 21% in the Southwest zone (19.7% unmet need) [8, 9, 13]. Reducing the unmet need by 25% to 100% reduced maternal deaths by 4% to 17% in the Southwest and 3% to 13% in the Northeast zone (Table 4, see Additional file 1). Because the unmet need is based on survey-derived preferences of Nigerian women questioned now, and the number of desired children remains high, modeling elimination of the unmet need only reduces the TFR from 5.9 to 4.9. Anticipating changes to fertility preferences over time, we conducted a secondary analysis in which the use of modern contraceptives was increased by 25% to 50%, corresponding to a contraceptive prevalence rate of 34.7% and 59.7%, respectively (Table 4). The TFR was reduced from 5.9 to 4.4 (25% increase), and to 2.9 (50% increase). With a contraceptive prevalence rate of 59.7%, approximately 1 out of 2 maternal deaths was prevented. All of the above strategies were extremely cost-effective with incremental cost-effectiveness ratios less than $10 per year of life saved. Additionally, increasing the use of modern contraceptives in the model led to a decline in abortion related deaths (more details in the Additional file 1).

Figure 2 shows the benefits expected from improving intrapartum care (upgrades), superimposed with the additional benefits from reducing the unmet need for contraception and increased coverage of safe abortion services. Strategies that improved intrapartum care alone had higher cost-effectiveness ratios (i.e., least attractive), reflecting the higher costs required for infrastructure improvements. Strategies that only improved family planning and safe abortion had very low cost-effectiveness ratios (i.e., very attractive), but reduced mortality by 20.5%. A strategic approach that involves simultaneous improvement in intrapartum care, family planning and safe abortion was the most efficient and associated with incremental cost-effectiveness ratios between the two aforementioned strategies. Further details are provided in the Additional file 1.

Table 5 provides robust insight into the importance of investing in all three domains of family planning, safe abortion and intrapartum care – any approach that focuses on only one of these, to the exclusion of the other, will be less effective and cost-effective. But decision makers will still need to decide on how to proceed with such stepwise investments in all three domains. While these choices will be a function of many contextual factors, we sought to provide decision makers with information on the expected benefits, value and efficiency of different approaches. We applied a time dimension of twelve years to the analysis to stacked cohorts, and conducted a stylized exercise to identify scale-up approaches that would be more and less efficient.

For purposes of generating main themes, and not trying to compare an unlimited number of hypotheticals, we restricted our approaches to three general options:

We used UN population projections for Nigerian women aged 15–45 years over a 15-year period (from 2006 – 2022) and model projected estimates of annual probability of mortality, proportionate mortality ratio and cost [56]. Upon applying data specific for “current status” to the cohort for 2006 and 2007, our population projections approximated UN projections (with age-specific variations in population ranging between −3% and 1% of UN data), as did our estimate of the number of maternal deaths each year (48,483 and 49,810 in 2007 and 2008 respectively). Further details about these methods are provided in the Additional file 1.

We applied model generated probabilities of mortality and costs to the three approaches described above, scaling the interventions over a 12-year period (from 2011 to 2022), and estimated the number of maternal deaths averted with each approach. Approach 1 could avert over 270,000 maternal deaths, while Approach 2 and Approach 3 could respectively avert over 340,000 and 380,000 maternal deaths over 12 year period (see Table 5).

Increasing availability of transportation for deliveries referred to EmOC facilities (from 30% to 100%) led to a 1% reduction in maternal deaths, and an ICER that ranged between $2,700 and $15,800 per YLS. Similarly, increasing the recognition of referral need during skilled home deliveries and the quality of care in EmOC facilities in an isolated manner resulted in 3% reduction in maternal deaths. Providing prenatal care to all pregnant women was not an attractive single intervention with only a 3% reduction in maternal deaths (ICER = $3,400 /YLS). However, if it is assumed to increase the odds of a subsequent facility delivery, an additional 5% - 33% of maternal deaths could be averted (ICER < $800 per YLS). We assumed that 30% of facility births occurred in centers that could provide all EmOC services, and that 10% of these occurred in cEmOC facilities. Shifting all routine EmOC deliveries to cEmOC facilities was much less efficient (ICER ranged between $2,000 and $8,500 per YLS) than shifting births to bEmOC centers or birthing centers with reliable attendance and transport to cEmOC centers if needed.

To further assess for uncertainty, we varied several model inputs to limits suggested by empiric evidence (see Table 2). These inputs include the incidence and case fatality rates of direct maternal complications, as well as the effectiveness and cost of maternal interventions (cost were varied between 50% and 100% of the original inputs: see Tables 1 and 2). While varying these inputs, we improved the coverage of effective interventions (i.e. we reduced the unmet need for contraception, increased access to safe abortion and post-abortion care, as well as access to optimal intrapartum and postpartum care). In all instances, the increasing availability of these interventions was cost effective (with ICERs less than $550 per YLS). Additionally, while there were significant differences in outcomes (i.e. the absolute number of maternal deaths, MMR, lifetime risk or maternal death and proportionate mortality risk: see the Additional file 1), the relative impacts of the interventions were constant (except in instances where the effectiveness of the interventions were altered: see Table 6). Nevertheless, the cost-effectiveness (CE) ratios, the ICERs and the predicted reduction in maternal deaths were largely robust (see Table 6).