CHAPTER THREE: METHODOLOGY

Introduction

In this chapter, the methodology of this study is presented. This chapter also aims to justify the research design. The theoretical and practical concerns of using a population-based approach and the indirect sisterhood method in particular, are explored. This chapter also discusses the data collection procedure involving staffing and recruitment of data collectors, team leaders and supervisors; the training of the research team and how data collection was conducted in the field. The survey time frame was eight months, which was February – September 2005. Data collection involved the following procedures: staffing and training and the actual collection of data

3.1. Study method

The methodology for this study is survey. “A survey is a method of collecting information from people about their ideas, feelings, health, plans, beliefs, and social, educational, and financial background” (Fink and Kosecoff 1998, p: 11). To calculate maternal mortality rates from survey data, two important and innovative methods are currently being used and they are the indirect and direct sisterhood method (Stecklov 1995). The Indirect Sisterhood method was developed by Graham et al in the late 1980’s (Graham et al 1989). It provides a method for calculating the indexes of maternal mortality in countries or regions where data on vital events are not routinely and reliably collected (Hanley, Hagen and Shiferaw 1996). The Direct Sisterhood method was a variant of the Indirect (Ruternberg and Sullivan 1991).

3.2. The Sisterhood Method

The sisterhood method is an indirect technique for deriving population-based estimates of maternal mortality. This method is often recommended by the WHO and UNICEF especially for countries with inadequate registration system and low income resources (WHO 1996). The sisterhood method is based on a rationale that is similar to of the “sibling-survivorship” method. The sibling-survivorship method starts from the proportion of brothers or sisters already dead among survivors currently age x to estimate the probability of dying between birth and age x. Adopting this method for maternal mortality, identified some issues: the necessity of the respondents knowing the cause of death; all maternal deaths had to have occurred between ages [Alpha] and [Beta] (the minimum and maximum ages of childbearing); the most widely used index – the Maternal Mortality Ratio (MMR) – having a fertility component (Garenne and Friedberg 1997).

The sisterhood method is based on answers to four questions:

• How many sisters have you ever had who ever married (or who survived until age [Alpha])?

• How many are dead?

• How many are alive?

• How many died while they were pregnant, during delivery, or within six weeks after delivery (that is, died of maternal causes)?

Source: Graham et al 1989

The sisterhood method of indirectly estimating the MMR was developed because of the difficulty and expense in getting such data in other ways, the relative ease of data collection, the relatively small sample size needed, and the relative ease of calculation (Graham et al 1989). This method has become an important tool in developing countries and has been widely adopted and adapted (WHO 1997c). The original sisterhood method is indirect, this is because the MMR is not measured directly but derived mathematically from information provided by respondents about their sisters (Graham et al 1989).

The second method which is an adaptation of the sisterhood method is known as the direct method (Ruternberg and Sullivan 1991). This is a direct method of estimation that unlike the indirect does not require model fertility and mortality schedules (Stecklov 1995). The direct sisterhood method has been routinely added to many surveys including the Demographic and Health Surveys (DHS) (WHO 1996). The direct method relies on the same basic principle as the indirect that is the use of information reported by siblings to expand the sample size and to compensate for the fact that the deceased are not available for interview (Graham et al 1989). However, more data are collected on the circumstances and timing of the sister’s death, which allows a more direct calculation of the MMR. The direct method requires a larger sample size, a larger number of questions, and a more difficult calculation than the indirect method (Rutenberg and Sullivan 1991). However, (table 9) due to the larger sample size, the direct method can be modified to obtain an estimate for a more recent period, and the greater number of questions provides internal validity checks (WHO 1997c).

3.2.1. Justification for the indirect sisterhood method

This study was conducted using the indirect sisterhood method. Comprehensive vital registration systems able to capture a reasonable proportion of maternal deaths are only available in very few developing countries (WHO 1997c). Where the vital registration system is inadequate, population-based surveys have to be used to estimate maternal mortality.

major difficulties to using the population based approaches have been identified (WHO 1997c; LeCoeur, Pictet, M’Pele and Lallemant 1998). These include the fact that maternal mortality is a rare event. It has been identified that even in places where levels of maternal mortality are high, the actual number of maternal deaths is likely to be relatively small because maternal deaths are rarer than infant deaths for a fixed reference period (WHO 1997c). This thus implies that large sample sizes would be needed to gather reliable results. This large sample size may involve visiting 200,000 households, which is an impossibly large number in any setting and totally unrealistic in small countries (Kwast 1985; WHO 1997c; LeCoeur et al 1998). This method is also totally unrealistic for a PhD that has to be completed within a specific time frame and has limited financial resources.

