Multivariate analysis of the role of school-attendance status in the introduction of variola minor into the household

Abstract

A discriminant-function analysis was applied to the set of 169 cases introducing variola minor into the corresponding households during the 1956 epidemic occurring in the capital city of Braganca Paulista County, Brazil. The analysis was aimed at discovering how well the four groups of households whose introductory cases were either an adult, a JT-school pupil, a JG-school pupil or a preschooler could be distinguished on the basis of six selected variables: (1, 2) relative location (x,y coordinates of the dwelling): (3) time (in days) from onset of the first case of the epidemic; (4) whether vaccinated or not; (5) total susceptible population in the household, and (6) distance from the residence of an introductory case to the residences of the case which started the chain. Prior allocation of introductory cases to school-attendance status groups was justified since discriminant-function analysis was able to distinguish among the four groups.

Cluster analysis was applied to the same data, in order to partition the cases into as homogeneous groups as possible, regardless of their actual status as pupils, adults or preschoolers. The results suggest that the simple division of cases by school-attendance status, while significant, was probably not the most meaningful. A cluster analysis was further applied to cases (excluding those from two schools), using only three variables: (1, 2) relative location and (2) time from onset. This analysis proved useful as a means of identifying many of the real subchains of contagion.