National Academies Press: OpenBook

Leveraging Longitudinal Data in Developing Countries: Report of a Workshop (2002)

Chapter: Report: Leveraging Longitudinal Data in Developing Countries

« Previous: Front Matter
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

PART I
REPORT

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
This page in the original is blank.
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Leveraging Longitudinal Data in Developing Countries

INTRODUCTION

Longitudinal data collection and analysis are critical to social, demographic, and health research, policy, and practice. They are regularly used to address questions of demographic and health trends, policy and program evaluation, and causality. Panel studies, cohort studies, and longitudinal community studies have proved particularly important in developing countries that lack vital registration systems and comprehensive sources of information on the demographic and health situation of their populations. Research using data from such studies has led to scientific advances and improvements in the well-being of individuals in developing countries. Yet questions remain about the usefulness of these studies relative to their expense (and relative to cross-sectional surveys) and about the appropriate choice of alternative longitudinal strategies in different contexts.

For these reasons, the Committee on Population convened a workshop to examine the comparative strengths and weaknesses of various longitudinal approaches in addressing demographic and health questions in developing countries and to consider ways to strengthen longitudinal data collection and analysis. This report summarizes the discussion and opinions voiced at that workshop.

The term longitudinal studies encompasses all studies in which a defined population is interviewed over an extended period of time. This workshop focused only on studies that gather information from the same

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

respondents in two or more waves of data collection.1 Therefore, the discussion in this report may not be applicable to longitudinal studies that select respondents in each wave from a common sample pool (such as a community or other sampling unit) rather than follow the same individuals.

The workshop distinguished three types of longitudinal studies. Panel studies2 are usually broad-based in sample and topical coverage, and frequently use the household as the sampling unit. In that case, information is collected from all or a sample of members of the selected households. Cohort studies, a subset of panel studies, follow a sample of people selected on the basis of a common age- or time-specific characteristic (such as birth year, age, or class membership). In some cases, the households to which cohort members belong may be included in the study. Longitudinal community studies, also known as population laboratories or demographic surveillance studies, systematically collect data (generally on fertility, mortality, and in- and out-migration) from all individuals (at least all individuals of interest in all households) in geographically demarcated communities. Such studies usually collect data at more frequent intervals than cohort or panel studies, although on a smaller range of topics. The most important distinction, however, relates to the focus of these studies on the community: data are collected from individuals, but the actual unit of observation is the community. Thus in general new people enter the sample as they move into the community, but those who leave the community are not followed.

Background

Information on population and health issues in developing countries during the first half of the 20th century was based on the few censuses that included relevant questions and on a few intensive longitudinal studies. Well-known examples of these studies include the Instituto de Nutrición de Centro América y Panamá (INCAP), which addressed child nutrition in Guatemala, (for a review, see Scrimshaw and Guzman, 1997) and the

1  

A paper by Andrew Foster presented at the workshop and reproduced in Part II explores other longitudinal designs.

2  

Barry Popkin presented definitions of the three types of longitudinal studies considered at the workshop.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Khanna study of health and fertility in India (Wyon, 1997; Wyon and Gordon, 1971).

Beginning in the 1960s, individual research institutions initiated multicountry programs of household surveys that greatly expanded the availability of developing country demographic and health information. These programs, which included the Knowledge, Attitude, and Practice (KAP) surveys on contraception (in the 1960s and 1970s), the World Fertility Surveys (1972 -1984), the Demographic and Health Surveys (DHS, which began in 1984 and continues today), and other survey series sponsored by individual research institutions, generally provide nationally representative, widely accessible, and comprehensive data.

Most of these efforts were cross-sectional (data were collected at only one point in time) and focused on the fertility and health status of women and children. Capitalizing on the explicit goal to develop comparable information for a wide range of countries over time, researchers currently use these data widely. The data have proven especially useful for cross-national comparisons. The coverage of household surveys has increased to the point where, by mid-2001, the DHS database alone contained over 100 datasets for 68 countries. Increased general use of DHS (and other) data can be credited to the development of standard recode files. Moreover, the speed with which survey data are made available to the public and with which analytic studies are conducted and their results published has increased remarkably. The DHS surveys are currently available on the World Wide Web.

The number of longitudinal community studies also has increased dramatically (Kahn and Tollman, 1998). A few of these studies have their roots in the 1950s and 1960s such as the Matlab study in Bangladesh (Aziz and Mosley, 1997); the Khanna (Wyon and Gordon, 1971), Singur (Garenne and Koumans, 1997), and Narangwal (Taylor and De Sweemer, 1997) studies in India; The Medical Social Research Project at Lulliani in Pakistan (Garenne and Koumans, 1997); and the ORSTOM (l’Institut Français de Recherche Scientifique pour le Développement en Coopération) study in rural Senegal (Garenne and Cantrelle, 1997; Garenne and Cantrelle, 2001; Cantrelle, 1969). However, most began in the late 1980s and 1990s (Alderman et al., 2001; Mosley, 1989). Many of these studies were undertaken to evaluate specific interventions such as family planning programs, vaccine-trials, or treatments for specific diseases (Das Gupta et al., 1997). However, research inquiries were often broadened beyond their original intent, extending the life of the study long after

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

the original questions were addressed and enhancing their value for other researchers. Some critics believe these studies, which are time-consuming and expensive because of the repeated collection of data at short intervals, have resulted in findings and publications that may be too limited to justify adding new sites and, in some cases, continuing existing efforts.

The current situation is therefore one in which the number of studies of various types (including but not limited to longitudinal studies and cross-sectional surveys) is large and growing. Yet, at the same time, policy makers are increasingly demanding rapid analysis and policy-relevant findings, and new analytical tools are expanding the ways in which data from studies of different types can be used. In addition, an increasing number of institutions and data collection sites are requesting access to limited funds in circumstances of growing competition and often more costly research environments.

Purpose of the Workshop

In this context, the Committee on Population convened a Workshop on Leveraging Longitudinal Data in Developing Countries in Washington, D.C., in June 2001. The primary goals of the workshop were to examine the comparative strengths and weaknesses of several longitudinal approaches in addressing demographic and health questions in developing countries and to consider ways to strengthen longitudinal data collection and analysis. Workshop participants addressed a wide range of scientific, practical, and strategic issues, concentrating on longitudinal community studies, panel studies, and cohort studies. Africa received special emphasis for two reasons: (1) the ongoing information crisis in the region and (2) the interest of funding agencies in evaluating the potential of expanded investment in longitudinal studies for addressing issues currently crucial to the region.

The intention of the Committee on Population was to provide an arena for discussion among researchers with diverse topical interests, disciplinary backgrounds, experience with longitudinal methods and approaches, and motivations for conducting longitudinal research; it was not to make recommendations about the best longitudinal approaches for various research questions or in various settings. This report provides a summary of the invited presentations and short papers, the discussants’ comments, and the general discussion. Two commissioned papers are reproduced in Part II of this report. The technical discussion ranged broadly from comparison of the approaches themselves, to examples of longitudinal studies, to data col-

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

lection and data management issues, to the relevant innovations in computer science. It also covered additional important and diverse issues, including ethics, collaboration and networking across studies, funding mechanisms, data sharing, capacity building, and researcher/participant/ community relations. The workshop agenda is in Appendix A and a list of participants is presented in Appendix B.

At this point, it is important to clarify what topics were not covered at the workshop and therefore are not covered in this report. To maximize the time devoted to comparing longitudinal approaches, workshop participants did not address topics related to cross-sectional data, including a comparison of longitudinal and cross-sectional approaches, and to techniques such as synthetic cohort analysis and retrospective data, which can be used with cross-sectional data to simulate longitudinal data. Other aspects of collecting and analyzing longitudinal data also were beyond the scope of the workshop. Relevant topics that were not adequately addressed include: tracking respondents (methods or costs); dealing with “split-offs” or changes in the sample produced when members of a household in the study leave the household (such as adult children moving out to establish their own household and separations and divorces); “refreshing” a sample (adding new respondents for those who drop out); changing survey questions if a better approach is developed over the course of a study; deciding on the optimal interval between interviews; analyzing longitudinal data (strategies and techniques); keeping data users informed about features of the data (e.g., if oddities or errors are discovered in the data); and including retrospective data collection in the first wave of a study.

Organization of the Report

This report has two parts. Part I includes an overview of the presentations and discussion at the workshop presented in four sections. The first section considers the benefits of longitudinal data in general. The section that follows compares the advantages and disadvantages of panel studies, cohort studies, and longitudinal community studies and presents considerations for determining the best approach. The third section examines challenges to longitudinal research, highlighting those associated with funding, relationships with respondents, attrition and population change, research biases, and ethics. The final section explores several ways in which longitudinal research efforts can be strengthened to increase returns to researchers, respondents, policy makers, and the scientific community. Part II of the

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

report includes two of the papers presented at the workshop. The first paper, by Andrew Foster, compares panel, cohort, and longitudinal community studies in low-income countries from a methodological perspective. The second paper, by Richard A. Cash and Tracy L. Rabin, presents an overview of ethical issues in developing country research with special reference to longitudinal data. The workshop agenda and the list of workshop participants are included as appendixes.

BENEFITS OF LONGITUDINAL DATA

Throughout the workshop discussion, participants noted the strengths of longitudinal research, even though identifying the advantages of longitudinal studies relative to those of cross-sectional studies was not an objective of the workshop.3 Yet while mentioning the virtues of longitudinal efforts, they continually noted that the use of longitudinal data and the specific approach adopted depend on the research question at hand. For many time-dependent research questions, synthetic cohorts (using cross-sectional data in a way that builds on age groups, representing birth cohorts, to examine how events of interest change over time) or retrospective data from cross-sectional studies may be equally or even more useful and have the additional benefits of lower cost and time intensity. Even when the research question demands longitudinal data, without sufficient time and money for follow-up, longitudinal efforts may be futile.