To overcome this problem of large sample size, an alternative and more efficient technique was developed. The sisterhood method is one of such approach (Graham et al 1989) and it was specifically designed to overcome the problem of large sample sizes. This is done by obtaining information from respondents during interview about the survival of all their adult sisters (WHO 1997c).

According to the WHO, many countries have used the sisterhood method during the past few years and they include: Gambia (Graham et al 1989), The Republic of Djibouti (David et al 1991), Malawi (Chiphangwi et al 1992), Niger (De Groof et al 1993), Ethiopia (Shiferaw and Tessema 1993), Zimbabwe (Oosterhuis 1993), Indonesia (Wirawan and Linnan 1994), Bangladesh (Shahidullah 1995) Kenya (Mace and Sear 1996), northern Nicaragua (Danel et al 1996), Tanzania (Olsen et al. 2000; Font et al 2000; Mbaruku, Vork, Vyagusa, Mwakipiti, and Roosmalen 2003), Uganda (Orach 2000), Ghana (Smith et al 2001) and Swaziland (Lech and Zwane 2002). However, before the indirect sisterhood method can be used to measure maternal mortality a number of issues must be taken into account (Graham et al 1989; WHO 1997c; Olsen et al 2000) these are:

That the sisterhood method is only appropriate when:

• there is no reliable estimate of the level of maternal mortality;

• an approximate level of maternal mortality is needed for advocacy purposes and to draw attention to the problem;

• resources do not permit any other approach for measuring maternal mortality in the immediate term; and

• a starting point is needed for more detailed follow-up of maternal deaths identified during the recent past.

When there is no reliable estimate of the level of maternal mortality

In Nigeria, maternal mortality ratio (MMR) estimates from nationaland international sources differ substantially. MMR values rangefrom 740 to 1500 per 100 000 live births for the last decade (WHO 2003; UNFPA 2000). The 1999 Multi Indicator Cluster Survey, estimated a maternal mortality ratio (MMR) of 704 deaths per 100,000 live births for a period of six to twelve years preceding the survey (Federal Office of Statistics (FOS) and United Nations Children Fund (UNICEF) 2000). This survey found a significantly higher rural than urban MMR (828 versus 531), and considerable variance across regions, ranging from 165 in the Southwest to 1549 in the Northeast (National-Planning-Commission (NPLC) and UNICEF 2004; Federal Office of Statistics (FOS) and United Nations Children Fund(UNICEF) 2000).

An analysis of maternal death at the University of Ilorin Teaching Hospital (U.I.T.H.) Ilorin over a 12-year period (1972-1983) identified 624 deaths making a maternal mortality rate of 4.50 per 1000 births (Adetoro 1987). The MMR in Ahmadu Bello Teaching Hospital (ABUTH) Kaduna was 652 per 100 000, and there was no consistency in the yearly trend, which varied between 321 and 730 per 100,000 live births from 1990 to 2000 (Onwuhafua et al 2000). Given the lack of consistentestimates to aid health policy and district management, thisstudy reports the life time mortality risk and MMR for Ibadan North and Ido Local Government areas obtained with the sisterhood method, as well as themain causes of maternal deaths.

When an approximate level of maternal mortality is needed for advocacy purposes and to draw attention to the problem

Worldwide, Nigeria has the 47th highest gross domestic product (GDP) and remains the world’s 8th largest exporter of petroleum and Sri Lanka on the other hand, is 78th on the GDP list (Weeks 2007). When the MMR of the two countries are compared, Nigeria has an MMR of 800 per 100,000 live births but Sri Lanka has an MMR of only 92 (WHO, 2004a).

