Epidemiology and the World Wide Web: is there a 'net benefit?

Abstract

In the early 1960s, practical epidemiology began to come under the influence of computing power. Before that time, epidemiologists had used ingenious devices such as the edge-punched card and hand-driven and even electrically driven calculating machines, all dependent on well known technology ingeniously applied. Regression analysis of study data was essentially unknown because the inversion of matrices beyond 3 x 3 by hand was far too time-consuming or well nigh impossible. Looking toward the future at that time, one would never have envisaged what in fact has taken place over the last 40 years. Large mainframes housed in dedicated buildings beckoned research workers to their halls so that jobs could be delivered to the technical staff as a roll of punched paper tape or a pile of punched cards. Paper tape was quickly discarded, as mistakes in keying the information onto the tape required remaking the tape or troublesome splicing. The IBM punch card allowed corrections to be made more easily, as replacement cards could be quickly inserted into the original deck. An industry grew up to produce heavy electromechanical machinery to record the input data. Magnetic tape, read on machines occupying the space of several filing cabinets, ensured consistency of input of a data set from one analysis to the next, but it was not convenient for storing programs likely to be modified from one run to another. There were few generalized programs suitable for analysis of epidemiologic data, so researchers tended to write their own. The time and effort spent on preparing and verifying data input, writing bespoke programs, and traveling between laboratory and computer center were considerable. At that time, it was obvious what was needed to reduce the