If almost nothing goes wrong, is almost everything all right? Interpreting small numerators. Export

@article{citeulike:6015787, abstract = {Twelve years ago in an article entitled "If Nothing Goes Wrong, Is Everything All Right?" Drs Hanley and Lippman-Hand1 provided a helpful rule of thumb for interpreting numerators of 0. They showed that if 0 events are observed in N trials, the upper limit of the 95\% confidence interval (CI) for the frequency of events is approximately 3/N. Thus, if there are no deaths among 50 patients treated with a new medication, the upper limit of the 95\% CI for the risk of death is about 3/50, or 6\%. On the other hand, if no deaths occur in a study of 1000 patients, the upper limit of the 95\% CI for the mortality rate is 3/1000, or 0.3\%. I have found that similar approximations work for other small numerators and that they can be a helpful shortcut when scanning journal articles. In each case, as the denominator . . .}, author = {Newman, T. B.}, citeulike-article-id = {6015787}, citeulike-linkout-0 = {http://view.ncbi.nlm.nih.gov/pubmed/7563445}, citeulike-linkout-1 = {http://www.hubmed.org/display.cgi?uids=7563445}, day = {4}, issn = {0098-7484}, journal = {JAMA : the journal of the American Medical Association}, keywords = {statistics, xrw}, month = {October}, number = {13}, posted-at = {2009-10-27 13:25:09}, priority = {2}, title = {If almost nothing goes wrong, is almost everything all right? Interpreting small numerators.}, url = {http://view.ncbi.nlm.nih.gov/pubmed/7563445}, volume = {274}, year = {1995} }

TY - JOUR ID - citeulike:6015787 L3 - citeulike-article-id:6015787 N2 - Twelve years ago in an article entitled "If Nothing Goes Wrong, Is Everything All Right?" Drs Hanley and Lippman-Hand1 provided a helpful rule of thumb for interpreting numerators of 0. They showed that if 0 events are observed in N trials, the upper limit of the 95% confidence interval (CI) for the frequency of events is approximately 3/N. Thus, if there are no deaths among 50 patients treated with a new medication, the upper limit of the 95% CI for the risk of death is about 3/50, or 6%. On the other hand, if no deaths occur in a study of 1000 patients, the upper limit of the 95% CI for the mortality rate is 3/1000, or 0.3%. I have found that similar approximations work for other small numerators and that they can be a helpful shortcut when scanning journal articles. In each case, as the denominator . . . IS - 13 JF - JAMA : the journal of the American Medical Association SN - 0098-7484 TI - If almost nothing goes wrong, is almost everything all right? Interpreting small numerators. VL - 274 KW - statistics KW - xrw AU - Newman, TB PY - 1995/10/04/ UR - http://view.ncbi.nlm.nih.gov/pubmed/7563445 ER -