Reducing maternal mortality and morbidity emerged as a global priority two decades ago, yet levels of maternal death have not appeared to have declined significantly in the developing world (Thompson 1999). Although knowledge of medical and public health interventions is substantial, and most maternal deaths are preventable (Fathalla 2007), half a million women die each year from complications of pregnancy and childbirth. Another 15 million suffer life-threatening disabilities (WHO 2004a). Maternal mortality is an enormous tragedy worldwide, hence ought to attract a lot of attention. However, political leaders are burdened with thousands of issues, multiple political pressures, and limited resources (Thompson 1999; Weeks 2007; Shiffman and Okonofua 2007). To change that, maternal mortality ought to be seen not just as a moral or a medical issue but as a political issue – one that requires good research that allows advocates to be strategic in affecting policy (Weeks 2007; Shiffman and Okonofua 2007). In 2004 at the tenth Anniversary meeting to mark the International Conference on Population Development political will was cited as the missing ingredient in global reduction of maternal deaths (International Conference on Population Development 2004).

Compared to other health concerns, maternal mortality reduction is given low priority on government health agendas and sometimes it might be missing altogether (Shiffman et al 2006; International Conference on Population Development 2004; Shiffman and Okonofua 2007). Shiffman et al (2006) identified the key factors that must be available in a country to generate political will. These range from credible evidence of a significant problem to a united network of advocates who have identified a clear set of national policy priorities. Using available data, plan for advocacy can be developed, and a sensitisation campaign organised to inform policymakers and community members about the devastating impact that not putting additional resource toward improving maternal and health care would have on the country’s economy.

When a starting point is needed for more detailed follow-up of maternal deaths identified during the recent past.

According to WHO (1997c), the original indirect method is relatively simple to implement. This simplicity is highly essential for the Researcher since to the knowledge of the Researcher, this study is the first community based study on maternal mortality in Nigeria. This is the first time also that the indirect sisterhood method would be used in Nigeria to obtain a population based estimates of maternal mortality. This meant that there is no reference point for the investigator, so a simple tool will be easier for her to manage to prevent confusion.

The indirect sisterhood method also involves only four simple questions. This made training and supervision of data collectors relatively straightforward (WHO 1997c). The questions for the direct sisterhood method are more complex and time-consuming to administer (Rutenberg and Sullivan 1991). Thus requiring additional efforts directed to the training and supervision of data collectors to prevent misreporting. Furthermore, as stated earlier, this study is a PhD project which had a time frame and required the use of a less complex and a less time- consuming tool, if the project was to be concluded in the specified three years.

In addition, the indirect sisterhood method is practical to add on to comparatively small scale surveys (WHO 1997c; Graham et al 1989; Wirawan and Linnan 1994; Hanley et al 1996). Although the direct sisterhood questions can also be added to an ongoing survey it requires more time than the four questions of the original method. This may have financial implication to a resource-limited country, such as Nigeria. The direct sisterhood method requires larger sample sizes and is, therefore, more expensive to implement (WHO 1997c).

Table 9: Summary of strengths and weaknesses of original indirect and direct variants of the sisterhood method

Source: WHO (1997c). The sisterhood Method for estimating maternal mortality: Guidance for potential users. WHO: Geneva

3.2.2. Calculating MMR in Indirect Sisterhood Method

The Maternal Mortality Ratio (MMR) is a measure of the risk of death directly related to the pregnancy once a woman has become pregnant. It is usually measured in terms of maternal deaths per 100,000 live births. The maternal mortality ratio, which is obtained by dividing the age-standardized maternal mortality rate by the age-standardized general fertility rate, is often considered a more useful measure of maternal mortality since it measures the obstetric risk associated with each live birth.

It is calculated by dividing the number of maternal deaths regardless of pregnancy outcome (ectopic pregnancy, fetal deaths and live births) by the number of live births in a defined population (WHO 1994). Maternal mortality rate on the other handrefers to the number of maternal deaths in a period (usually a year) per 100,000 women of reproductive age (usually defined as aged 15-44 or 15- 49). This indicator takes into account both the risk of becoming pregnant and the risk of dying for reasons related to maternal complications during pregnancy (WHO 1994).

The calculations underlying the measurement of the MMR by the Indirect Sisterhood Method are described by Graham et al. In the Indirect Sisterhood Method, respondents aged 15-49 years old are grouped into five year age groups and this provides the number of sisters per age group. These counts are only for those sisters who have entered the reproductive period (>=15 years). Since the number of sisters who have entered the reproductive period reported by respondents in the younger age groups will exclude those sisters who are yet to reach reproductive age; a compensation is made to arrive at the expected number. The expected number of sisters reaching reproductive age for respondents in the younger age groups (age 15-19 and age 20-24) is calculated by multiplying the number of respondents by the mean number of sisters among respondents aged 25-49. This approximation is based on the assumption that the mean number of sisters per respondent is the same in each age group of respondents.