The benefits of longitudinal research discussed at the workshop can be grouped into two general areas: (1) contributing to scientific knowledge and (2) promoting careful research practices and designs.

Contributing to Scientific Knowledge

Workshop participants suggested that longitudinal research contributes to understanding causal relationships by collecting more accurate and detailed information on the timing and sequence of various events than might otherwise be obtainable. It also seems to permit greater accuracy by4:

3  

This section is based on presentations by Linda Adair, Ties Boerma, Andrew Foster, Barry Popkin, and Stephen Tollman.

4  

This section is based on the presentation by Andrew Foster.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
  • examining changes in various behaviors and related events over time with observations close to the time of the change or event

  • addressing selectivity problems (such as the background characteristics of respondents that may confound the relationship between the variables of interest) in statistical analyses with fewer assumptions

  • studying programs or sources of change in which there are lags between the introduction of an intervention and its possible effects.

The advantage of longitudinal data relates specifically to researchers’ ability to look at change (e.g., before and after differences) for a given individual while, in the process, netting out the effect of (unobserved) characteristics of the individual that do not change over time. With cross-sectional data, researchers compare different individuals at a point in time and must be concerned that differences along the dimension of interest might be due to other (unobserved) differences among individuals.

Longitudinal research of various types has led to a substantial body of scientific and policy-relevant findings. Scientifically, the availability of longitudinal data has allowed researchers to better understand human, social, and economic development processes, to test more dynamic and complex theories of social and health behaviors, and to refine their understanding of causal relationships. Studies have illuminated the health, social, and economic needs of individuals, communities, or subgroups of populations; evaluated the effectiveness of a range of programs and interventions; and enabled policy makers and planners to set priorities based on evidence. Table 1 pulls together some specific examples of study findings that were mentioned at the workshop. In the opinion of several workshop participants, the contributions of longitudinal research to science have been much greater than those to policy to date.

The benefits of longitudinal research can become clearer when research is related to a particular topic, and Ties Boerma did just that in his presentation on the human immunodeficiency virus (HIV) and other sexually transmitted diseases (STDs). Through longitudinal studies of HIV/STDs, researchers now better understand the complex interactions among the biomedical and social determinants of HIV and other STDs. These studies include investigations of the trends and determinants of HIV infection; the impacts of HIV and other STDs on fertility, adult and child mortality, and population size; and the interactions among demographic factors, such as age and migration, and socioeconomic, cultural, and biological factors in the acquired immunodeficiency syndrome (AIDS) epidemic. Longitudi-

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

TABLE 1 Examples of Lessons Learned from Longitudinal Data Presented at the Workshop

Study

Findings

Cohort Studies

INCAP (Guatemala)

Inter-relationships of diet, nutritional status, and infectiona (particularly the sequencingb)

Khanna (India)

Breastfeeding alone is insufficient to supply the calories needed by infants six months and older; supplementary foods are required for infants to fight common intestinal and respiratory diseases

Cebu (Philippines)

Long-term effects of stuntingc

INCAP

Economic, health, and developmental effects of key developmental patternsd

Rationale for multipurpose child care focus in development

Fetal programming (Barker hypothesis): adult health outcomes affected by prenatal and early postnatal environment

Panel Studies

IFLS (Indonesia)

Household adjustments to macroeconomic shocks

CHNS (China)e

Increased malnutrition among rural poor (leading to government policies to lower food prices and initiate anti-poverty efforts)

RLMS (Russia)f

Privatization’s effect on poverty: major expansion of long-term poor

Increased gender and economic inequality

Longitudinal Community Studies

Rufiji, Tanzania

Location of health facilities, use of health services, and infant and child health

Mortality burden of malaria (especially for children)

Bandim, Guinea-Bissau

Risks of Diptheria, Pertussis, and Tetanus (DPT) vaccination for young infants (less than 3 months old); importance of when vaccinations are administered

Matlab, Bangladesh

DPT vaccinations among children between three and five months and decreased mortality

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Study

Findings

Manhica, Mozambique

Rapid increase in mortality of children under age five in year 2000 (possibly as a consequence of the stress of the January-February 2000 floods)

Ifakara, Tanzania

Insecticide treated bednets and reduced under-five mortality

SOURCE: Based on presentations by Linda Adair, Barry Popkin, Stephen Tollman, and INDEPTH information.

NOTE: This table reflects the experience of particular workshop presenters rather than a comprehensive or systematic picture of the field. It should be viewed as illustratious of the range of rich findings generated through longitudinal analyses of various types. INCAP=Instituto de Nutrición de Centro América y Panamá; IFLS=Indonesian Family Life Survey; CHNS=China Health and Nutrition Survey; RLMS=Russian Longitudinal Monitoring Survey.

aMartorell et al. (1990).

bRamakrishnan (1999a); Schroeder et al. (1999); Martorell (1995); Martorell et al. (1995); Ruel et al. (1995).

cMendez and Adair (1999) and Adair and Guilkey (1997).

dRamakrishnan et al. (1999b).

eGuo et al. (2000) and Bell et al. (2001).

fLokshin et al. (2000); Lokshin and Popkin (1999); Popkin and Mroz (1995).

nal studies have provided some information (based on verbal autopsies) on the impact of AIDS on mortality. Longitudinal studies undertaken to evaluate interventions aimed at reducing HIV infection (including community trials) have yielded important information on HIV and sexually transmitted diseases.

The future of HIV/STDs research is likely to continue to be dominated by intervention studies (albeit focused on various aspects of transmission, prevention, and cure). Boerma foresees that more gains in knowledge about HIV/STDs and related population and health issues will require more studies using population-based samples (as opposed to clinic clients or other ad hoc groups) and larger comparison populations. He expects to see more

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

studies that examine HIV/STDs and other aspects of health in a broader perspective and over a longer-term than the current short-term intervention studies. A challenge lies in how to develop ways to overcome the short-lived nature of intervention studies and establish long-term population cohort studies.

Promoting Careful Research Practice and Designs

Another major contribution of longitudinal research, in addition to the many and diverse specific findings it has yielded, is related to the iterative and dynamic processes of inquiry these studies require and the careful research practices and designs they promote. The importance of this kind of contribution was a key theme of presentations by Barry Popkin, Stephen Tollman, and Robert Willis.

Barry Popkin traced the historical evolution of cohort and household panel studies from the very focused studies of specific topics prior to the 1970s, particularly in the field of public health, to the multipurpose and broad-based studies of today. He argued that six changes in approach have improved the scientific and policy impact of these types of longitudinal surveys:

  1. the shift from single purpose to multipurpose surveys.

  2. the shift from retrospective to prospective data.

  3. the shift from individual-level data collection to multilevel data collection, especially community data collected concurrently with individual and household data.

  4. the integration of qualitative work.

  5. the addition of a biomedical perspective to social science work. and

  6. the addition of natural sciences and ecology in the mid-1990s.

Stephen Tollman, in discussing the contributions of longitudinal community studies to science and policy, pointed to specific findings (see Table 1) that have emerged from longitudinal community studies and the overall process of serendipity and feedback that makes a critical contribution to such studies. Drawing on experience from the INDEPTH network (an International Network of field sites with continuous Demographic Evaluation of Populations and Their Health in developing countries), he emphasized the long-term research benefits facilitated by two other key features of

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

longitudinal community studies: (1) the engagement with the community and (2) the development of research infrastructure.

Finally, in a presentation on the lessons learned from the long history of longitudinal studies in the United States, Robert Willis stressed the importance of sound and appropriate designs using the Health and Retirement Study (HRS) and the Asset and Health Dynamics of the Oldest Old (AHEAD) study as examples. These two studies, developed as cohort studies to examine effects of aging on the health and economic well-being of Americans and related policy issues, began with two different age groups but were merged in 1993 and expanded in 1998 with the addition of new cohorts (of the same age as the original cohorts were at entry). HRS designers switched to a steady state design in order to present an ongoing picture of the U.S. population over age 50. Their goals were to promote better understanding of the effects of aging on the well-being of individuals and any changes in those effects that may occur over time, and to improve the capacity of researchers to analyze the effects of social and policy changes as they occur in the future (Willis, 1999).

COMPARISON OF DIFFERENT APPROACHES TO LONGITUDINAL DATA

Panel studies, cohort studies, and longitudinal community studies are, of course, not equivalent in their objectives or their ability to answer specific social sciences questions. As noted earlier, the research question and the goals of the research project, along with context, funding, and other considerations, must be taken into account in determining the best study approach.

Workshop participants raised some of the issues that should be considered in determining the best longitudinal approach for a particular study. The issues include the following eight items:

  1. What is the research question? How can the question best be answered? What types of indicators and measures and what frequency of observation are required to answer the research question (and are these particularly amenable to a certain longitudinal approach)?

  2. What is the context for the study? (What other data on the study area are available? What data are needed? What burden, if any, will the study pose for respondents? Could other studies or activities in the area pose the risk of contamination?)

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
  1. Does the study population need to be representative to a larger population (such as a country or region)?

  2. Is there adequate spatial and temporal variability in the main independent and outcome measures?

  3. Which is more important to address the research question properly: geographical coverage of a particular area or sample size?

  4. What are the main purposes of the undertaking and future plans (e.g., scientific research, capacity building, local investment and infrastructure)?

  5. For topic or sample coverage, which is more important—scope or depth?

  6. What ethical considerations are most relevant for the intended research goals?

Andrew Foster compared the strengths and weaknesses of panel, cohort, and longitudinal comparative studies for three purposes: measurement/description, program evaluation, and structural analysis (hypothesis testing and statistical modeling). A summary based on his presentation and further elaboration by others at the workshop is presented in Table 2.

