Statistical power and sampling size

A further requirement for economic evaluation is recruitment and retention of a sufficient sample of patients and/or facilities to show statistically significant changes between groups. Many economic evaluations have previously been underpowered, mainly due to the skewed distribution of costs among subjects. To take an example, Gray et al. (19 showed (retrospectively) that at 80 per cent power, their case management study (n = 30) was sufficient to detect between-group differences of approximately 30 per cent for total costs, but to detect a 20 per cent difference in health-care costs alone, over 700 subjects per arm would have been required! This raises a number of questions, including what constitutes a worthwhile difference in costs, and what is the additional research cost of basing sample size estimates on economic (as opposed to clinical) variables. What remains clear, however, is that the uncertainty introduced into economic findings by the variability of costs must be addressed not only by sensitivity analysis but also by statistical power considerations at the design stage.

Invisible Viagara

Invisible Viagara

You are about to discover the "little-known" techniques, tricks and "mind tools" that will show you how to easily "program" your body and mind to produce an instant, rock-hard erection. Learn how to enjoy all of the control, confidence and satisfaction that comes from knowing you can always "rise to the challenge" ... and never have to deal with embarrassment, apologies, shyness or performance anxiety in the bedroom, ever again.

Get My Free Ebook


Post a comment