The total number of sisters who had died irrespective of the cause for respondents in each age group is calculated as well as the total number of sisters who died from maternal causes for respondents in each age group. The number of deaths from all causes across the age groups of respondents is summed up to give an overall total number of dead sisters. Summation of the number of maternal deaths across the age groups of respondents is done to provide an overall total of maternal deaths of sisters.

The total number of women-years of exposure to maternal mortality is calculated over the entire reproductive period. This is also called the sisters units of exposure (Graham et al 1989). This is calculated for each age group of respondent by multiplying the total number of sisters of respondents reaching reproductive age by an adjustment factor. To calculate the overall total of women-years of exposure, the number of women-years of exposure to maternal mortality across the age groups of respondents is summed up.

To calculate the probability of maternal death, the total number of maternal deaths is divided by the total number of women-years of exposure to maternal mortality for each age group. This is the lifetime risk of dying from maternal causes for a sister within the age group of the respondent.

To know the proportion dying from maternal causes within the age group of respondents, the total number of maternal deaths is divided by the total number of dead sisters within each age group. This is the proportion of maternal deaths in females of reproductive age (PMDF) within the age group of the respondent.

The overall total number of maternal deaths is divided by the overall total number of women-years of exposure to maternal mortality across all age groups of respondents from 15-19 to 45-49 to give a probability Q, which is the overall lifetime risk of dying from maternal causes for a sister of a respondent. Divide the overall total number of maternal deaths by the overall total number of dead sisters to give the overall proportion of maternal deaths in females of reproductive age (PMDF).

To calculate MMR, Graham et al (1989) used the formula 1 – [(1 – Q) ^1/TFR], where Q is the overall lifetime risk of dying from maternal causes and TFR is the Total Fertility Rate. When this number is multiplied by 100,000, it gives the estimated number of maternal deaths per 100,000 live births. To correct for pregnancies not ending in a live birth using the suggestion in Hanley et al (1996), the formula 1 – [(1 – Q)^1/1.2TFR] is used, where Q is the overall lifetime risk of dying from maternal causes and TFR is the Total Fertility Rate.

3.2.2.1. Adjustment factors

Adjustment factors were derived by Graham et al (1989) from a theoretical consideration of age patterns of maternal mortality, based on a model describing the relationship between the proportion of sisters dead by the age of a respondent and the lifetime risk of maternal death. Graham et al (1989) give a table of adjustment factors, which are as shown in Table 10 for this study.

Table 10: Adjustment factors for estimating women-years of exposure to maternal mortality by age group of respondent

Source: Graham et al (1989)

3. 3. Study methodology

This study was divided into four main parts:

• Identification of the total live births in the study settings to serve as denominator

• Identifications of cases of maternal deaths based on vital statistics and hospital records

• Identification and measurement of maternal deaths based on community survey using indirect sisterhood method; and

• Investigations of the causes of maternal death.

3.3.1. Study design

This was a retrospective study. Information was obtained directly from individuals at the community level using structured interviews to obtain data on maternal mortality. This was to enable a comparison to be made between the primary data collected at the community level and the secondary data available at the Ministry of Health to assess the adequacy and the quality of registration of maternal death in the state.

3.3.2. The Target Population:

The target population of this survey was all adult females and males over age 15 years in the study setting. This means that the subjects for this study consisted of one female or male from each household who was between the ages of 15-49.

3.3.3. Sampling design

A multi-stage sampling with stratification and clustering was considered appropriate for this study. Sampling is the process of selecting a portion of the population to represent the entire population (Polit, Beck and Hungler 2001). A multi stage sampling is an extension of cluster sampling where a hierarchy of clusters are chosen going from larger to smaller (WHO 2004b). A cluster is a naturally occurring unit or grouping within the population. It is a unit for which the administrative level has a clear, non-overlapping boundary (Polit et al 2001). Clustering was very useful for this study as it avoids having to compile exhaustive lists of every single person in the population. Clusters should be as heterogeneous as possible within and as homogenous as possible between.

