Revista de Saúde Pública
On-line version ISSN 1518-8787Print version ISSN 0034-8910
SILVA, Eunice Pinho de Castro. Methods of sample selection for estimates of vaccination coverage. Rev. Saúde Pública [online]. 1986, vol.20, n.5, pp.377-384. ISSN 1518-8787. http://dx.doi.org/10.1590/S0034-89101986000500007.
It is presented two sampling methods for selecting a sample of children of a given age range, living in a geographical area of interest, to estimate vaccinal coverage: the R.H. Henderson and T. Sundaresan method, the Department of Epidemiology and Quantitative Methods of National Public Health School method. A third method of selection is then proposed. The geographical area of interest is divided into parts which are used as primary sampling units (PSU's). First method: 30 PSU's are selected with probability proportional to the population living in the PSU; a starting point ("household") is selected by random selection within each PSU selected: selection of 7 children from within each of the PSU's, begins with the starting household, and then continues to the next nearest household until a total of 7 individuals is obtained (all individuals of the appropriate age living in the last household falling within the sample are included). Second method: the first stage is like that of the first method; previously to the second stage there is a census in each PSU selected in order to prepare a list of all dwelling units where one or more children live; applying systematic selection, 7 dwelling units are selected from this list. Third method: 30 PSU's are selected with probability proportional to Di (number of dwelling units in the PSU i) given by the Census; the dwelling units in the selected PSU are considered in a fixed order; at the second stage, each selected PSU is exhaustively visited, according to the order of the dwelling units; during this visit, only the dwelling units where one or more children live are listed; to these dwellings a systematic selection is applied, the interval Di/b of which is such that with these two stages there is an equal probability selection method (epsem); the over-all sampling fraction is n/^C with n given by 210/a% (a%is the expected response rate) and ^C is the estimate of the number of children living in the area of interest during the survey. The probability of selection of each individual of the appropriate age, in the first method, depends on the ratio Pi/Di'(Pi is the population of the ith PSU; Di' is the number of dwelling units in this PSU, during the survey), it depends too on the number of dwelling units with no children living in them, situated between two dwellings with children, as well as on the number of children living in each dwelling unit. When the ratios Pi/Di' are equal for all the PSU's, and the second item has the same values for the whole area of interest and this happens also with the third item, we shall then have a selection method of equal probability (epsem) to the first. The second method is a probability one, since Pi/D'ci can be known for all the PSU's of the area of interest (D'ci is the number of dweling units with children living in, during the survey; this method will be an equal probability selection when that ratio is the same for all PSU's. The third method is an epsem, independently of the values of Di. The first method is used by the Expanded Programme on Immunization and is considered effective, simple and inexpensive. The second and third present some changes and consist of alternative methods which attempt to diminish the mean square error of the estimator, although they are less simple and more expensive than the first one. The third requires more data than the second. Since the estimator used to estimate the vaccinal coverage presupposes self-weighing samples, the principal purpose of the third method was to provid a method which will be epsem, i.e., will give to every child of the age group studied, living in the area of interest, equal probability of selection.
Keywords : Vaccination; Estimation technics; Sampling studies; Probability.