Contraceptive Prevalence Survey Research Articles

Multinomial logit models with implicit variable selection

The multinomial logit model is the most widely used model for the unordered multi-category responses. However, applications are typically restricted to the use of few predictors because in the high-dimensional case maximum likelihood estimates frequently do not exist. In this paper we are developing a boosting technique called multinomBoost that performs variable selection and fits the multinomial logit model also when predictors are high-dimensional. Since in multi-category models the effect of one predictor variable is represented by several parameters one has to distinguish between variable selection and parameter selection. A special feature of the approach is that, in contrast to existing approaches, it selects variables not parameters. The method can also distinguish between mandatory predictors and optional predictors. Moreover, it adapts to metric, binary, nominal and ordinal predictors. Regularization within the algorithm allows to include nominal and ordinal variables which have many categories. In the case of ordinal predictors the order information is used. The performance of boosting technique with respect to mean squared error, prediction error and the identification of relevant variables is investigated in a simulation study. The method is applied to the national Indonesia contraceptive prevalence survey and the identification of glass. Results are also compared with the Lasso approach which selects parameters.

Introduction to Use of health impact metrics for programmatic decision making in global health

Over the past twenty years, the international landscape for health programming in developing countries has changed dramatically, and global awareness of the consequences of poor health has grown. One result is the international commitment to the Millennium Development Goals (MDGs) that has put poverty reduction and health at the forefront of the global agenda. And, as new funding agencies and mechanisms (e.g., GFATM, GAVI, PEPFAR, PMI) appeared, attention focused more closely on the impact of their expenditures. Major donors - multilateral, bilateral and private - have embraced ‘results-based programming’. They ask whether they are getting ‘value for money’. And the commitment of national leaders to the MDGs confirms that they, as well as the donors, expect accountability for the significant expenditures now being made on health programs. Ultimately, what the global health community wants to know is which programs of service delivery can most efficiently improve health. Without strong evaluations designed to measure impact, decision-makers lack sufficient evidence to fund expansion of service delivery or behavior change programs, and doing so without data on impact might even be considered unethical [1]. But there are many constraints to measuring the impact of large, complex programs, and studies to evaluate impact are expensive and take time [2]. The global health community has instead relied almost exclusively on measurements of intervention coverage, behaviors, and levels of mortality and fertility from household surveys to assess progress. This approach has a long tradition, starting in the 1970s with the World Fertility and Contraceptive Prevalence Surveys, and continues today with the Demographic and Health Surveys and the Multiple Indicator Cluster Surveys, which are the main vehicle for measuring progress toward many of the MDGs. But from the point of implementing organizations, surveys do not provide timely information, and rarely report below regional or provincial level, when more specific data are often needed for program management. Surveys are also expensive and they leave a large gap - a black box - between program inputs and their outcomes and impact. Many implementing organizations are looking for new approaches to express the impact, equity and breadth of their programs. The papers in this Supplement provide a middle ground solution. They give us the opportunity to examine how two organizations that implement global health programs are addressing the need to assess performance, make strategic decisions about their programmatic priorities based on performance and goals, and at the same time be accountable to funders and national stakeholders. The papers provide details about how these program implementers are translating routinely-collected data on program outputs - either commodities distributed or services provided - into estimates of impact on health status, as expressed in disability-adjusted life years (DALYs) averted and on outcomes as expressed in changes in contraceptive prevalence and contribution to national indicators. The papers describe a range of organizational uses for data on impact, as well as analytic methods for examining equity in health status and program exposure that may be used to guide program development.

Demographic and health surveys: a profile

Demographic and Health Surveys (DHS) are comparable nationally representative household surveys that have been conducted in more than 85 countries worldwide since 1984. The DHS were initially designed to expand on demographic, fertility and family planning data collected in the World Fertility Surveys and Contraceptive Prevalence Surveys, and continue to provide an important resource for the monitoring of vital statistics and population health indicators in low- and middle-income countries. The DHS collect a wide range of objective and self-reported data with a strong focus on indicators of fertility, reproductive health, maternal and child health, mortality, nutrition and self-reported health behaviours among adults. Key advantages of the DHS include high response rates, national coverage, high quality interviewer training, standardized data collection procedures across countries and consistent content over time, allowing comparability across populations cross-sectionally and over time. Data from DHS facilitate epidemiological research focused on monitoring of prevalence, trends and inequalities. A variety of robust observational data analysis methods have been used, including cross-sectional designs, repeated cross-sectional designs, spatial and multilevel analyses, intra-household designs and cross-comparative analyses. In this profile, we present an overview of the DHS along with an introduction to the potential scope for these data in contributing to the field of micro- and macro-epidemiology. DHS datasets are available for researchers through MEASURE DHS at www.measuredhs.com.