For measurement, or description, of demographic and health processes or patterns, the major differences among the three approaches stem primarily from the differences in the target populations. Panel studies generally cover a broad population, whereas cohort and longitudinal communities focus on specific subpopulations (based on a common characteristic for cohort studies and on a specific community or geographic area for longitudinal community studies). These differences affect the degree to which findings from a longitudinal study can be generalized to other areas of the country and the extent to which inferences can be made. In addition, the greater depth and breadth of data collected in most panel and cohort studies enhances their analysis options, whereas most longitudinal community studies typically measure selected (but limited) events in much greater detail and have larger sample sizes, enabling better observation of many rare events.

As for program evaluation, the representativeness of the study population to the general population is less salient than other characteristics in determining the suitability of each approach for assessing the impact of community-level interventions. Evaluation is especially sensitive to changes in the study population, so continuity of the sample across time is more important than whether the sample is representative of a larger population.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Thus, longitudinal community studies, which deliberately follow all members of the community, are ideal for assessing program impacts. Panel and cohort studies may be affected by losses in the study population (such as through migration or other changes in household composition) unless they try to follow respondents who move. Cohort studies are further limited by the selection of the sample according to characteristics that may or may not be appropriate for evaluating the program of interest.

On the other hand, longitudinal studies that are located in many communities often take advantage of the community variation (e.g., use community characteristics to achieve statistical identification). Studies that focus on only one community (or a few communities), including most longitudinal community studies, cannot do this.

Differences in the heterogeneity of the sample between longitudinal community studies, on the one hand, and panel and cohort studies, on the other, create different opportunities for program evaluation. Because longitudinal community studies are often tied directly to interventions in specific areas (with control groups in other areas), it is not necessary to have a lot of variation within the sample; the interesting variation is imposed through the experimental design. By contrast, panel and cohort studies, with their broad geographic coverage, require a heterogeneous sample in order to produce variability in outcomes of interest to study.

Under opportune circumstances, however, carefully situated and planned evaluations can be effectively conducted with panel and cohort studies, and they are likely to require less investment than longitudinal community studies and to inspire fewer concerns about the generalizability of findings and the aspects of the study community that may influence findings.

The three approaches differ least in their usefulness for structural analysis. The techniques used in these analyses and their data requirements vary greatly, so that the variation within a longitudinal approach is often greater than the variation across different approaches.

In terms of enhancing community participation and strengthening local research capacity, longitudinal community studies outperform the other longitudinal approaches. The primary objectives of longitudinal community studies extend beyond their potential to advance immediate knowledge. Specifically, enhancing community participation and strengthening local research infrastructures are typically key components of these studies, especially when controlled trials are to be undertaken. The involvement of local communities and capacity strengthening may be, and often are, im-

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

TABLE 2 Session 1—Comparisons of Panel Cohort and Longitudinal Community Studies Discussed at the Workshop

 

Panel Studies

General definition

Usually broad-based in sample and topical coverage. Household is frequently the sampling unit.

Capacity strengthening

Generally work with established institutions and researchers in country

 

Offer temporary jobs, mostly for data collecting, data entry, and coding

 

Generally work with researchers and policy makers at the national level

Ethical concerns

Protecting confidentiality of individual respondents, particularly when sharing data

 

Obligations to respondents to provide results that may be produced from the data years after data collection

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Cohort Studies

Longitudinal Community Studies

Subset of panel studies. Follow a sample of people selected on the basis of a common age- or time-specific characteristic (such as birth year, age, or class membership). May include information on households to which cohort members belong.

Systematically collect data (generally on fertility, mortality, and in- and out-migration) from all individuals in geographically demarcated communities. Data are generally collected at more frequent intervals than for cohort or panel studies, although on a smaller range of topics. Communities are the primary unit of observation (though data come from individuals). Also referred to as population laboratories or demographic surveillance studies.

Generally work with established institutions and researchers in country

Establish long-term research center at site

Offer temporary jobs, mostly for data collecting, data entry, and coding

Offer range of employment opportunities

Generally work with researchers and policy makers at the national level

Generally work with researchers and policy makers at the local (community) level

Protecting confidentiality of individual respondents, particularly when sharing data

Protecting confidentiality of individual respondents

Obligations to respondents to provide results that may be produced from the data years after data collection

Protecting confidentiality of the communities in which respondents live (especially for small-area studies)

 

Obligations to respondents to provide results that may be produced from the data years after data collection

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Comparison for different research goals:

 

Panel Studies

Research goal: measurement/description.

Measuring and describing patterns of demographic change (computation of vital rates and other aspects of individual and household welfare) and changes in these measures and patterns over time.

Advantages

Most likely to be representative of large-area population

 

Greater depth and breadth in socioeconomic and health measures

Disadvantages

Need to refresh panels to remain representative over time

 

Often have few observations of rare events

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Cohort Studies

Longitudinal Community Studies

Representative of group born in a particular cohort

Provides detailed information on specific community(ies)

Greater depth and breadth in socioeconomic and health measures

Decrease the costs and increase the quality of longitudinal data collection in specific community(ies) by establishing research infrastructure

 

Allows precise estimation of mortality through larger (person-year) samples and greater focus on rare events

 

Utilize greater periodicity of measurement

Not representative of or generalizable to general population

Not necessarily representative of other communities; no way to obtain statistical estimate of cross-community variability to derive estimates.

Often have few observations of rare events

 

Outmigrants excluded from subsequent surveys

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

 

Panel Studies

Research goal: program evaluation

Obtaining estimates of program and intervention effectiveness by collecting indicators at two or more points in time before, during, or after program implementation. Often looking at the same individual over time is less important than looking at community effects, or comparisons between individuals or groups of similar ages at time 1 and time 2.

Advantages

Capture greater spatial variability.

 

Allow evaluations of programs focused on geographic areas to control for endogeneity of program placement.

Disadvantages

Significant changes in the sample will inhibit ability to yield good estimates.

 

May have insufficient sample sizes in communities where interventions are placed to study program impacts.

Research goal: structural analysis

Uncovering the mechanisms underlying observed outcomes; hypothesis testing.

Advantages

Generally have a broad range of variables.

 

Useful in looking at short- and long-term outcomes (depending on length of study).

 

Conducive to large, representative samples.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Cohort Studies

Longitudinal Community Studies

Capture greater spatial variability.

Ideal for program evaluation at the community level because of deliberate tracking of entrants, leavers, and relevant behaviors of all members that can be compared at any given points in time.

Allow evaluations of programs focused. on geographic areas to control for endogeneity of program placement.

Significant changes in the sample will inhibit ability to yield good estimates.

Limited geographic coverage, raises concerns about

• spatially autocorrelated variables

• logistical considerations surrounding placement of program villages

• generalizability of findings.

Limited range of variables may impede ability to control for many confounding factors (though several studies include a broad range of variables).

Because of age-specific nature of sample, any program effects related to age are more difficult to evaluate using cohort data (unless the intervention is specific to the cohort being studied).

Data can become dated or useless if changes in the social and political environment change the nature of the cohort of interest (e.g, a change in Medicare laws will affect studies of the health of the elderly).

Generally have a broad range of variables.

Because of the population coverage, generally include information necessary for a panel study and a cohort study—opening the possibility for a range of longitudinal studies and approaches.

Useful in looking at short- and long-term outcomes (depending on length of study).

Useful for understanding life events and cumulative and lifelong exposures to various effects.

Useful in looking at short- and long-term outcomes (depending on length of study).

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

 

Panel Studies

Disadvantages

Attrition can compromise representativeness and estimates

 

Often miss rare events

 

SOURCE: Based on presentations by Andrew Foster, Linda Adair, Jim Phillips, Duncan Thomas, and Barry Popkin and on comments by participants in the workshop.

portant elements of panel and cohort studies, but they are generally not explicit objectives of these approaches. Although some capacity strengthening does accompany panel and cohort studies, especially for the in-country collaborating institutions and for studies connected to U.S. institutions in the past decade through the Fogarty International Center (FIC) at the U.S. National Institutes of Health (discussed later in this report), comprehensive approaches to capacity strengthening and community participation have been concentrated within longitudinal community studies.

At several points in the discussion, workshop participants expressed the sentiment that, though particular methods of longitudinal data collection and analysis may better suit particular research questions, the potential for knowledge multiplies when different longitudinal approaches are used in conjunction (this point is discussed later in the section on “Strengthening Longitudinal Efforts”).

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Cohort Studies

Longitudinal Community Studies

 

Good at capturing rare events because of the population coverage

 

Maximize ability to incorporate serendipity and feedback because of regular intervals of data collection and research infrastructure

Attrition can compromise representativeness and estimates

Attrition can compromise representativeness and estimates.

Often miss rare events

Nonrepresentative beyond community in which data are collected

Problems with estimation: endogeneity (including altered behavior in response to various factors of interest) and difficulty identifying age and period effects

Often include small sample sizes

Randomized at the community level, and so introduces statistical inefficiency

CHALLENGES TO LONGITUDINAL RESEARCH

Despite the benefits of longitudinal research for science and policy, some challenges must be addressed. Good longitudinal efforts tend to be resource-intensive in funding requirements, design and planning, research subject and community participation, and investigators’ investment of time. Institutional (or researcher) continuity is needed to provide the memory and direction for long, complex studies, particularly those requiring several years of observations to generate findings. Additional problems relate to the mobility of respondents and change in the communities under study, as well as the ethical considerations unique to this type of research. Five challenges to working with longitudinal data are described in this section: (1) funding, (2) relationships with communities and respondents, (3) changes in samples, (4) changes in study protocols, and (5) ethical research practices.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Funding

Researchers face the challenge of obtaining and sustaining investment in their projects over a long period. Because of several issues related to the costs, competition for funding, and time requirements of longitudinal data, researchers find it difficult to determine when to terminate, scale back, or maintain or increase research efforts. Workshop participants recognized the possibility that continuing data collection in one site may have diminishing returns compared with starting over in another site, especially when sample attrition or other sample changes affect the representativeness of the sample. However, the cost-effectiveness and other strategic interests associated with continuing, moving, or ending studies was not a topic addressed by the workshop.

Longitudinal studies are more expensive than cross-sectional studies with similar sample size and design (though not necessarily more expensive than repeated cross-sections with the same number of waves). Data are collected repeatedly (at least twice) from respondents. To maintain the integrity of the study, respondents must be tracked over time. Throughout the process, the huge amounts of data generated must be managed and analyzed by project staff.

As the number of longitudinal study sites has grown, the competition for funds has increased. This competition affects both researchers’ abilities to start-up new studies and the long-term sustainability of existing sites. Because of the time requirements and the long-term nature of longitudinal studies, researchers often must rely on limited substantive results for continued funding. With the demands of data collection and management and proposal writing, researchers frequently find themselves lacking the time and resources to analyze the data. Studies therefore often fail, in the eyes of many, to achieve their research potential.

It also takes years for longitudinal studies to produce substantive results. A lag between study initiation and study results is often expected because longitudinal studies are intended to look at trends and long-term effects. The long germination period required by many longitudinal studies was demonstrated with two examples at the workshop. First, Barry Popkin and Linda Adair cited David Barker’s work on the effects of the prenatal and early-natal environment on adult health outcomes as an example of how significant returns to longitudinal studies may emerge years after study initiation (Adair et al., 2001; Barker, 1998; Popkin et al., 1996).

Second, Jane Menken presented results from a study in Bangladesh in

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

which she and her colleagues examined the effects of early life characteristics on later survival. They linked detailed information on women collected in the 1976 Determinants of Natural Fertility Study (including health, nonpregnant weight, fertility history, and household socioeconomic status) with survival data from the International Centre for Health and Population Research’s Demographic Surveillance System in Matlab, Bangladesh, and demonstrated that the effects of socioeconomic and health status in early adulthood persist over the next 20 years. However, the effects of certain key variables on the odds of dying became statistically significant only after 10 years of follow-up (body mass index), 15 years of follow-up (no schooling), or even later (being Hindu was approaching statistical significance at 20 years of follow-up).

Relationships with Communities and Respondents

All longitudinal studies depend on the cooperation of respondents and their communities for the duration of the study. Yet, the demands on respondents and their communities, and the extent to which they are integrated into the research process, differ with the type of study because of different objectives, study populations, practices, and researcher presence in the community. Generally, the longer the relationship with the community and the greater the demands on the participants, the greater is the need for community participation and perceived returns to participants and communities in the research process.

On the one hand, in two-wave panel studies researchers survey respondents and then survey them again some time later. The demands on the respondents and their communities are minimal, generally limited to sampling and data collection needs. The potential benefits to the respondents and their communities are also limited—to any direct compensation for participation or relevant findings that emerge (if shared with the communities). On the other hand, longitudinal community studies have a strong long-term presence in a specific (generally relatively small) study site, and researchers collect data at regular (and often short) intervals from all households. It is therefore vital that researchers have a good rapport and working relationships with community leaders, respondents, and local professionals throughout the process.

Community participation is crucial in three situations: (1) when sensitive topics are included in the research scope; (2) when studies are concentrated in smaller physical areas; and (3) when research involves multiple

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

revisits or complicated questionnaires or research requirements (such as medical tests or time-use diaries). This point was illustrated by James F. Phillips, who described the process through which a study of female genital mutilation (FGM) was integrated into an ongoing longitudinal community study in Navrongo, Ghana, without disrupting ongoing research efforts. The researchers held many meetings and discussions with community leaders and members over several months about whether and how to add questions on FGM to the survey. With community support, they were able to gather data on FGM without compromising their overall goals or the continuing participation of respondents. In the Navrongo project, community leaders have participated in many aspects of the study.

Another dimension of project-community relations involves the returns to communities and individuals from the research activities. These returns range from direct benefits at the research site (such as employment on the project) to increased knowledge and well-being as the results of the studies filter back to the community through information sharing and new policies and programs. Benefits to the community, while desired by many researchers, generally are achieved only through intentional objectives and strategies.

Changes in the Sample

Study populations and samples change over time, affecting longitudinal research. People are added to the study population through in-migration and births and removed from study populations and samples through out-migration, deaths, and refusals to participate. Current demographic patterns in many developing areas suggest that longitudinal researchers need to be aware of these changes; however, information is often insufficient to allow researchers to anticipate specific changes.

Studies differ in the extent to which they accommodate changes by tracking migrants or refreshing the sample. In Cebu, Philippines entire slums have been razed and the number of communities where respondents live has grown from 33 to over 260 as respondents have dispersed due to migration over 18 years of tracking the original mother-infant cohort. HRS/AHEAD’s steady state design, in which new cohorts of those in the youngest age group are added at regular intervals, was incorporated to address losses in the sample (to death) and to keep the sample fresh. The willingness of researchers to track or add respondents and the study objectives affect the need to adjust the study. As mentioned earlier, study areas

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

also differ in the extent to which sample adjustments, tracking, or other efforts prove effective.

Studies differ in their need to address certain population changes. When communities are a key component of the research, studies are more vulnerable to changes, and it is more important to incorporate necessary adjustments. For community-based studies, it is critical to deal with entrants and leavers, though not necessarily through tracking, because the composition of the population in the community (and changes in it) forms the basis for the research. Robert Willis also noted that when context variables (such as community or neighborhood factors) are central aspects of research, sample attrition or change compromises the study if not properly addressed.

Sample attrition is a potential problem for any longitudinal study where researchers want to generalize results. Research has revealed that migrants (respondents who leave) differ in important ways from those who stay. For example, research using the Indonesian Family Life Survey shows that long-distance migrants, short-distance migrants, and nonmigrants differ in educational attainment and earned income levels.5 Examining longitudinal household surveys from Bolivia, Kenya, and South Africa, Alderman and colleagues (2001) saw significant differences between respondents who were retained and those who were lost in follow-ups. Researchers face additional challenges in areas where migration is circular, with people regularly coming and going from a study area.

Two examples of ways in which to deal with sample changes were presented at the workshop. First, Duncan Thomas described three ways to address change in study samples due to migration:

  1. Track migrants. Researchers can attempt to follow respondents who have migrated out of the study area and conduct the follow-up survey(s) with respondents who are found.

  2. Adjust the sample. Researchers can sample a subset of all migrants from the study area and adjust the remaining sample accordingly. Adjustment works best when individuals are the unit of analysis; studies in which households or dwellings are the sampling or identification unit are less amenable to adjustment.

5  

This section is based on a presentation by Duncan Thomas.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
  1. Ignore the changes. Researchers, in some cases, can ignore attrition and population changes.

Thomas reported that efforts between 1997 and 1998 to track migrants in the Indonesian Family Life Survey substantially reduced sample attrition and provided useful information. However, the costs of tracking and interviewing a migrant were about 20 percent more than those for a nonmigrant. Tracking also increases the spread of the study because it entails going to areas that were not originally part of the sample.

Pierre Ngom described his experiences tracking migrants in a study of urban slum residents in Kenya (see Box 1) and pointed out the problems that arise because of changes in the definition of critical units, such as households or dwellings, over time. The costs and difficulties associated with tracking respondents are closely related to the geographic scope of the study. The costs of tracking migrants in the IFLS will be lower than those for a longitudinal community study site because the IFLS has national coverage and therefore study teams at many of the migrant destination sites.

Effects of Changes in Longitudinal Study Protocols

The existence of a research infrastructure (for example, trained interviewers, research center, baseline data) provided by longitudinal studies, especially long-term ones such as longitudinal community studies, may provide an incentive for others to locate independent studies or programs in the area or for researchers to add variables or types of data to their protocols (for example, the existing research infrastructure and baseline information may facilitate evaluation).

Such changes in the study area or research protocol may affect research in two ways: (1) by instigating changes in the study participants and (2) by contaminating the research process. Speaking to the first point, workshop participants raised concerns about whether communities and participants change in reaction to the study—the Hawthorne effect—particularly when multiple revisits to the same respondents or multiple studies in a small area occur. For example, do multiple exposures to survey questions change the reactions of respondents to the questionnaires (e.g., making them more cooperative, more accurate, or more hostile)? Does survey experience stimulate change in respondents’ knowledge, attitudes, or behavior by raising awareness of various issues? Survey experience also may affect research

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

BOX 1 Challenges of Longitudinal Community Studies in Urban Slums in Kenya

In Africa, rapid urbanization, growing poverty and inequity in health outcomes, lack of assessment tools for and knowledge about health and social interventions, and increasing data needs that involve tracking movers in longitudinal studies clearly justify increased longitudinal research in the cities, particularly slums and poor areas. Although longitudinal community studies in Africa have a relatively long history and fill a critical role in collecting demographic and health data, sites have been concentrated in rural areas. Recently, researchers at the African Population and Health

Research Center have initiated a longitudinal community study in an urban slum in Kenya, only the second longitudinal community study conducted in an urban area. Pierre Ngom presented five challenges to conducting longitudinal community studies in urban slums:

  1. Defining residential units, households, and individuals is much more difficult in urban areas than rural areas. Whereas in rural areas households and residential units are easily identifiable and self-contained, urban residents live in units that are impermanent, easily altered, and flexible.

  2. The population is highly mobile. Researchers do not find the same people in the same dwelling unit between two successive interviews.

  3. The right visitation cycle length and residency time thresholds are difficult to determine because of the transient nature of the population and the high numbers of temporary absentees and visitors.

  4. Insecurity and community fatigue (because of a high concentration of research and programs in slum areas) are common characteristics of urban slums. Violence and explosive situations can easily disrupt research and put respondents and researchers in danger. Slum residents are distrustful of researchers.

