NATIONAL AND SUB-NATIONAL ANALYSIS OF THE HEALTH BENEFITS AND COST-EFFECTIVENESS OF STRATEGIES TO REDUCE MATERNAL MORTALITY IN AFGHANISTAN
Table of Contents Part I: Overview of Model
Part II: Overview of Model Parameterization, Calibration, Performance
Part III: Overview of Costs and Estimates
Part IV: Supplemental Results
Part V: References
Part I: Overview of Model
The best available data were synthesized using a computer-based model to assess the costs and health outcomes of different strategies to reduce disability and death due to pregnancy-related complications in Afghanistan. The model captures the natural history of pregnancy and relevant co-morbidities in an individual woman, aggregates clinical outcomes to the population or subgroup level, and reflects setting-specific epidemiology, and access to health care through factors such as infrastructure, human resources, technology, health facilities and transport. We prioritized data from Afghanistan or the South-Central Asian region to estimate initial ranges for age-specific probabilities of pregnancy, miscarriage, abortion, risk of maternal complications, and case-specific fatality and morbidity rates. Separate models were adapted to four provinces of Afghanistan ranging from urban to extremely remote and rural by superimposing data on coverage rates for prenatal care, antenatal care, family planning, facility births and skilled birth attendants [SBAs]. After integrating assumptions on the availability of transport, facilities, and quality of care, model-projected outcomes (e.g., maternal mortality ratio [MMR], total fertility rate [TFR]) are compared with available data at the national and provincial levels.
Strategies relied on improving coverage of effective interventions and providing access to key services. Interventions could be provided individually, paired, or packaged into a bundle of integrated services; phased approaches involved scaling up access to services over time. Model outcomes include clinical events (e.g., postpartum hemorrhage), aggregate population measures (e.g., life expectancy), and economic costs (e.g., average per person lifetime costs). Monte Carlo simulation was used to track the number of per-woman events such as pregnancies, live births, facility-based births, and maternal complications, allowing estimation of measures and indicators such as TFR, MMR, proportionate mortality ratio (proportion of deaths among women aged 15-49 that are pregnancy-related), and lifetime risk of maternal death.
We conducted national and sub-national analyses at the provincial level, representing a gradient of settings from rural and remote to urban, to explore alternative approaches in settings that differ according to underlying maternal risk, health and socioeconomic status, access to health providers, means of referral and transport, and availability of facilities capable of providing different levels of emergency care. The comparative performance of alternative strategies is described using the incremental cost-effectiveness ratio, defined as the additional cost of a specific strategy divided by its additional clinical benefit, compared with the next least expensive strategy. Strategies are considered “inefficient” or dominated if they were more costly and less effective, or more costly and less cost-effective, than an alternative strategy. We followed standard recommendations for economic evaluation. [WHO CHOICE] Sensitivity analyses were conducted to assess the impact of parameter uncertainty on our results.
The Global Maternal Health Policy Model is a computer-based model that simulates the natural history of pregnancy (both planned and unintended) and pregnancy- and childbirth-associated complications. This model defines health states to reflect important characteristics that affect prognosis, quality of life, and resource use. The time horizon incorporates a woman’s entire lifetime and is divided into equal time increments during which women transition from one health state to another. Non-pregnant girls enter the model and in each time period may become pregnant depending on age, use of contraception, and clinical history. (ManuscriptFigure 1) Once pregnant, women have a chance of spontaneous abortion (i.e., miscarriage), induced abortion, or continued pregnancy. A proportion of induced abortions will be unsafe (i.e., surgical or medical abortion conducted by untrained personnel). Labor and delivery may be associated with a direct complication of pregnancy (e.g., hypertensive disorders of pregnancy, obstructed labor, hemorrhage, sepsis). Case fatality rates are conditional on the type and severity of complication (e.g., moderate sepsis requiring antibiotics in facilities offering basic emergency obstetric care [bEmOC] versus severe hemorrhage requiring blood transfusion in facilities offering comprehensive emergency obstetric care [cEmOC]) and underlying comorbidity (e.g., moderate versus severe anemia). Nonfatal complications include neurological sequelae, rectovaginal fistula, severe anemia, and infertility. (Manuscript Figure 1) In addition to death from maternal complications, women face an annual risk of death from age-specific all cause mortality.
