Abstract
High mortality entry criteria for new therapies such as ECMO require the use of retrospective data often with small numbers of patients. A new computer-generated statistical method called “Bootstrapping” has made this possible. We conducted a 2 year retrospective study of term infants(n=30) with respiratory distress admitted to our NICU to determine criteria that wouldpredict an 80% mortality for future pts. These criteria would permit an informed decision as to when to advance from conventional therapy to ECMO. Variables considered were: pH, pCO2, pO2, BE, MAP, FIO2, IMV, PIP, PEEP, AaDO2. At the time of maximal conventional therapy, all parameters were analyzed comparing survivors to nonsurvivors. Maximum, minimum, median, and quartile values for each variable were compared. To quantify the differences in the distributions, we compared an overlapping statistic for each variable which showed AaDO2 andPL predictive of survival. Graphing these result revealed an AaDO2 ≥605 with a PL ≥38 to predict 84% mortality. To confirm this Bootstrapping was applied. This method permits resampling with replacement of the original data base(X100) with recomputation of the mortality predicted by the variables of interest. With this method we determined that 75 out of 100 cycles would produce 91% mortality with the median 84.6%. Bootstrapping allows small pop. to be analyzed with confidence.
Article PDF
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Beck, R., Short, B., Sacco, W. et al. 166 THE USE OF A NEW STATISTICAL METHOD TO DEVELOP HIGH MORTALITY ENTRY CRITERIA FOR NEW THERAPIES SUCH AS J.W EXTRACORPOREAL MEMBRANE OXYGENATION(ECMO). Pediatr Res 19, 138 (1985). https://doi.org/10.1203/00006450-198504000-00196
Issue Date:
DOI: https://doi.org/10.1203/00006450-198504000-00196