RETHINKING THE PAST COMMUNITY DEVELOPMENT OF INDONESIA: HIV TRANSMISSION RISK IMPLICATIONS OF ESTIMATES OF PREMARITAL CONCEPTION

As the countries of Asia and the Pacific come to realize that they will need to confront the AIDS epidemic most are lacking information to help identify populations at risk, and most have no data with which to gauge levels of risk behavior in key segment of the general population. Using data from the 1987 Indonesian Contraceptive Prevalence Survey (The Demographic and Health Survey), we use marital month of first birth to derive estimates of premarital conceptions for marriage cohorts. The estimates produced suggest that the level of premarital sexual intercourse has been rising steadily since the late 1950's, but particularly fast since the mid to late 1970's. Rates specific for age at first marriage and the type of place where a women grew-up, as well as for education suggest that premarital intercourse is now prevalent and increasing rapidly in many segments of the country. The implications of these observations for the development and targeting intervention programs are discussed.

The Potential Impact of Community-Based Distribution Programmes on Contraceptive Uptake in Resource-Poor Settings: Evidence from Ethiopia

Community-based reproductive health services (CBRHS) have been in implementation in Ethiopia for over a decade. This paper assesses the impact of such programmes on contraceptive use. The data were obtained from the 2000 Ethiopian Demographic and Health Survey (DHS) and the 2003 Contraceptive Prevalence Survey (CPS). While the 2000 DHS contained a nationally representative random sample of women of reproductive age, the 2003 CPS focused on areas with active CBRHS programmes. The findings indicate that expanding the current CBRHS programme has the potential of doubling the current national contraceptive prevalence rate.

The potential impact of community-based distribution programmes on contraceptive uptake in resource-poor settings: evidence from Ethiopia.

Community-based reproductive health services (CBRHS) have been in implementation in Ethiopia for over a decade. This paper assesses the impact of such programmes on contraceptive use. The data were obtained from the 2000 Ethiopian Demographic and Health Survey (DHS) and the 2003 Contraceptive Prevalence Survey (CPS). While the 2000 DHS contained a nationally representative random sample of women of reproductive age, the 2003 CPS focused on areas with active CBRHS programmes. The findings indicate that expanding the current CBRHS programme has the potential of doubling the current national contraceptive prevalence rate.

Using modular neural networks for business decisions

Understanding large amounts of information and efficiently using that information in improved decision making has become increasingly challenging as businesses collect terabytes of data. Businesses have turned to emerging technology including neural networks, symbolic learning, and genetic algorithms. In the current study, four classification methods were compared using results from an Indonesian contraceptive‐method preference survey. The four methods are linear discriminant analysis, quadratic discriminant analysis, backpropagation neural networks, and modular neural networks. The modular neural network is a more complex and less frequently used neural network model. This comparative study gives insight into its performance on classifying observations from a challenging data set, the 1987 National Indonesia Contraceptive Prevalence Survey.

Prevalence of contraceptive use in a periurban area of Sousse, Tunisia

A contraceptive prevalence survey was conducted among 1277 women of reproductive age, to explore knowledge of and attitudes to contraceptive use and study factors influencing contraceptive use. We found that family planning methods were generally well known. The prevalence of modern contraceptive use was 53.6%, IUDs being the most commonly used followed by sterilization and pills. The public sector was the main source for obtaining contraceptives although pills and condoms tended to be obtained from the private sector. There was a significant correlation between contraceptive use and urban residence.