Strategies in the model to reduce maternal mortality consist of improving coverage of effective interventions, which may be provided individually or packaged as integrated services. In addition to family planning, antenatal care, and safe abortion, the model includes both intrapartum interventions that reduce the incidence of a complication (e.g., misoprostol for postpartum hemorrhage [PPH], clean delivery for sepsis) as well as those that reduce the case fatality rate through appropriate management in a referral facility (Manuscript Figure 1). The effectiveness of interventions to either reduce the incidence of complications or to reduce case fatality rates associated with complications depends, in part, on access to specific services (e.g., trained SBA) and to specific levels of facilities (e.g., cEmOC capacity for blood transfusion). Accordingly, the ultimate impact of interventions depends on several setting-specific factors. These include delivery site, presence of birth attendant, quality and type of referral facility, as well as successful referral when necessary. The model therefore explicitly considers the location of delivery, type of assistance, access to basic or comprehensive obstetrical care, and the ability to overcome a series of barriers around the timing of delivery (e.g., recognition of referral need, reliable transport, timely treatment at an appropriate facility); these factors collectively determine the health services a woman can access and the specific interventions that would be included. (Manuscript Figures 2 and 3)
Delivery setting is differentiated by provider (e.g., family member, traditional birth attendant [TBA], SBA) and by site (e.g., home versus facility). The facility categories are flexibly modeled such that particularities of the public health infrastructure in different settings (country, province, rural versus urban areas) can be accurately represented in terms capacity and cost. Facilities are categorized as (1) primary-level facilities, which may not provide all services necessary to qualify as a bEmOC facility, but could function as birthing centers with SBA staff who provide expectant management of labor, 24-hour intrapartum care, and reliable referral connections when necessary; (2) secondary facilities with bEmOC capacity, assumed to be capable of administering injectable antibiotics, oxytocics, and sedatives or anti-convulsants, performing manual removal of placenta, removal of retained products, and assisted vaginal delivery; and (3) tertiary facilities with cEmOC capacity, which are also able to provide blood transfusion, cesarean section, and management of advanced shock. (Manuscript Figure 2)
We recognize that some tertiary sites will not have a blood bank and some secondary sites may eventually be able to perform c-section; further, we recognize that in the strategies that include stepwise investments in infrastructure and facility improvements, not all facilities will be expected to be fully implemented as one of the three distinct types. However, because the costs, functions and staffing are fairly closely aligned with basic or comprehensive EmOC capacity, this simple categorization captured the most important dimensions for purpose of this analysis. Part II, Subsection B provides a stylized example of how public health facilities in Afghanistan, based upon the Afghanistan National Health Resources Assessment (ANHRA) [MSH 2002], may be superimposed on our general model framework.
This model also allows us to evaluate phased approaches that involve scaling up access to services over time; we designate such stepwise investments in infrastructure as “upgrades”. In addition to reducing unmet need for family planning and unsafe abortion, these strategies incrementally shift home births to facilities, increase skilled attendance, and improve access to, and quality of, emergency obstetrical care. For women delivering at home or in birthing centers, these strategies also improve recognition of referral need, access to transport, and expedient referral to an appropriate facility.
All models were built using TreeAge Pro 2008 (TreeAge Software Inc., Williamstown MA) and analyzed using IBM/Lenovo Dual-Core VT Pro Desktop computers running Microsoft Windows XP, using Microsoft Excel 2007 and Visual Basic for Appplications 6.5 (Microsoft Corp., Redmond WA). We used Monte Carlo simulation to generate the number of per woman events such as pregnancies, live births, facility-based births, and maternal complications. This output is useful for both calibration exercises, as well as for assessing internal consistency and projective validity of the model by generating outcomes in similar formats to clinical studies. We used first-order Monte Carlo simulation to assess first-order uncertainty and one- and two-way sensitivity analyses to assess parameter uncertainty.
Part II: Overview of Model Parameterization, Calibration, Performance
Data and Assumptions: Initial Natural History Parameters. The best available data are sought on clinical parameters governing the natural history of pregnancy. Examples of required model inputs include the age-specific probability of pregnancy, miscarriage, unsafe and safe abortion; incidence, morbidity, and case fatality rates for each maternal complication (PPH, sepsis, obstructed labor, hypertensive disorders of pregnancy); prevalence of co-morbidities (e.g., anemia).