Stratification on the other hand is the process by which the population is divided into subgroup (WHO 2004b). Sampling is then conducted separately within each subgroup. Strata should be as homogenous as possible within and as heterogeneous as possible between. This means that strata should be formulated in such a way that individuals belonging to a stratum should be as similar to each other as possible with respect to key variables from individuals belonging to a different stratum. Stratification was important in this study because of the existing geographical structure of the state and the fact that registration of deaths, although inadequate, is available in the state Ministry of Health, which is located in the state capital, to enable an assessment of the system of registration.

As earlier stated the sampling design for this study is a multi-stage sampling with stratification and clustering. Although the sampling unit is the individual subject, the sampling was conducted on the household level. The subjects were chosen by selecting a household and one eligible subject in the household was included in the sample.

Applying the multi stage stratified cluster sampling

• Strata: Provinces (Ibadan)

• Primary Sampling Unit (PSU): Counties (Ibadan North LGA/ IdoLGA)

• Secondary Sampling Unit (SSU): Enumeration Areas (Eight political wards)

• Elementary units: Households

• Final Units: Persons

3.3.3.1. Selection of LGA

Oyo state is divided administratively and geographically into 33 Local Government Areas (LGAs), which are subdivided into 361 political wards (Ogundeji 2002). The sample design used the Local Government Areas as strata. Two randomly selected LGAs, an urban (Ibadan North) and a rural (Ido) LGA were selected to create a basis for comparison (see appendix 2 for the map of Ibadan and Ibadan North LGA).

Ibadan North LGA has a population of about 403,513 people (FOS 2001) (see appendix 3 for number of households in Ibadan North LGA per political ward). It is an urban, mostly congested and overcrowded area where most of the inhabitants are traders, artisans and small scale farmers (see appendix 4 for street scenes, houses and waste disposal in Ibadan North LGA). This LGA has one teaching hospital, seven Primary Health Care (PHC) facilities and two general hospitals (see appendix 5 for profile of Ibadan North LGA). The Ido LGA on the other hand is randomly selected out of the nine rural LGA that surround the metropolis of Ibadan. Ido has a population of about 71,529 people (FOS 2001). This LGA has ten PHC centres and most of the inhabitants are small scale farmers (see appendix 6 for profile of Ido LGA). It is important that a rural LGA is selected so that data obtained from the study can be representative of the total population. These two study settings allow for comparison to be made about the actual number of deaths in an urban as compared with a rural setting both in Oyo state.

3.3.4 Sampling frame

The sampling frame should cover 100% of the eligible population in the surveyed country (Fowler 2001; WHO 2004b) this means that every eligible person in the country (or selected region) has a chance of being included in the survey sample. The choice of sampling frame usually depends on using some existing and reasonably up-to-date population lists, or constructing one by undertaking a census of members of households in a defined geographic area (WHO 2004b). Three characteristics of a sampling frame that a researcher should evaluate include comprehensiveness, probability of selection and efficiency (Fowler 2001).

For this study, the electoral rolls were used as a sampling frame since enrolment to vote is compulsory in Nigeria. A household register was also available and used although getting access to this information took considerable time and effort which is the norm in Nigeria when access is needed to any Government documents.

3.3.5. Sample size requirements for sisterhood method

The original work by Graham et al (1989) does not show any method for calculating sample size requirements for surveys employing the sisterhood method to estimate maternal mortality. Graham et al (1989) suggest that “for an estimate of the broad level of maternal mortality, interviews with 3,000-6,000 adults will be required, depending on the expected level of maternal mortality and the number of sisters who can be roughly expected to have reached reproductive age per adult” ( Graham et al ,1989. P: 132).

The WHO however gave a formula to calculate sample size requirement for the sisterhood method:

n = 4P (1-p)/e²where n = number of respondents; P = proportion of respondents to maternal deaths; and e = maximum error that will be tolerated in the proportion, generally 10%. p = average number of adult sisters per respondents divided by estimated lifetime risk of maternal death (WHO 1997c). This formula assumes a simple random sampling scheme. To allow for design effect of using more complex sampling schemes, such as cluster sampling, the derived number of respondents should be rounded up to at least the next 1000 (WHO 1997c). Tables 11 and 12 provide some estimates of sample size requirements for the two approaches compared with household surveys using direct estimation.

Table 11: Number of respondents needed to establish a maternal mortality ratio of 300 per 100,000 live births correct to within 20%