References Table

Notes

1. Systematic research on this computing problem is lacking. Few papers on the computer science challenges of longitudinal demographic studies exist in the scientific literature despite frequent references to computer problems that have posed a major constraint on research [45,62]. More typically, however, research designs and collection designs are published without detailed discussion of issues in computer systems design [57,58]. Some analysts have provided useful presentations of problems [7-9] although tractable solutions remain the subject of investigation.
2. Most field stations utilize some form of demographic surveillance. Demographic surveillance, however, is sometimes used in cluster samples when longitudinal studies require dispersed national or regional samples [1,44].
3. See the review by Terris [60] of the classical studies by Goldberger, Wheeler, and Sydenstricker [23-26].
4. There are several examples of this type of epidemiological investigation in the early public health literature, although the most thoroughly documented are the Hagerstown study and the East Health District investigation [15,36,53,56]. Field stations have played an important role in research on the epidemiology of chronic disease [46].
5. For example, the Muskogee County study in Georgia was a longitudinal investigation of tuberculosis that was eventually used to research the efficacy of BCG vaccine and other tuberculosis control modalities [13,22,49].
6. A summary of the Singur project is in Notestein [47] and in Rao and Mathen [52]. Singur was the first factorial demonstration of the impact of family planning service delivery on fertility. To date, only three projects have demonstrated demographic impact: Singur, Matlab, and Danfa. Both the Khana and Singur initiatives involved family planning only, and were constrained to offering foam tablets and information about the rhythm method. The Singur project was the first experiment to demonstrate that family planning service delivery can have an impact on fertility. The Khana experiment failed to confirm this impact, however. In 1968, a study was launched in 26 villages of Narangwal, Punjab to test the efficacy of integrating health services with family planning. Results were subject to considerable debate, and fertility impact was not definitively demonstrated [31,59].
7. Over a thousand papers are based on Matlab investigations and trials. Useful overviews of the Matlab story include [10-12,17-21,40,50].
8. A useful review of strengths and pitfalls of longitudinal community and household cohort studies appears in Mosley [41]. See also Defo [14].
9. Notable exceptions exist of cohort studies that serve multiple research protocols and whose past data contribute to new studies (for example, the Whitehall and TANESA studies [2,30,39,48]).
10. Two versions of the HRS have been developed. HRS1 is a DOS-based system written in MicroSoft FoxPro 2.1. In January 1999 the HRS2 became available. This is a new MS-Windows version which improves on the HRS1 in a number of significant ways. The system is WINDOWS-95, -98, and -NT compatible. HRS2 has a more user-friendly data entry screen design than HRS1. The layout of the data entry screen resembles the paper forms used for data recording. Consistency logic can be attached to the database table allowing easier project-specific changes and limiting the possibility of inserting data without proper logical checks. Core relationship information now allows for extra-familial relationships, such as social networks, kinship structure, or economic associations. The HRS system permits entry of data on "externally defined" individuals. This, in turn, permits research on ways in which events to nonmembers of households affect risk among members. Also, new HRS2 "external" individual features permit periodic registration of individuals temporarily present in households. This information is sometimes required for prevalence studies. Procedures for recording pregnancy outcomes have been refined so that information about births is registered, not just information about mothers and pregnancies. This permits research requiring data on births (such as birth weight) or data on the characteristics of multiple births. HRS2 is designed to support a number of different ID formats with little to no changes.
The over-arching aim of these system changes is to develop a single model of surveillance that can accommodate almost all longitudinal surveillance operations. By incorporating features of systems that enhance versatility, the HRS2 works for any specified interviewing interval, permits studies on social groups other than households, facilitates detailed specification of household relationships, and allows for complex extended family structures. These features permit health research on social characteristics of families, networks, kinship structures, and other complex relationships.
11. When the clerk types in the household ID, the first screen to appear is the list of household members. In order to assign an event to an individual, the clerk must select that person from the list of members. If the clerk miskeys the household number but the mistake corresponds to a valid ID for another household, that household's member list will appear. At this point the clerk's error should be apparent (the member list on the screen will not match the list on the paper register from which s/he is typing), and s/he can "revert" (back out of data entry mode for that record) without leaving a permanent trace. Only after an individual has been selected and an event button has been clicked is a permanent visitation record associated with the household for that round.
12. Quite detailed field manuals have been developed for the Bangladesh Sample Registration System, a version of the HRS that operates in a dispersed cluster sample of rural households [32-34,42,43]. Using the Bangladesh system as a model, corresponding field manuals have been developed for the Indonesian system, located in Indramayu Regency of West Java [61]. The Navrongo field manual is documented in Binka et al. [5,6].

 

References Table

The Household Registration System:
Computer Software for the Rapid Dissemination of Demographic Surveillance Systems

James F. Phillips, Bruce B. MacLeod, Brian Pence
© 2000 Max-Planck-Gesellschaft ISSN 1435-9871
http://www.demographic-research.org/Volumes/Vol2/6