Data and Assumptions: Intervention Effectiveness. The effectiveness of interventions to reduce the incidence of complications and/or reduce case fatality rates is from published data, and varies by complication type and severity. Initial estimates assume an intervention can be delivered appropriately, but are then modified according to several setting-specific factors. These include delivery site, presence of birth attendant, quality and type of referral facility, as well as successful referral when necessary
Data: Coverage Inputs and Selected Services. Data on coverage rates for interventions, facility births, skilled birth attendants, antenatal care, and family planning, are stratified by rural and urban status, and for validation analyses are state-specific.
Data and Assumptions: Barriers to Effective Referral.Effective referral relies on the ability to overcome three critical delays (a) recognition of referral need and willingness to be referred (by provider and delivery location); (b) expedient transfer to referral facility (determined by distance, affordability, available transport); and (c) timely treatment in an appropriate facility capable of high-quality emergency obstetrical care (e.g., 6 signal functions in bEmOC, blood transfusion and surgery in cEmOC).
Calibration Exercises. Calibration targets include the distribution of causes of maternal mortality (e.g., PPH, obstructed labor, sepsis), maternal mortality ratio (MMR), total fertility rate (TFR). After integrating assumptions about the availability of health services, model-projected estimates of MMR, TFR, and distribution of direct causes of maternal mortality are compared to empiric data. Selected uncertain parameters (such as the case fatality rates conditional on severity of complication) are varied across a pre-specified plausible range, in a systematic fashion, to ensure output is consistent with key empiric indicators.
Model Performance. Model performance is assessed by comparison of model-based projections with independent measures such as life expectancy, proportionate mortality ratio, and population-based outcomes. Projective validity of the empirically-calibrated model is further assessed by simulating four provinces, and comparing projected maternal health indicators with reported data.
Data and Assumptions
Age-specific probability of pregnancy
To estimate a fertility rate in the absence of any family planning, we used turn of the century data from Afghanistan, when contraceptive use and abortion rates were low and women’s access to modern, high quality health care extremely limited [Amowitz 2002, Bartlett 2005, CSO 2003, UNIFEM 2007]. We synthesized data on family planning, demographics (i.e., crude birth rate in Afghanistan of 48 births per 1,000 population [PRB 2005]), and assumptions related to miscarriage and stillbirths [Harlap 1980, Bartlett 2005] to approximate an average annual natural fertility rate of 31%. The model allows for age-specific inputs for fertility and use of contraception. We assumed 15% of all pregnancies end in spontaneous abortion, of which approximately one-third result in incomplete abortion requiring medical intervention [Harlap 1980, Menken 2006]. We assumed women with long-term complications such as infertility or untreated obstetric fistula did not become pregnant again. We assumed women with complications that were treated (e.g., severe anemia, surgically-repaired fistula) could become pregnant again.
Anemia, poor health, and young age at first pregnancy
Anemia is very common among Afghan women. The prevalence of anemia in Afghanistan has been estimated at 48% (for 2007) in women of reproductive age [UNIFEM 2007] and 61% (for 2006) in pregnant women [WHO 2008]. The 2000 Multiple Indicator Cluster Survey (MICS2) conducted in the Eastern region of Afghanistan found even higher rates of anemia - 71.4% among pregnant women and 88.7% among non-pregnant women. [Afghanistan MICS2 Steering Committee 2001] For the baseline calibration models representing maternal mortality in Afghanistan from 1999-2002, we use the MICS2 estimate of anemia prevalence among pregnant women. We use the WHO estimate for 2006 as the estimated anemia prevalence among pregnant women for 2007, in the “progress towards MDG5 analysis”. None of the anemia estimates could be distinguished by severity level. We assume 30% of anemic women in Afghanistan are moderately anemic, and 5% are severely anemic, leading to national prevalence estimates of severe anemia of 3.6% and moderate anemia of 21.4%. We base these assumptions on the Indian National Family Health Survey-3 (NFHS-3), in which 28.8% of anemic women in India could be classified as moderately anemic and 3.8% as severely anemic. [IIPS 2007]
We assume that anemia prevalence is not constant across Afghanistan. The 2005 Afghanistan National Risk and Vulnerability Assessment (NRVA) shows that Laghman, Kandahar and Badakhshan had much higher proportions of households in the worst food consumption category (Low dietary diversity / Very Poor Food consumption) compared to Kabul (24-33% compared to 15% in Kabul). [MRRD 2007] Since the most likely cause of anemia in Afghanistan is iron deficiency anemia resulting from too little iron in the diet [Woodruff 2002], these data suggest that the prevalence of anemia is likely to be higher in these semirural, rural, and remote rural regions with relatively worse dietary intake. We therefore assume a lower prevalence of moderate (15%) and severe (2%) anemia for Kabul, and a higher prevalence (25% moderate, 5% severe) for the provinces of Laghman, Kandahar and Badakhshan, compared to the national estimates. We allow for uncertainty surrounding these estimates and provide the ranges used in sensitivity analyses in the table below.