  5. Software designed for maintaining and using longitudinal data that pertain to rural populations is not easily transferable to urban settings because of different household arrangements and levels of mobility; specific software for urban areas is needed.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

findings in evaluation studies if it makes respondents more (or less) accessible to programs—for example, more informed or more open to programs and treatments. These potential influences are not applicable to research based on objective measures such as biomarkers and anthropometry. The data could also become less accurate over time if respondents learn to answer in certain ways to shorten the interview or avoid sensitive questions (for example, respondents may learn that a particular answer to one question leads to a set of additional questions, and in follow-up waves answer the question in such as way as to avoid follow-up questions).

Second, adding new studies, new types of data (such as the collection of biomarkers or a component on FGM), or new programs to a study area may interfere with original study plans. For example, the presence of several concurrent interventions in a study area makes it difficult to demonstrate precise causal processes and effects. Researchers also must grapple with whether to maintain consistency in variables across surveys or change variable measurement if better measures are developed during the course of the study. Community relationships and participation also may be influenced by such changes.

Ethical Research Practices

More rigorous institutional requirements for research on human subjects and greater knowledge about how research practices can affect individuals and communities have led to a heightened emphasis on ethics in demographic and health research. Yet, growing concerns about reducing the costs of longitudinal data collection and increasing public accessibility to the data often come in direct conflict with ethical concerns.

Longitudinal studies are subject to the same ethical considerations as studies in which data are collected at a single-point in time. The three basic principles of ethics in biomedical research—respect for persons, beneficence, and justice—guide most discussions of ethical considerations.6 These principles encompass broad concerns, including ensuring respondent confidentiality, obtaining informed consent from respondents, and assuring that research benefits outweigh potential risks to respondents (see the paper by Cash and Rabin in Part II).

6  

Some researchers, particularly those in the medical sciences, also include a fourth principle (nonmaleficence) of ethics in biomedical research (see the paper by Cash and Rabin in Part II).

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Many ethical considerations are compounded in longitudinal data collection. First, identity protection becomes a greater concern because an identification variable must be maintained in the data to link data from the same respondent in subsequent waves, possibly requiring extra measures to protect respondents. Second, the additional variables, attributes, and information contained in the data increase the likelihood that respondents can be identified. This is especially true when studies are concentrated in small physical areas in developing countries. Richard Cash argued that the developing country context has forced researchers to extrapolate the core ethics principles from the individual level to that of the community (further discussed in the Cash and Rabin paper in Part II), which is particularly relevant when researchers return to the same communities over time.

Third, the longer life of longitudinal data generates additional considerations related to researchers’ obligations to respondents (for example, providing research results to participants who may move during or after the study or offering treatment for illnesses identified with the research). Obligations to participants are especially important for biomedical research; the concerns may be quite different for social science research in which the studies or interventions pose little risk to participants.

The following ethics-based questions were raised by participants, pointing to the challenges facing researchers who collect and work with longitudinal data:

  • Are researchers obliged to track respondents in case of unanticipated results with implications for respondents? If so, for how long?

  • What are the time limitations on sharing results with respondents? If results take years to show up, do researchers need to find respondents to share the results with them?

  • How do researchers deal with obtaining informed consent from respondents who are minors in one wave of a study and adults in subsequent waves? From whom do researchers need to obtain consent?

  • In addition to obtaining consent from respondents for each wave of data, do researchers also need to obtain consent for analysis and research questions that use multiple waves of data? Do researchers need to obtain permission from respondents to use data in ways other than those originally intended?

  • Should researchers go back and ask participants for consent for an add-on or unanticipated study? What if returning to get permission may compromise the confidentiality of respondents?

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Ethical research practices, particularly related to identification, are especially critical in longitudinal community studies. The geographically specific sites combined with ongoing data collection render longitudinal community studies vulnerable to problems with maintaining the confidentiality and anonymity of respondents. The need to protect the identity of the community itself, particularly if the research topics are sensitive, is another concern for researchers. As awareness of the research spreads, these issues become increasingly challenging. Capacity strengthening efforts also raise concerns about ethics, especially related to what researchers should and can be expected to provide to the community in terms of education, jobs, information, and the like. These concerns were not explored at this workshop.

Common ethical practices in research are evolving to address emerging issues and new technologies. At the U.S. National Institutes of Health, a subject of current debate is whether informed consent should be obtained for third-party subjects. Specifically, should informed consent be solicited from indirect respondents (e.g., other household members about whom primary respondents are asked), and do researchers need to provide the results or benefits of research to these indirect participants? At the same time, ethical issues raised by new technologies, such as geographic information systems (GIS), are affecting researchers undertaking any study, but are of special concern to longitudinal studies where geographical areas are more easily identifiable. Two discussions at the workshop—on incorporating biomedical testing and GIS into social research in developing countries— highlight some of the ethical dilemmas (see Box 2).

STRENGTHENING LONGITUDINAL EFFORTS

Because many longitudinal research efforts tend to be resource intensive and difficult, maximizing the returns to these efforts is a worthy aim. Much of the discussion at the workshop centered on eight ways in which longitudinal efforts could be strengthened: (1) building a strong, but flexible base survey; (2) increasing variables and types of data; (3) networking and collaborating; (4) linking data within and across studies; (5) strengthening research capacity; (6) expanding access through data sharing; (7) increasing data access and use through computer science innovations and technology; and (8) strengthening longitudinal research through funding mechanisms.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Building a Strong, but Flexible Base Survey

Many important findings that have emerged from longitudinal research were unintended at the onset of the study. 7 Because longitudinal studies involve multiple waves of data collection, researchers are able to incorporate changes in study communities during the observation period or employ technological developments to investigate a topic. Flexibility and adaptability are features of longitudinal research. The power of longitudinal data lies not just in repeated measures of the same variables, but also in the ability to add a rich set of non-time-varying covariates for analysis in new domains as well as contextual information for prespecified hypotheses. However, the ability to add on to a study or use data in ways originally unintended requires a strong research design—one that includes a broad range of carefully defined variables providing a solid base on which to build.

Strong study designs are central to researchers’ ability to capitalize on opportunities that arise during the research process. Speaking of longitudinal studies in the United States, Willis argued the most successful studies have clear objectives and a solid base of information, with designs flexible enough to capitalize on any new opportunities accompanying theoretical and empirical developments or a changing context. A well-conceived core of variables increases the likelihood that additional research topics can be easily built in with additional questions or research modules and that collected waves of data can serve as baseline information for new programs, policies, or changes in the study environment.

Recent examples demonstrate the importance of building on a solid foundation. Even though economic shocks were not anticipated at the outset of the studies, researchers were able to, because of a strong foundation, examine carefully the economic crisis in Indonesia through RAND’s Indonesian Family Life Survey (Frankenberg et al., 1998, 1999), and the increasing long-term poverty and need for changes in the social safety net in Russia through the Russian Longitudinal Monitoring Survey conducted by the University at North Carolina at Chapel Hill (Lokshin and Popkin, 1999; Popkin and Mroz, 1995). Likewise, in Mozambique researchers were able to detect the doubling of infant mortality in the study region during 2000, possibly related to the stress of the January and February floods,

7  

This section is based on presentations by Robert Willis, Monica Das Gupta, and Agnes Quisumbing.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

BOX 2 Ethical Issues Associated with Adding Geocoding and Biomarkers to Longitudinal Research

The ethical issues raised by new technologies are affecting all kinds of research, but especially longitudinal studies where geographical areas are more easily identifiable. These issues described here are exacerbated by, but not unique to, longitudinal studies.

Geocoding and GIS data

Because of technological advances in the collection and use of geospatial data and the availability of geographic identifiers, researchers are increasingly incorporating geospatial data into survey data.* Not only are geospatial variables useful as analytical variables, but they also enhance research in two ways: by facilitating the linkage of independent datasets through common geospatial references (e.g., latitude-longitude coordinates, administrative area codes, or enumeration district codes) and by providing the basic information needed to generate other variables (such as those related to distance, availability, direction, and proximity of facilities).

The use of these techniques, however, raises two ethical considerations. First, it is more likely that study respondents could be identified through attributes, locations, or other information obtainable through these codes. And, second, it is more likely that data could be used for purposes other than those intended by the original data collectors, raising issues about informed consent.

Collecting Biomarkers

Scientific developments over the past few decades have dramatically increased researchers’ abilities to collect biological data from testing of respondents requiring small amounts of biological material (such as blood spots and saliva).** These developments have not only helped researchers to provide better diagnoses of some infections and diseases than those based on self-reports or

*  

This section is based on the presentation by Stephen Matthews.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

other measures, but also enabled them to examine some individual characteristics through simple biomarkers. The use of bioindicators in longitudinal social science and public health research raises the following questions:

  • How much time elapses between taking the sample and getting the results? What are the implications of this timing for researchers? Do the results influence the timing of the next visit?

  • How should researchers deal with inaccurate tests (such as false positive and false negatives)? Should tests be repeated? When? Under what conditions?

  • Are the results of use to the individual respondent? Should respondents be given all results?

  • What is the public health significance of the results? How does this affect researchers’ obligations to individual participants?

  • Do the results have social implications (e.g., social stigma) for respondents?

  • Is treatment available or not? What does treatment require?

  • Can researchers store blood or other samples for tests that are developed in the future? Are researchers obligated to store blood or other biological samples? For how long?

  • Do researchers have to track respondents over time to provide them with results discovered later (either unanticipated ones or those that take time to show up)? For how long?

  • Do researchers have an obligation to inform participants about very complex interactions between biological and social characteristics? What are the responsibilities of researchers when treatment for a particular condition requires both medical and social responses?

  • What are researchers’ obligations to secondary respondents or families that may be implicated in findings (such as those indicating the presence of HIV or susceptibility to hereditary diseases)?

**  

Based on presentations by Noreen Goldman and Stan Becker.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

through the Manhica longitudinal community study in Mozambique (described in the presentation by Stephen Tollman).

The next step to strengthening longitudinal efforts involves developing research designs that not only accommodate serendipity, but also encourage it. Monica Das Gupta argued that the number of unintended findings with great consequence to knowledge and public well-being (Aaby, 1997) suggest that researchers should strive to develop designs that foster serendipity and unanticipated findings in the research process.

