Methods for the Field and Computation
The search for epidemiological methods by pioneer investigators of cardiovascular diseases led to increased communication among them, then to improved methodology. A certain collegiality emanated as well from early conferences on survey methods, study design, and work-in-progress.
Essential to early CVD prevalence surveys and the prospective studies that followed them were standard, validated survey methods for symptom questionnaires, blood pressure and other physical measurements, the electrocardiogram, blood lipid and other chemical measures, and objective diagnostic criteria. New field methods were rigorously and systematically developed, often collaboratively, among the World Health Organization, the London School of Hygiene and Tropical Medicine, and St. Thomas’s and Guy’s Hospitals, the Laboratory of Physiological Hygiene in Minnesota, and committees of the American Heart Association and the International Society of Cardiology (Holland 1962; Rose and Blackburn, 1968).
For analysis, in the early 1960s, those in the new field of CVD epidemiology were able to put aside their cumbersome mechanical calculators and start sorting punched cards to get their row and column sums, summed squares, and products. But no sooner had researchers become comfortable with cross-classifications, two variables at a time, than electronic computers burst upon the scene, rapidly expanding the horizons of computation and inference. Computer facilitation of data storage and management, of computation of correlations and regressions, along with the capacity for handling multiple variables simultaneously, was technically akin to its contemporary phenomenon of blasting off into space!
Electronic computers made possible one of the major contributions of modern epidemiology: analysis of the effect of multiple variables while adjusting for their interactions. For example, the multiple logistic regression was first implemented in CVD epidemiology by Jerry Cornfield (Truett et al. 1967), soon followed by Cox’s proportional hazards ratios and life table regression (Cox 1972).
Bradford Hill of London is given the credit due him for assimilating epidemiological-statistical methods from the past and developing new methods for improving the design of observational studies, for the design and analysis of randomized clinical trials, and for compiling the now-classic considerations for causal inference from statistical correlations in observational studies. Hill also trained many pioneers in chronic diseases epidemiology.
Automated chemistry vastly improved the reliability and efficiency of survey measurements as did the role of the World Health Organization (WHO), the National Institutes of Health (NIH), and Centers for Disease Control (CDC) in providing laboratory standards and testing and manuals of survey methods.
The evolution of epidemiological methods went hand in hand with the increasing sophistication of investigators and their computers, leading to rapid progress in the new discipline of CVD epidemiology. (Henry Blackburn)
References
Cox, D. 1972. Regression models and life tables. Journal of the Royal Statistical Society Series B, 34, pp. 187-220.
Holland, W.W. 1962. The reduction of observer variability in the measurement of blood pressure. In J. Pemberton, ed. Epidemiology. Reports on research and teaching, Oxford University Press, London.
Rose, G.A., Blackburn, H. 1968.Cardiovascular Survey Methods. Geneva: World Health Organization.
Truett, J., Cornfield, J. and Kannel, W. 1967. A Multivariate Analysis of the Risk of Coronary Heart Disease in Framingham. Journal of Chronic Diseases, 20: 511-524.
For more detail on computing methods development, please click here to visit the essay, “John Vilandre on How Data Processing and Computing Developed in One Center of CVD Epidemiology.”