Estimated anemia prevalence among pregnant women nationally and for selected provinces
2 – 5
15 – 25
1.5 – 4.5
10 – 20
3.5 – 6.5
25 – 35
3.5 – 6.5
25 – 35
3.5 – 6.5
25 – 35
The relative risk (RR) of death from maternal complications has been found to be 3.51 times greater with severe anemia and 1.35 times greater with moderate anemia, compared to a woman without anemia. [Brabin 2001] We assumed severe and moderate anemia were associated with these higher relative risks of death from pregnancy- and delivery-related complications, although anemia differentially affected mortality from postpartum hemorrhage and sepsis, and complications following unsafe abortion. We conservatively assumed that severe and moderate anemia did not impact the case fatality rates of untreated obstructed labor or hypertensive disorders.
Poor general health, malnutrition, and vitamin A deficiency also contribute towards increased risk of dying from pregnancy- and delivery-related complications. Chronic malnutrition during childhood can cause stunting, leading to the development of a smaller than normal pelvis, and put a woman at increased risk of obstructed labor. [WHO 1999] According to the UNICEF Nutrition Database, between 45% and 64% of the Afghan population suffered from chronic stunting during the 2001-2003 period. [UNICEF 2006] While very limited data are available on the prevalence of vitamin A deficiency in women of reproductive age, WHO has classified Afghanistan as “Severe subclinical” vitamin A deficiency [WHO 1995]. Regression-based estimates of the prevalence of vitamin A deficiency in pregnant women in Afghanistan range from 12% to16% [WHO 2009], and a 2002 survey of women of reproductive age in Badghis province found a 4.7% prevalence of xerophthalmia (ocular manifestations common to vitamin A deficient populations) [WHO 2007a]. A study by the WHO on major risk factors associated with global and regional burden of disease found that vitamin A deficiency was associated with a 4.51 (CI 2.91-6.94) increased relative risk of all-cause maternal mortality. [Rice 2004] The authors estimate that 20% of all-cause maternal mortality can be attributed to this condition. [Rice 2004] We do not directly use the relative risk estimate within the model, as it was based on a single study conducted in Nepal, and there is a lack of further studies looking at this association. We do, however, consider the potential impacts of vitamin A deficiency, chronic malnutrition and stunting on case fatality rates during the calibration process.
Pregnancies at a young age and high parity are also known to increase the risk of death from pregnancy- or delivery-related complications. Women in Afghanistan tend to marry young and have high fertility rates, especially in rural areas. While the legal minimum age for girls to marry in Afghanistan is 16, marriage of girls under the age of 16 is common. [FIFC 2004] The 2005 NRVA report stated that the most common age of marriage was 20 years, and that 1.3% of girls aged 10 to 11 years were married. [MRRD 2007] Other sources have estimated that most girls are married before the age of 16 or 18 years. [FIFC 2004, UNIFEM 2007] For instance, UNICEF data analyzed by Tufts University showed that 16% of Afghan girls are married under the age of 15 years, and 52% under the age of 18. [FIFC 2004] Another source has estimated that 57% of girls were married before the age of 16. [UNIFEM 2007] A household survey of mothers with children under 5 years conducted in two rural districts (Karokh and Chesht-e-Sharif) of Herat province found that in both districts, approximately 50% (51% in Karokh and 46% in Chesht-e-Sharif) of women were younger than 16 years of age at the time of their first pregnancy. [Ahmed 2004] A further 43.5% and 49.0% (in each district respectively) of women were 16 to 21 years at the time of their first pregnancy. [Ahmed 2004] A Tufts University study of rural populations in a number of different provinces found an average of 8 live births per woman, with a range of 0 to 17 live births. [FIFC 2004]
While not nationally or provincially representative, and consisting only of rural households, the MICS 2003 survey [UNICEF 2006] found differences in child brides (defined as percentage of women 20-24 years of age who were married before they were 18 years old) across provinces that support the general trend of maternal indicators across the four provinces we model: