Medical error statistics

Pharmacoepidemiology and Statistical Science. In most cases, medication errors can t be blamed on a single person. ... 

Statistics must also project in the IND the magnitude of difference necessary to demonstrate significance in variations between the laboratory values and clinical findings in patients treated with the test agent and those given placebo or standard medication. These determinations, as well as many of the particulars of the selected randomization patterns in the phase 2 and 3 trials, rest on a fundamental ability of statistics, that is to ascertain the likelihood of error in a particular statement and to estimate the confidence that can be placed in any experimental value. 

Proof in the strict sense cannot be delivered by a controlled clinical study, which has been called the sacred cow . (5) However, the probability of statistical error in terms of the chosen target criteria can be fixed in advance. In clinical studies, different interpretations of results are still possible, since intuitive medical observation and judgement remain indispensable in the individual case (E. Buchborn, 1982). The triad of empiricism, intuition and logic is necessary in both diagnosis and treatment (R. Gross, 1988). 

In conclusion, planning a method comparison study to achieve a given power for detection of medically notable differences should be considered. In this way, a method comparison study is likely to be conclusive either the null hypothesis of no difference is accepted, or the presence of a relevant difference is established. Otherwise, a statistically nonsignificant slope deviation from unity or intercept deviation from zero or both may either imply that the null hypothesis is true, or be an example of a Type II error (i.e., an overlooked real difference of medical importance). 

For a given sample size n, alpha and beta errors are inversely related in that, as one reduces the a error rate, one increases the j3 error rate, and vice versa. If one wishes to reduce the possibility of both types of errors, one must increase n. In many medical and pharmaceutical experiments, the alpha level is set by convention at 0.05 and beta at 0.20 (Sokal and Rohlf, 1994 Riffenburg, 2006). The power of a statistic (1 — /3) is its ability to reject both false alternative and null hypotheses that is, to make correct decisions. 

A second and more plausible explanation for the failure to reach statistical significance may have been due to the small sample size. This is a likely cause since all three general trends (menses duration, inter-menses duration and cycle duration) were generally consistent across all three cycles. Many volunteers failed to complete the project or were excluded in our analyses due to occasional failure to apply the solutions or use of medication. This resulted in an increase in the likelihood of a Type II error and reduced power of the test. As such, future replications must ensure a larger sample of subjects who complete the entire project. 

Wiles, A. D. et al. "A statistical model for point-based target registration error with anisotropic fiducial localizer error." IEEE Transactions on Medical Imaging 27.3 (2008) pp. 378-390. 

---