Data Collection and Validation in Resource-Poor Settings
Much of the input to the workshop emphasized the point that patterns of morbidity and mortality are shifting both in terms of cause and in terms of age distribution. The workshop paper by Byass, de Savigny, and Lopez notes that changing therapeutic options tend to increase prevalence as compared with incidence for some key diseases; perhaps the most obvious example of this phenomenon in sub-Saharan Africa can be found in the use of antiretroviral therapy against HIV/AIDS, which keep AIDS patients alive for longer periods of time, thus increasing rates of prevalence. Changing patterns of risk factors—such as the prevalence of mosquito breeding sites in the case of infectious diseases, or factors such as tobacco and alcohol consumption in the case of noncommunicable diseases—constitute a further critical factor in the changing patterns of mortality and morbidity (Dalal et al., 2011; Danaei et al., 2011). All of these factors are changing rapidly against a background of sparse and sometimes dubious detailed information about what is actually happening, which makes it even more vital to proactively consider changes in health data systems in sub-Saharan Africa in order to increase the visibility of the continent’s long-term trends and needs in population health (Byass, 2009).
Nevertheless, it is unrealistic to suppose that over the next 10 to 20 years all the countries of sub-Saharan Africa will develop national health information systems that have sufficiently high coverage and achieve global standards of timeliness, completeness, and quality. Thus it is necessary to consider a transitional approach to improving the supply of health information in the short term in ways that are relevant to the essential policy actions that sub-Saharan African countries will need to take as the epidemiological transition unfolds.
In his presentation for the session on data collection and validation, Peter Byass identified several key questions that countries and international agencies should consider:
• What mix of national and local-area data sources are needed for monitoring epidemiological transition, and with what sampling approaches?
• How can continuous longitudinal, repeated cross-sectional, and one-time survey data be effectively integrated within a national information system to reveal epidemiological transitions?
• What are the economic and human resource implications for upgrading national health information systems in order to measure epidemiological transitions?
• What are the ethical and political issues related to long-term improvements in national health information systems?
Byass, de Savigny, and Lopez also described a typology of data sources that can potentially contribute to national health information (see Table 7-1).
Level | Model | Sample | Approach | Examples |
National | National census | All | Complete cross-section | Most countries |
Ongoing registration | All | Complete longitudinal | Industrialised countries | |
Sentinel districts | 1-2% of population | Longitudinal sample | China | |
Cluster surveys | Cluster sample size | Repeatable cross-section | DHS surveys, WHO-SAGE | |
Fixed panel surveys | Cohort sample size | Longitudinal cohort | Millennium Cohort Study | |
At health facilities | All or sample of facilities | Self-selected group | Annual health reports | |
Provincial | Complete population | All | Complete longitudinal | In registered countries |
Cluster surveys | Cluster sample size | Cross-section | Vaccine coverage | |
At health facilities | All or sample of facilities | Self-selected group | Annual health reports | |
Local area | Individual surveillance | Defined area population | Complete in defined area | INDEPTH centers |
One-time surveys | Survey sample size | Cross-sectional | ||
Specific research | Context- dependent | Specific issues of interest |
SOURCE: Byass, de Savigny, and Lopez (2011).
This session also included a discussion of how lessons learned to date from HIV surveillance efforts might be applied. Thomas Rehle described the key features of
second-generation HIV surveillance, which include (1) combining survey methods for greater explanatory power; (2) developing strategic partnership between surveillance and program evaluation; (3) a shift in emphasis toward measuring incidence rather than focusing mainly on prevalence; and (4) recognition of biological (HIV, AIDS, sexually transmitted infections) and behavioral surveillance as integral components adapted to the stage and type of the epidemic. These features imply a surveillance that is more focused on subpopulations at high risk of infection and that has an emphasis on trends over time.
Rehle further described the new South African National Health and Nutrition Examination Survey (SANHANES). It combines questionnaires with physical examinations and biomarker testing, combines longitudinal and cross-sectional design elements, and is designed to make it possible for health and nutritional status to be explored in much greater detail than was previously possible.