Increasing Variables and Types of Data

The addition of more or different variables to longitudinal data collection can increase the potential and use of longitudinal data. This is especially true when the added variables substantially improve researchers’ ability to incorporate more accurate measures of important factors or test more complex models. Two specific examples of such variables were discussed at the workshop: geocoding and collecting biomarkers, particularly in conjunction with social surveys.

Stephen Matthews described how the use of geospatial data (data that indicates the physical location and characteristics of a community or site) strengthens longitudinal research by enabling researchers to create new variables and better measures, and to better model complex relationships. The heightened awareness of geospatial and geocoded data, the increasing availability of this data, and technologies to access and use it offer researchers opportunities to add value to existing data. Matthews argued that, when used properly, geospatial data enhances demographic modeling by incorporating geographic relationships and structure into analyses and improving the robustness and validity of spatial demographic modeling and longitudinal methods.

Matthews cautioned, however, that GIS is not a panacea. Selecting appropriate units of analysis and properly aggregating data to new units, accessing and integrating data, dealing with small sample sizes and data clusters, and recognizing problems in geospatial data (e.g., misspecified data, missing data, inconsistent geocodes) require theoretical and practical considerations. From a modeling perspective, GIS adds a huge overlay of complexity in terms of both spatial aggregation and identification.

Likewise, collecting bioindicators (such as blood or saliva samples), particularly in conjunction with social data greatly enhances the quality of several health measures and expands the range of research questions that

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

can be addressed by researchers. Reporting on their experience of adding biomarker data to a broad longitudinal survey in Taiwan, Noreen Goldman and Maxine Weinstein described how they were able to address several limitations of social data, permitting them to incorporate individuals’ health trajectories into their models; examine the relationships between social, economic, psychosocial, and physiological factors; and obtain markers across different physiological systems with recent information, using a large, nonclinical sample. Their experience collecting biomarkers in Taiwan was successful (acceptable response rates, high compliance with protocol, very few complaints from respondents) and promising for social and health research.

In short, achievement of study value will depend on additional variables contributing to, not detracting from, the major research goals, and the benefits of including such variables should outweigh the additional resources they entail.

Networking and Collaborating

The research potential of longitudinal data is greater when combining studies of different types or covering different topical or regional areas. By networking and collaborating across sites and study approaches, researchers can easily leverage longitudinal data, increasing their value. Networking of researchers involved in various studies facilitates the development of comparable datasets across study sites. The value of the research programs such as the (largely cross-sectional) World Fertility Surveys, Demographic and Health Surveys, and the World Bank’s Living Standards Measurement Surveys has been enhanced by obtaining comparable data across countries, enabling comparative research. Perhaps with this advantage in mind, networks are emerging around various longitudinal studies.

The example of the INDEPTH network highlights the added value accrued through relationships between researchers at various sites or involved in projects in different types of data collection or objectives. Founded in November of 1998, INDEPTH is a network of longitudinal community field sites based on health and demographic surveillance systems to capitalize on the research and policy-informing capabilities organized by researchers from various field sites. The network not only provides an opportunity for sites to work with and learn from the experiences of other sites, but also promotes activities and technologies that enable researchers to exploit the comparative capabilities of the research. For example, the network has

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

initiated a monograph series that adds value to findings from individual sites by presenting information (e.g., age-specific mortality rates and life tables) in a comparative manner, enabling researchers to identify seven distinct patterns of mortality in Africa (INDEPTH Network, 2002). INDEPTH plans to produce other volumes presenting demographic and health topics in a similar fashion. The topics include cause-specific mortality and a new initiative on malaria transmission intensity and the mortality burden in Africa, health equity, and migration patterns. These efforts can strengthen individual sites as well as promote standardized datasets and increase the potential for further cross-site comparisons and multisite research.

Recognizing the limitations of longitudinal community studies (primarily in terms of their limited geographical coverage), INDEPTH also is collaborating with the African Census Analysis Project8 to capitalize on the opportunities for examining African populations by means of combining detailed longitudinal data from longitudinal community sites with the coverage of censuses. A collaborative effort between these two organizatons, which may focus on HIV/AIDS, is in the early stages.

Collaboration, which often builds on networks, occurs when one or more groups of researchers, using comparable datasets, carry out research in common or in conjunction with other groups. Plans are under way to initiate studies comparable to HRS/AHEAD in several European countries, expanding the uses of these data. Similar collaborations could be encouraged in developing countries.

Linking Data Within or Across Studies

Linking data to other sources of data in a study or in a particular site (such as data from another study, GIS or community level data) can add value by increasing, without additional data collection, the information available on individuals or other study units when common variables exist. Linking occurs within and across datasets.

Within a single dataset, researchers can link information on couples,

8  

The African Census Analysis Project (ACAP) is a joint initiative of the University of Pennsylvania and several African demographic research and training institutions. The main objective of this initiative is to consolidate African censuses to make them more accessible and increase their comparative potential.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

family members, or households to increase research potential. Allan Hill argued that most analyses of longitudinal data fall short of exploiting their full potential by failing to build on linkages and other internal features of longitudinal designs. By linking and comparing couples’ responses on desired and (subsequent) actual fertility in the Farafenni (Gambia) longitudinal community study, Hill demonstrated that discordant reporting of live births between men and women was significantly related to the woman’s age and the time since the marriage, controlling for the man’s age.9 He illustrated how linking couples improved data quality (particularly age reporting of children), identified incomplete responses, and provided insight into complex family processes, such as polygyny and fostering, which vary over time. Although linking couples generated additional technical, theoretical, and ethical challenges for researchers, it added considerable value to existing studies.

In the study by Menken and her colleagues described earlier, linking data from two studies at the same site—one with detailed information on health and socioeconomic variables at one point in time and one with detailed information on mortality—enabled researchers to contribute to substantive (the influence of early characteristics on later survival probabilities) and methodological (the influence of the length of observation on findings) knowledge.

Particularly relevant to issues around linking data is the development of technologies for generating or handling data. The generation of common variables in multiple datasets through geocoding has already been discussed. Developments to ease the extensive technological demands of large and complex datasets, such as linked data, are presented in the later section “Increasing Data Access and Use Through Computer Science Innovations and Technology.”

Strengthening Research Capacity

Efforts to expand the participation of developing country scientists in research and strengthen their analytical skills improves the quality of current research and research capacity in developing regions.

9  

The older the woman and the more time since marriage, the greater the predicted probability of discordant reporting, and the greater the predicted probability that the man will report more live births than the woman will.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

As a first step, Francis Dodoo reminded workshop participants that capacity strengthening demands a conscientious effort by researchers to identify what communities and developing country scientists want. It is critical that researchers carefully define what capacity strengthening involves, what skills will be enhanced, and how these skills will be developed.

For example, Dodoo argued that capacity strengthening may require both training researchers to be sophisticated users of the technology required to manipulate and analyze data and training them to raise thoughtful and provocative research questions and to develop their own projects. These goals are different and require different strategies. The first involves technological and statistical skills; the second requires analytical and writing skills.

Studies differ in the strategies they employ to strengthen research capacity. Most current longitudinal community studies explicitly incorporate local research development into their projects; indeed, strengthening local research capacity is often a key objective guiding these studies (as discussed earlier in the section “Relationships with Communities and Respondents”). Researchers carrying out panel and cohort studies have generally worked with research institutions in developing countries rather than individual scientists. All studies require that researchers and funding agencies grapple with issues of project ownership and ensure that developing country collaborators receive due credit for their work on joint projects.10

A serious commitment to capacity strengthening requires integrating these goals into the criteria, review process, and evaluation of research and supporting grants. The following two examples of supportive strategies for capacity building presented at the workshop demonstrate the explicit plans and commitment required to strengthen capacity successfully.

First, Kenneth Bridbord described the strategy for improving the research capacity of scientists from developing nations at the U.S. National Institute of Health’s Fogarty International Center (FIC). Serving the overall goal of “promot[ing] and support[ing] scientific research and training internationally to reduce disparities in global health,” the FIC promotes training of scientists and health professionals through training and research grants, international training grants for U.S. citizens, and institutional grants to universities and nonprofit research institutions with demonstrated

10  

This paragraph is based on the presentation of Francis Dodoo.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

collaboration with foreign research institutions. The program is built on five principles for capacity building in developing country scientists:

  1. a long-term effort and commitment

  2. in-country coordination with local and community leadership

  3. promotion of “South to South” collaboration

  4. training, networking, and mentoring

  5. support for a wide range of health professionals (including nurses, nurse midwives, laboratory staff, counselors, physicians, traditional healers, program administrators).

Second, Cheikh Mbacke argued that, “because they entail long-term involvement with communities in geographically defined areas, longitudinal studies have great potential for building both individual and institutional capacity.” Speaking specifically about longitudinal community studies in Africa, he highlighted the need for explicit strategies for capacity strengthening in order to exploit the opportunities for capacity development inherent in these projects. He then presented six critical aspects of capacity building at African longitudinal sites.

The two most important requirements of capacity building are a solid institutional grounding and a capacity for effectively mobilizing resources. The second two most important requirements are demonstrating the relevance and benefit of the research to the local community and reducing the costs of research endeavors through innovative data collection, changes in methodology, and new technologies. The last two important aspects of capacity building are attracting and keeping scientists, especially local scientists, and enhancing networking capacities to increase the analytic potential of the data and project (for example, through comparative studies) and to promote growth opportunities.

Expanding Access Through Data Sharing

Data sharing, or making data available to secondary users, has substantial potential for quickly advancing longitudinal efforts and the related scientific and policy benefits. However, data sharing also raises several concerns about protecting respondents and ensuring data quality. A critical question addressed at the workshop was: how do researchers balance the tension between protecting the property rights of producers of the data and

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

serving the desires of secondary users, including the public, to get the maximum use and benefit out of the data?