History and current status of reproductive health surveys at CDC

Since the development of immunization surveys in the 1950s to evaluate poliomyelitis immunization programs, the Centers for Disease Control and Prevention (CDC) has traditionally used populationbased surveys to evaluate public health programs.1 The use of population-based surveys continued overseas in the smallpox eradication program during the late 1960s and 1970s.2,3 Family planning epidemiology was begun at CDC in the late 1960s, first as an activity, then as the Family Planning Evaluation Division, and later as part of the Division of Reproductive Health (DRH). At first, service statistics systems were developed for local and state health departments.4,5 It quickly became apparent that population-based household surveys would be needed to obtain data on family planning users in the private sector and nonusers.6–9 Jack Smith, chief of the statistics activity, spearheaded the early efforts to improve service statistics systems and to develop populationbased surveys for evaluating family planning programs. In 1979, Smith was the lead researcher for a joint household survey with the Mexican Ministry of Health that was carried out along the U.S.–Mexico border to obtain family planning data on non-Hispanics and Hispanics.10–13 Since that effort, with the development of the random-digit dialing method, statewide surveys on reproductive health have been carried out by telephone in Idaho, Arizona, New York, Hawaii, and Georgia.14,15 The switch to telephone as the survey mode has been due largely to cost considerations. Parallel to the development of family planning survey methodologies in the United States, the Family Planning Evaluation Division, in collaboration with the U.S. Agency for International Development, developed the Contraceptive Prevalence Survey (CPS) for use in developing countries.16 The first of these surveys was conducted in El Salvador in 1975.17 The CPS is a regional or national probability household-based sample survey designed primarily for managing and evaluating family planning programs. This survey collects information on knowledge and use of contraception in relation to desire for more children, choice of contraceptive methods, availability of family planning services, and other program-related factors. The survey provides rapid feedback to program administrators on the use of family planning services and on program success in serving women at risk of unintended pregnancy. To ensure representativeness, the CPS (and subsequent surveys described below) uses a three-stage sampling design. The first stage is a selection of census sectors with probability proportional to size. Within the selected census sectors, a random sample of households is then chosen. In the third stage of the sample design, one woman of reproductive age is selected from each household. In a standardized interview, each woman is asked about her knowledge, use, and attitudes concerning contraceptive methods and sources of contraceptive supplies and services. Results are weighted to compensate for geographic oversampling or undersampling, the selection of one woman per household, and nonresponse differentials. Contraceptive Prevalence Surveys do more than reveal the extent of contraceptive use. They also help administrators evaluate family planning programs, check service statistics, and identify groups of women in need of family planning services. Each CPS provides the rapid feedback necessary to evaluate and improve family planning information and service delivery programs. Because questionnaires are short and usually only several thousand women are interviewed, the surveys can be repeated every few years to gauge the impact of new or reoriented programs undertaken on the basis of earlier findings. The CPS data are especially useful in determining how many women are at risk of unintended pregnancy; that is, how many do not want to become pregnant but may do so because they are fecund, sexually active, and not using contraception. Included among these are women whom family planning services have not yet reached. Identifying these women is of prime importance to program administrators. Around 1980, the CPS evolved into the Family Planning and Maternal Child Health Survey (FP-MCH), which expanded the content of the CPS to include data From the Behavioral Epidemiology and Demographic Research Branch, Division of Reproductive Health, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, Atlanta, Georgia. Address correspondence and reprint requests to: L. Morris, MD, CDC-MS K-35, Behavioral Epidemiology and Demographic Research Branch, 1600 Clifton Road, Atlanta, GA 30333. E-mail: lmorris@ cdc.gov.

