Monte carlo simulations comparing fisher exact test and unequal variances t test for analysis of differences between groups in brief hospital lengths of stay

Abstract

BACKGROUND: We examined type I and II error rates for analysis of (1) mean hospital length of stay (LOS) versus (2) percentage of hospital LOS that are overnight. These 2 end points are suitable for when LOS is treated as a secondary economic end point. METHODS: We repeatedly resampled LOS for 5052 discharges of thoracoscopic wedge resections and lung lobectomy at 26 hospitals. RESULTS: Unequal variances t test (Welch method) and Fisher exact test both were conservative (ie, type I error rate less than nominal level). The Wilcoxon rank sum test was included as a comparator; the type I error rates did not differ from the nominal level of 0.05 or 0.01. Fisher exact test was more powerful than the unequal variances t test at detecting differences among hospitals; estimated odds ratio for obtaining P < .05 with Fisher exact test versus unequal variances t test = 1.94, with 95% confidence interval, 1.31-3.01. Fisher exact test and Wilcoxon-Mann-Whitney had comparable statistical power in terms of differentiating LOS between hospitals. CONCLUSIONS: For studies with LOS to be used as a secondary end point of economic interest, there is currently considerable interest in the planned analysis being for the percentage of patients suitable for ambulatory surgery (ie, hospital LOS equals 0 or 1 midnight). Our results show that there need not be a loss of statistical power when groups are compared using this binary end point, as compared with either Welch method or Wilcoxon rank sum test.