The ecology of medical care: origins and implications for population-based healthcare research

Health Services Research, April, 1997 by Kerr L. White

Finally there are specialized denominators such as all patients using specific practices, services, or institutions. Here denominators can be estimated by various sophisticated statistical maneuvers such as the "capture-recapture" methods first used by wildlife biologists. Randomized clinical trials involve a different set of problems. That is why the 1971 WHO Expert Committee on Health Statistics, in first defining terms for evaluating health services, distinguished between "efficacy" and "effectiveness" (White 1971). Estimates of the relative "efficacy" of an intervention or management strategy are generated from highly selected sets of patients participating in a clinical trial. "Effectiveness" describes the usually different results obtained from assessing the same intervention's benefits when used by large numbers of unselected physicians treating unselected patients living in their natural habitats. Many undocumented factors influence the practical application of an "efficacious" intervention in the "real world."

It was William A. Guy who, in the 1850s, first noted the impact of selective bias arising from differences among those sick persons admitted and not admitted to hospitals for similar diseases (Guy 1856). But it took another century before Berkson in 1946 first demonstrated mathematically the importance of selective bias in epidemiological investigations (Berkson 1946). Acknowledgment of the ubiquitous impact of selective bias is now incorporated in etiological, intervention, and healthcare studies. But the identification and measurement of selective bias continues to be a significant problem in studies that employ large databases, especially those involving multi-organizational databases.

Substantial problems remain also with the reliability and validity of the data generated by countless patients, physicians, and other healthcare personnel, as well as with numerous coding and entry clerks in a wide variety of practice settings, institutions, and systems. For example, there are major difficulties in defining all the terms, definitions, and standards that govern the acquisition of the original data; if comparisons across systems are to be made, these need to be standardized. Help arrived with the advent of Uniform Minimum Data Sets, now widely used for hospital discharge abstracts and claims forms (Murnaghan and White 1970). They have also been developed for ambulatory care (Murnaghan 1972) and, to a lesser extent, for long-term care (Murnaghan 1975).

But these are not new ideas! It was Sir William Petty, widely regarded as the father of economics and epidemiology, who in the seventeenth century first suggested comparing the outcomes of different health systems, specifically of the hospitals of London and Paris (Greenwood 1948). It was Florence Nightingale who first urged recording the outcomes of hospital admissions and the development of uniform hospital statistics (Nightingale 1860). And it was J. A. Glover, a health officer in Britain, who during the 1930s first documented substantial differences in the tonsillectomy rates for children in what were otherwise similar cities and jurisdictions-later referred to as the "Glover Phenomenon" (Glover 1938).