Trends in the Demand for Family Limitation in Developing Countries

Data from 108 Demographic and Health Surveys World Fertility Surveys and Contraceptive Prevalence Surveys were assembled for 41 developing countries covering the period extending from the mid-1970s to the late 1990s. The percentage of women who want no more children has risen slowly but steadily in sub-Saharan Africa since the 1970s having reached a level of 20-40% in many countries by the late 1990s. Yet overall levels remain far below those seen in Asia and in North Africa where the level of demand for limiting births clusters in the 40-60% range. The proportion of women wanting to stop childbearing is also high in Latin America and shows more evidence of leveling off than in Asia. Unmet need for the means to limit births is increasing fairly uniformly for most sub-Saharan African countries; in contrast in Asia and North Africa and Latin America and the Caribbean it is generally declining with the adoption of contraceptive use. While the evidence indicates that most women in sub-Saharan Africa who practice contraception do so to space rather than to limit births trend data suggest that the proportion of users practicing contraception to limit births has been increasing in recent years; in some countries this proportion approaches half of all method use and is higher than expected elsewhere. In contrast there has been little change in this balance in Asia and North Africa and in Latin America and the Caribbean with the great majority of users in both regions seeking to limit rather than space births. While demand for contraception is increasing throughout the developing world most of the demand in Asia and North Africa and in Latin America is already being met while much of the demand in sub-Saharan Africa is not. In both Asia and Latin America where contraceptive use is already high providers need to gear their services toward helping clients to continue use and to improve the effectiveness of their contraceptive practice. In sub-Saharan Africa where use is low programs must aim to encourage adoption of modern methods. (authors)

In Pakistan, Contraceptive Prevalence is Increasing, but Fertility Remains High

This paper presents the findings of the Pakistan Contraceptive Prevalence Survey conducted in 1994-95. Data on marriage patterns contraceptive knowledge and use fertility and fertility preferences and family planning services were taken from a sample of 7922 married women aged 15-49. Overall it is noted that contraceptive prevalence has reached 18% in the country an increase of 50% above the level of 12% recorded in 1991. In addition the ideal family size has been found to have fallen from 4.9 to 3.4 children over the last decade. However the total fertility rate recorded in the 1995 survey was 5.6 live births per woman. Women aged 35 or older had borne an average of 6.8 children of whom 5.7 were living at the time of the survey. If such childbearing trends were to continue it is projected that Pakistani women will have an average of 5.6 births. Moreover it is found that child mortality remains a common experience in the country in which 1 out of the average 6.7 children born to each woman aged 35 or older will die.

Partitioning the Effect of Infant and Child Death on Subsequent Fertility: An Exploration in Bangladesh

A method of partitioning the fertility impact of infant and child death into two components – a physiological and a behavioural effect – is proposed by use of the Cox hazard model with three dummy variables that indicate the time of child death and the status of breastfeeding with reference to the return of menstruation postpartum. An application of the method to the 1991 Bangladesh Contraceptive Prevalence Survey data suggests that the effect from the physiological mechanism outweighed the effect from the behaviourial mechanism (the former effect raising the hazard of an additional birth by nearly 90 per cent). It appears that the effect of a child death declined over time and an early cessation of breastfeeding was not the sole cause for invoking the physiological mechanism. The risk of childbirth rose sharply among the educated if their children died, although the main effect of education itself was to reduce the risk.

Socioeconomic development, status of women, family planning, and fertility in Bangladesh: a district level analysis.

In this paper we examine the effects of socioeconomic development, the status of women, and family planning on fertility and the extent to which these effects vary among the nineteen districts of Bangladesh. The 1983 and 1991 Bangladesh Contraceptive Prevalence Survey data are used to examine the effects of these factors on differences in contraceptive use among currently married women aged 15-49. The proportion of currently married women aged 15-49 using contraception was 46.3 per cent in 1991, a considerable increase from 26.5 per cent in 1983. Contraceptive use rates for all the districts increased substantially over the period between 1983 and 1991 and the gap between Dhaka's rate and those of other districts was narrower in 1991 than in 1983. An analytical model composed of socioeconomic development, status of women, and family planning variables is tested using logistic regression. The logit model is used to evaluate the effects of a selected group of variables on the probability of using contraceptive methods. The analysis demonstrates clearly that socioeconomic development and women's status significantly impact the use of contraceptive methods in Bangladesh. The results also suggest that better-educated, employed women are more likely to use contraception than those who have little or no formal education and who are not employed. In 1983, rural-urban differences in contraceptive use were significant, but in 1991 area of residence was not significant, which suggests that family planning programs may have played an important role in providing contraceptive information and technology to rural areas. Our analysis also suggests that family planning programs operated more efficiently in some districts than in others, and regional differences remained important in 1991 as they were in 1983. Decomposition analysis suggests that shifts in population structure favored increased contraceptive use in Bangladesh. This analysis also indicates that change in rate is also important, contributing about 21 per cent of the overall increase in contraceptive use.