Christine Bachrach identified some of the benefits and costs associated with data sharing. Six benefits of data sharing were discussed, specifically how data sharing:

  1. advances science

  2. promotes timely analysis and dissemination of data

  3. increases the speed with which results get out

  4. facilitates linking independent datasets

  5. increases efficient use of scarce resources

  6. promotes hands-on training opportunities.

The costs, or drawbacks, of data sharing include four issues:

  1. Data sharing can pose a threat to the perceived intellectual property rights of investigators.

  2. Data sharing reduces the control of the principal investigator and the scientific community over the use of data, increasing the possibility for misuse, misleading results, or bad research based on the data.

  3. The process of preparing, documenting, disseminating, and supporting data incurs monetary costs.

  4. Data sharing can pose potential risks to privacy of research participants.

The issues raised in this list highlight a core issue in data sharing: successful data sharing requires balancing the potentially competing interests of three interest groups—the data controllers (collectors and primary users), the data users, and the data subjects (or respondent community). In his presentation, Kobus Herbst outlined the different interests and roles of these three groups (see Figure 1).

Data controllers, those who collect and maintain the data, are obliged to produce high-quality data, invest in the local community, attract and keep quality researchers, and protect respondents. These obligations and the nontrivial investments that data controllers make give them the right to use the data before others have access to it. Allowing some time for collectors to work with the data is important for ensuring the quality of the data, incorporating measures to protect the security of respondents, and developing the necessary codebooks and instructions for using the data.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

FIGURE 1 Interests of data collectors, data users and data subjects in longitudinal research projects.

SOURCE: Based on presentation by Kobus Herbst.

Data users, including secondary users, policy makers, and the general public, are interested in recognizing the benefits of data as quickly as possible. Broad use of data reinforces scientific inquiry, a diversity of perspectives of approaches, and investigations other than those planned by the data collectors. Data sharing serves the interests of this group by increasing the knowledge base in a timely manner.

The interests of data subjects include both ethical dimensions and the rights to information. Confidentiality, privacy, and safety are key concerns of data subjects and their communities. Data sharing can compromise the ability of data collectors to ensure these protections. Yet, respondents also have an interest in receiving the benefits of the research—which is encouraged and expedited through data sharing.

These issues again highlight the distinctions between cohort and panel studies, on the one hand, and longitudinal community studies, on the other. Whereas cohort and panel studies in developing countries tend to be fairly

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

accessible to secondary and public users within a few years of data collection, longitudinal community studies have remained under the tight grip of the researchers and institutions collecting the data. These different data-sharing practices arise largely from the different features of these studies. As discussed in previous sections, longitudinal community studies often work with massive amounts of data that are continually collected, and they have goals and relationships with communities that differ greatly from those associated with other longitudinal studies thus data collectors involved with those studies face multiple demands. Again, because they are located in small physical areas, researchers working with longitudinal community studies face added challenges in protecting their respondents.

Many workshop participants agreed that the dilemma of competing interests can be resolved by adopting explicit strategies and creative models that reflect the various needs and concerns and that are appropriate for the topics and goals of studies. Workshop participants then discussed the various approaches to data sharing now in use (generally from most accessible to most restrictive):

  • Datasets are available on the Internet.

  • Data can be acquired through a formal request process (with or without a charge).

  • Variables that could be used to identify respondents (ID variables) or communities (such as geocodes) can be removed.

  • Data can be made available to secondary users who come to multipurpose data centers or enclaves.

  • Subsets of data in the form of specific modules or variables can be made available

  • Data at various levels of aggregation can be made available.

  • Users can request summary tables or specific analyses that are then provided by data custodians.

Thus, depending on the research interests and critical issues associated with a particular dataset or study, different strategies of data sharing can be developed to promote wider use of data without compromising the confidentiality of respondents, the property rights of primary users, or the quality of research. To date, removing identifying variables from shared data, especially public access samples, is the most common approach.

Data collectors also have responsibilities for facilitating use of their data. The Demographic and Health Surveys are currently available on the

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Internet to any interested users that complete a short form. Martin Vaessen described issues that the DHS has addressed in making these data widely available. Specfiically, Vaessen pointed out that data should be provided in usable formats, and multiple formats when possible, to be most useful. Users also need the tools required to use the data, including software, training, and documentation. The documentation should address information about the contents and general features of the data, the history of its collection and use, and its nuances.

Determining when data should be made available is a critical consideration that requires a compromise between the various interests just described. Having a clearly set schedule that details when data will be more broadly available and the procedures for accessing them is perhaps the most important consideration.

Several workshop participants mentioned that often data-sharing practices are determined by the funding agencies of a project; expectations about when the data will enter the public domain is included in the original agreement. Christine Bachrach suggested that funding agencies become more involved with creative models to encourage data sharing and that they work data-sharing expectations into the granting process (as is currently the case with many panel studies supported by the U.S. National Institutes of Health).

Increasing Data Access and Use Through Computer Science Innovations and Technology

Many issues of data sharing, particularly the tension between primary and secondary users, can be addressed through improved technologies for collecting, maintaining and using longitudinal data.11 Although these issues are not specific to longitudinal data, in view of the amount of data produced in longitudinal studies, the technology and processes for managing and analyzing data are particularly salient. Once the intensive data collection and management requirements are reduced, data collectors can spend more time analyzing data. Many longitudinal community sites and the INDEPTH network are currently developing mechanisms for more efficient data storage and management. As primary users shift their emphasis from data handling to analysis with better data management and

11  

This section is based on presentations by Bruce MacLeod and Sam Clark.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

analysis software, their returns increase and the prospects of data sharing are enhanced.

Reducing the time and costs associated with data storage and retrieval is critical for the long-term success of longitudinal community sites. Bruce MacLeod calculated that data storage and retrieval currently account for 10-40 percent of research budgets. These figures are likely to go up if biomarkers or extra security measures are included.

Currently, software to automate construction of data management systems building on consistency logic and basic data types is being developed in the field by experienced users. Simplifying data management of single sites is an important goal for software design; however, a main objective is to develop templates and other mechanisms that facilitate standardized data formats (particularly definitions of key variables and data storage logic) across studies, thereby supporting cross-country comparisons and linkages between datasets.

A basic building block of this project is a relational data model, described by Sam Clark, with the capacity for self-generation (and change) of relational variables with minimal syntax. The Structural Population Event History Register (SPEHR) is an example.12 Requests for summary tables or specific analyses are filled by data custodians. Workshop participants discussed seven design considerations for relational data model for longitudinal population data:

  • standards and comparability: compatible data definitions and storage structures across datasets

  • flexibility and extensibility: ability to manage data with a wide range of realities around time dimension, structure dimension (marriages, households, residences), and relationships

  • self-documentation: management system that automatically generates descriptions of data, logic, and metadata.

12  

The SPEHR is a relational data model being developed by Sam Clark and his colleagues. The model is built on several components: events (birth, death), multievent processes, item-episodes (residency at place X from time 1 to time 2, marital union from date of union to separation), experiences (the manner in which an item-episode is affected by an event or vice versa), shared experiences, and attributes that change over time. A working demonstration is available at http://www.samclark.net/SPEHR/SPEHR.htm.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
  • easy maintenance: automated day-to-day maintenance tasks that include adding, editing, and deleting data and data structures; verifying data integrity; and generating operational reports

  • security: system that includes partitioning of data into secured units and legally inaccessible units and centralized control over data

  • validity: no duplication of data and checks for validity

  • analysis-friendly: user-friendly software based on intuitive design, easily recognizable views of the data, and defined summary measures and analytical building blocks, and defined output data formats

The ultimate goal of relational models and supportive software is to facilitate the sharing and comparing of data from different populations to improve the scientific and research base on demographic and health issues.

Strengthening Longitudinal Research Through Funding Mechanisms

A recurrent and important theme emerged throughout the discussions of each topic covered in this section of the report. For each topic, workshop participants recognized the importance of support from funding agencies to fully realize the objectives identified. If sustainability, capacity building, data sharing, community involvement, linking, and networking are worthy goals, funding agencies should encourage and support them through more precise and explicit strategies in proposals and award criteria. For example, sustainability strategies, including a fixed time limit for funding and a plan to shift to local funding mechanisms, should be built into original proposals to encourage in-country support and less reliance on international funding agencies. If capacity strengthening is a priority, plans for strengthening capacity, including a clear definition of what it entails, should be explicit in the proposals, funding criteria, and evaluations of the project. The same is true for the goals of networking and data sharing, dissemination of results back to the community, and collaborative work. Not only will these strategies enable researchers and funding agencies to better realize their own objectives, but incorporating aspects of interest into the funding mechanisms will help researchers determine which longitudinal approaches are most appropriate given the set of objectives on the table.

CONCLUSION

The main goal of this workshop was to compare the strengths and weaknesses of different longitudinal approaches—specifically, panel, cohort,

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

and longitudinal community studies. These approaches differ in their objectives for research and community participation, the study populations and samples, the potential for addressing various research questions, and the ethical concerns with which researchers must grapple. This report has highlighted how these approaches compare in confronting several challenges faced by longitudinal researchers and in adding value to existing and future longitudinal efforts. A clear theme of the workshop was the importance of using longitudinal approaches that best fit the research questions being asked or the overall goals of the project, which may include aspects of community strengthening and local investment along with the scientific objectives.

A second major theme that emerged was the importance of multiple research approaches to enhance scientific progress and improve the well-being of individuals through effective policies. Workshop participants identified how the weaknesses in one approach could be easily offset by linking the data collected using that approach with other data collected using similar or different approaches. Issues of access to data—clearly a critical need in increasing the use (and value) of longitudinal data—resurfaced throughout the workshop as an area that needs more careful and critical attention, particularly as it affects the ability to protect the confidentiality of respondents and communities. Overall, the workshop emphasized the importance of careful designs that provide the best science and information while protecting respondents and the communities in which they live, whatever the approach.

