December 1, 1980International Journal of Epidemiology
Control of Confounding in the Assessment of Medical Technology
Authors
Discussion
Loading...
Member takes
Implication
Methodological analysis demonstrates the necessity of clinical judgement for confounding control in medical technology assessments, highlighting risks of automated statistical variable selection.
Key Points
- Examine principles for identifying and controlling confounding variables when evaluating the effectiveness of medical technologies.
- Analyzed conceptual examples of confounding, specifically confounding by indication, in healthcare evaluations.
- Assessed the validity of clinical subject-matter judgement compared with automated statistical selection algorithms like stepwise regression.
- Determination of whether a covariate acts as a confounder requires clinical judgement and subject-matter knowledge rather than purely statistical criteria.
- Statistical selection procedures based solely on significance testing, including stepwise regression, can produce misleading conclusions in medical technology evaluations.
Cite This Study
Greenland et al. (1980) studied this question.