HIV/AIDS and its risk factors in Pakistan

Pakistan currently reports a low prevalence of HIV/AIDS. Concern over the potential for an expanded HIV epidemic in the country is mainly due to segments of the Pakistani population who engage in high-risk practices a low level of public knowledge about HIV/AIDS dangerous blood transfusion and inoculation practices and Pakistans proximity to India a country which has experienced a rapid increase in the prevalence of HIV/AIDS. The limited data currently available on HIV/AIDS prevalence in Pakistan are presented. By November 1996 64 cases of AIDS had been reported to Pakistans National AIDS Program. However the World Health Organization estimated that by the end of 1994 there were really 40000 people infected with HIV. Data from point-prevalence studies suggest that foreign and Pakistani nationals who have either worked abroad or as seafarers may comprise a sizeable proportion of the recognized HIV/AIDS cases. The studies also documented cases of HIV infection among people with multiple sex partners blood transfusion recipients and prisoners. Heterosexual contact appears to be the dominant mode of viral transmission. Lifestyle risk factors and health care practices which contribute to HIV transmission are described.

Infant feeding patterns on Mauritius Island, 1991

The 1991 Mauritius Contraceptive Prevalence Survey (CPS) included a special module on infant feeding patterns in Mauritius. Since 1985, when a similar CPS was conducted, the incidence of breast-feeding has fallen from 86% to 72%. The duration of any breast-feeding among those breast-fed remained constant at 13.6 months. The module allowed for an assessment of the World Health Organization (WHO) breast-feeding indicators on exclusive breast-feeding, timely complementary feeding and continued breast-feeding. Only 16% of infants 0–3 months old are exclusively breast-fed; only 29% of infants 6–9 months old receive breast milk and complementary foods and only 27% of children 12–15 months are still breast-fed. These patterns of limited breast-feeding and early supplementation may signal future declines in breast-feeding for other African and Asian countries. Published by Elsevier Science Ltd

Lower Egyptian Fertility Linked to Later Marriage, Increased Method Use

This paper reports on several analyses of contraceptive use trends and fertility determinants in Egypt. Analysis is based primarily on data from the 1992 Egypt Demographic and Health Survey (EDHS). One study used data from the 1980 Egypt Fertility Survey the 1984 Egypt Contraceptive Prevalence Survey and the 1988 and 1992 EDHS. This study documents the doubling of contraceptive use among married women during 1982-90. The second study uses birth history from the 1992 EDHS in the analysis of fertility decline and its determinants. The first study indicates that contraceptive prevalence rose from 24% in 1980 to 47% in 1992. Slower prevalence rate increases occurred in high-prevalence regions of Lower Egypt and Urban Governorates. A rapid increase in prevalence rate occurred in rural areas of Upper Egypt. There was a shift in choice of methods. IUD use increased and use of oral contraceptives decreased slightly. Increases in use occurred in all age groups. Levels of use were lowest among adolescents and women in their early 20s. 32% of women with 1 child used contraception in 1992 compared to only 16% in 1980. 59% of women with 3 children used contraception in 1992 compared to only 29% in 1980. Method use was highest among the most educated women. Findings from a logistic model indicated that urban women were 1.5 times more likely to use contraception. Women living in Lower Egypt or Urban Governorates were twice as likely to use contraception as women in Upper Egypt. In the second study researchers found that delayed age at marriage changes in postpartum insusceptibility and an increase in the proportion of reproductive age women contributed to a 13% decline in fertility during 1988-92. Women born between 1963 and 1977 had much lower fertility than women born earlier. The number of women with a second birth within 5 years of the first declined between 1988-92. Both studies confirmed a fertility transition.