REFERENCES

Aaby, P. 1997 Bandim: An unplanned longitudinal study. Pp. 276-296 in Prospective Community Studies in Developing Countries, M. Das Gupta, P. Aaby, M. Garenne, and G. Pison. Oxford: Clarendon Press.

Adair, L.S., C.W. Kuwaza, and J. Borja. 2001 Maternal energy stores and diet composition during pregnancy program adolescent blood pressure. Circulation 104:1034-1039.

Adair, L.S., and D.K. Guilkey 1997 Age-specific determinants of stunting in Filipino children. Journal of Nutrition 127:314-320.

Alderman, H., J.R. Berhman, H.-P. Kohler, J.A. Maluccio, and S.C. Watkins 2001 Attrition in longitudinal household survey data. Demographic Research 5(4):80-124.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Aziz, K.M.A., and W.H. Mosley 1997 The history, methodology, and main findings of the Matlab project in Bangladesh. Pp. 28-53 in Prospective Community Studies in Developing Countries, M. Das Gupta, P. Aaby, M. Garenne, and G. Pison, eds. Oxford: Clarendon Press.

Barker, D.J.P. 1998 In utero programming of chronic disease. Clinical Science 95:115-28.

Bell, C., K. Ge, and B.M. Popkin. 2001 Weight gain and its predictors in Chinese adults. International Journal of Obesity 25:1079-1086.


Cantrelle, P. 1969 Étude démographique dans la région du Sine-Saloum (Sénégal). Êtat-civil et observations démographiques. Travaux et Documents de L’ORSTOM, No. 1. Paris: ORSTOM.


Das Gupta, M., P. Aaby, M. Garenne, and G. Pison 1997 Prospective Community Studies in Developing Countries. Oxford: Clarendon Press.


Frankenberg, E., K. Beegle, B. Sikoki, and D. Thomas 1998 Health, Family Planning and Well-Being in Indonesia during an Economic Crisis: Early Results from the Indonesian Family Life Survey. DRU-2013-FGI/NICHD/ UNFPA. Santa Monica, CA: RAND.

Frankenberg, E., D. Thomas, and K. Beegle 1999 The Real Costs of Indonesia’s Economic Crisis: Preliminary Findings From the Indonesian Family Life Survey. DRU-2064-NIA/NICHD. Santa Monica, CA: RAND.


Garenne, M., and P. Cantrelle 1997 Three decades of reseach on population and health: The ORSTOM experience in rural Senegal, 1962-1991. Pp. 235-252 in Prospective Community Studies in Developing Countries, M. Das Gupta, P. Aaby, M. Garenne, and G. Pison, eds. Oxford: Clarendon Press.

2001 Prospective studies of communities: Their unique potential for studying the health transition: Reflections from the ORSTOM experience in Senegal. Pp. 251-285 in The Health Transition: Methods and Measures, Proceedings of an International Workshop, J. Cleland and A.G. Hill, eds. London.

Garenne, M., and E. Koumans 1997 Appendix: Prospective community studies in developing countries: A survey of surveys. Pp. 297-338 in Prospective Community Studies in Developing Countries, M. Das Gupta, P. Aaby, M. Garenne, and G. Pison, eds. Oxford: Clarendon Press.

Guo, X., T.A. Mroz, B.M. Popkin, and F. Zhai 2000 Structural changes in the impact of income on food consumption in China, 1989-93. Economic Development and Cultural Change 48:737-760.


INDEPTH Network 2002 Population and Health in Developing Countries, Volume 1: Population, Health, and Survival in INDEPTH Sites. Ottawa: International Development Research Centre.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Kahn, K., and S. Tollman 1998 The INDEPTH Network: A Comparative Perspective on Field Sites in Africa, Asia, Latin America, and the Middle East. Unpublished manuscript.


Lokshin, M., and B.M. Popkin 1999 The emerging underclass in the Russian Federation: Income dynamics 1992-1996. Economic Development and Cultural Change 47:803-829.

Lokshin, M., K.M. Harris, and B.M. Popkin 2000 Single mothers in Russia: Household strategies for coping with poverty. World Development 28:2183-2198.


Martorell, R. 1995 Results and implications of the INCAP follow-up study. Journal of Nutrition 125(4 Suppl):1127S-1138S.

Martorell, R., J.P. Habicht, and J.A. Rivera 1995 History and design of the INCAP longitudinal study (1969-77) and its follow-up (1988-89). Journal of Nutrition 125(4 Suppl):1027S-1041S.

Martorell, R., J.A. Rivera, and C.K. Lutter 1990 Interaction of diet and disease in child growth. Pp. 307-321 in Breastfeeding, Nutrition, Infection and Infant Growth in Developed and Emerging Countries. S.A. Atkinson, L.A. Hanson, and R.K. Chandra, eds. St John’s, Newfoundland: ARTS Biomedical Publishers and Distributors.

Mendez, M.E., and L.S. Adair 1999 Severity and timing of stunting in infancy and performance on IQ and school achievement tests in late childhood. Journal of Nutrition 129:1555-1562.

Mosley, W.H. 1989 Population Laboratories for Community Health Research. Working paper number 51. Mexico City: Population Council.


Popkin, B., and T. Mroz 1995 Poverty and the economic transition in the Russian Federation. Economic Development and Cultural Change 44:1-31.

Popkin, B.M., M.K. Richards, and C. Monteiro 1996 Stunting is associated with overweight in children of four nations that are undergoing the nutrition transition. Journal of Nutrition 126:3009-3016.


Ramakrishnan, U., R. Martorell, D.G. Schroeder, and R. Flores 1999a Roles of intergenerational effects on linear growth. Journal of Nutrition 129(2S Suppl):544S-549S.

Ramakrishnan, U., H. Barnhart, D.G. Schroeder, A.D. Stein, and R. Martorell 1999b Early childhood nutrition, education and fertility milestones in Guatemala. Journal of Nutrition 129(11):2196-2202

Ruel, M.T., J.A. Rivera, J.P. Habicht, and R. Martorell 1995 Differential response to early nutrition supplementation: Long-term effects on height at adolescence. International Journal of Epidemiology 24(2):404-412.


Schroeder, D.G., R. Martorell, and R. Flores 1999 Infant and child growth and fatness and fat distribution in Guatemalan adults. American Journal of Epidemiology 149(2): 177-185.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×

Scrimshaw, N.S., and M.A. Guzman 1997 A comparison of supplementary feeding and medical care of preschool children in Guatemala, 1959-1964. Pp. 133-156 in Prospective Community Studies in Developing Countries, M. Das Gupta, P. Aaby, M. Garenne, and G. Pison, eds. Oxford: Clarendon Press.

Taylor, C.E., and C. De Sweemer 1997 Lessons from Narangwal about primary health care, family planning, and nutrition. Pp. 101-132 in Prospective Community Studies in Developing Countries, M. Das Gupta, P. Aaby, M. Garenne, and G. Pison, eds. Oxford: Clarendon Press.


Willis, R.J. 1999 Theory confronts data: How the HRS is shaped by the economics of aging, and how the economics of aging will be shaped by the HRS. Labour Economics (special issue) 6(2):119-145.

Wyon, J.B. 1997 Determinants of rates of early childhood sickness and death, and of long birth intervals: Evidence from the Khanna study, rural Punjab, India, 1954-1969. Pp. 54-80 in Prospective Community Studies in Developing Countries, M. Das Gupta, P. Aaby, M. Garenne, and G. Pison, eds. Oxford: Clarendon Press.

Wyon, J.B., and J.E. Gordon 1971 The Khanna Study: Population Problems in the Rural Punjab. Cambridge, MA: Harvard University Press.

Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
This page in the original is blank.
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 1
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 2
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 3
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 4
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 5
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 6
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 7
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 8
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 9
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 10
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 11
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 12
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 13
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 14
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 15
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 16
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 17
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 18
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 19
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 20
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 21
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 22
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 23
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 24
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 25
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 26
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 27
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 28
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 29
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 30
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 31
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 32
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 33
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 34
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 35
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 36
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 37
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 38
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 39
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 40
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 41
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 42
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 43
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 44
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 45
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 46
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 47
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 48
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 49
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 50
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 51
Suggested Citation:"Report: Leveraging Longitudinal Data in Developing Countries." National Research Council. 2002. Leveraging Longitudinal Data in Developing Countries: Report of a Workshop. Washington, DC: The National Academies Press. doi: 10.17226/10405.
×
Page 52
Next: Demographic Analysis of Community, Cohort, and Panel Data from Low-Income Countries: Methodological Issues »
Leveraging Longitudinal Data in Developing Countries: Report of a Workshop Get This Book
×
Buy Paperback | $41.00 Buy Ebook | $32.99
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

Longitudinal data collection and analysis are critical to social, demographic, and health research, policy, and practice. They are regularly used to address questions of demographic and health trends, policy and program evaluation, and causality. Panel studies, cohort studies, and longitudinal community studies have proved particularly important in developing countries that lack vital registration systems and comprehensive sources of information on the demographic and health situation of their populations. Research using data from such studies has led to scientific advances and improvements in the well-being of individuals in developing countries. Yet questions remain about the usefulness of these studies relative to their expense (and relative to cross-sectional surveys) and about the appropriate choice of alternative longitudinal strategies in different contexts.

For these reasons, the Committee on Population convened a workshop to examine the comparative strengths and weaknesses of various longitudinal approaches in addressing demographic and health questions in developing countries and to consider ways to strengthen longitudinal data collection and analysis. This report summarizes the discussion and opinions voiced at that workshop.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    Switch between the Original Pages, where you can read the report as it appeared in print, and Text Pages for the web version, where you can highlight and search the text.

    « Back Next »
  6. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  7. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  8. ×

    View our suggested citation for this chapter.

    « Back Next »
  9. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!