Differential pattern of birth intervals in Bangladesh

Findings from this 1989 Bangladesh nationally representative Contraceptive Prevalence Survey reveal that the survival status of the index child was a strong factor in explaining differences in birth spacing. The effect of mortality of the index child at the infant stage was shortening of the birth interval. The effect was weaker for child mortality. The age of the mother at child birth had the next strongest effect on birth spacing. Increases in maternal age at birth were associated with increases in birth interval. Explanatory variables that had little influence on birth spacing were maternal educational status, sex of the index child, residence, and birth order. The data pertain to open and closed birth intervals during 1984-89. There were 11,034 birth intervals, of which 36% were closed birth intervals. The analysis included the distribution of closed subsequent birth intervals among different subgroups and the study of the differences. The product-limit approach for estimating the survival function was used in the analysis of both mean closed and open birth intervals. About 75% of birth intervals among women aged under 35 years ranged 13-24 and 25-36 months. About 66% of birth intervals among women aged over 35 years lasted 13-35 months; 19-24% lasted over 3 years. About 25% of births occurred within 12 months after women experienced the death of an infant. Only 5% of births occurred within 12 months among women without a prior index child death. About 50% of women had a subsequent birth within 13-24 months after the death of the index child. The product-limit estimates confirmed that the death of an index child during infancy was a strong determinant of a short subsequent birth interval.

Contributions of Alexander D. Langmuir to the epidemiologic study of population change and family planning.

This article traces the influence of Dr. Alexander Langmuir on the application of epidemiological methods to the study of population change and family planning. In the 1960s, global population growth had become a major issue but the role of public health professionals in meeting the challenge was unclear. By 1967, Langmuir had assigned eight Epidemic Intelligence Service (EIS) Officers to family planning (FP) evaluation, capitalizing on support from the Centers for Disease Control (CDC) and securing funding from many sources. In 1970, the year Title X of the Public Health Service Act designated funds for population activities with none of the funds designated for the CDC, Langmuir retired. EIS officers trained by Langmuir issued surveillance reports on abortion (in 1969) and sexual sterilization (the report defined the risk of morbidity from sterilization for women) and conducted studies of the safety and effectiveness of IUDs, (which led to association of the Dalkon Shield with high rates of pelvic inflammatory disease and with septic spontaneous abortion deaths) and the association of oral contraceptive use with various cancers. Epidemiologic evaluation procedures were applied to FP programs and evaluation training, to adolescent pregnancy, to international studies, to international contraceptive prevalence surveys, and to demographic studies. By mid-1993, 118 EIS officers had worked in FP evaluation. These EIS officers and alumni continue to work in the fields of reproductive health or public health throughout the world.

Estimating Contraceptive Needs from Trends in Method Mix in Developing Countries

To enable donor agencies and family planning program managers to estimate future needs for contraceptive supplies in developing countries data drawn from 106 Demographic and Health Surveys and Contraceptive Prevalence Surveys conducted in 35 Third World countries between 1974 and 1992 were reviewed. At least 2 surveys conducted a minimum of 2 years apart were used to permit calculation of changes in total and method-specific prevalence and annual rates of change. At the time of the most recent survey 44% of respondents were using contraception and 35.5% were using a modern method; these rates were 31.6% and 24.0% respectively at the initial survey. Total contraceptive prevalence increased at an annual rate of 5% while modern method use increased by 6%. Sterilization acceptance increased by 8% each year while use of the pill IUD and condom registered annual increases under 2%. In 30 of the 33 countries for which data were available the prevalence of sterilization grew at a substantially faster rate than that of other modern methods. Overall these findings suggest a need for caution against unrealistically high forecasts of the demand for supply-based contraceptive methods. On the other hand the population group eligible for contraception is increasing by 2-3% each year. It remains uncertain whether the increased availability of injectables and implants will result in greater overall contraceptive prevalence or simply substitute for pill IUD and condom use.

Reproductive change in Sri Lanka: Analysis of intermediate variables, 1982 and 1987

This study examines the intermediate determinants of fertility in Sri Lanka by making use of the data collected in the 1982 Sri Lanka Contraceptive Prevalence Survey and the 1987 Sri Lanka Demographic and Health Survey. The analysis shows that the most important inhibitor of potential fertility is deliberate control. The marital structure of the population is also an important fertility-inhibitor, but lactational infecundability is increasingly becoming an unimportant contributor. The findings show the success of the family planning program in Sri Lanka, which propelled fertility to a substantial lower level. Achievement of the replacement level fertility by the turn of the century, set by the Sri Lankan government, would largely depend on the efforts to increase the quality and quantity of contraceptive use and the duration of breastfeeding.