CLASSIFICATION TREES - A POSSIBLE METHOD FOR ISO-RESOURCE GROUPING ININTENSIVE-CARE

Classification and grouping of clinical data into defined categories o r hierarchies is difficult in intensive care practice. Diagnosis-relat ed groups are used to categorise patients on the basis of diagnosis. H owever, this approach may not be applicable to intensive care where th ere is wide heterogeneity within diagnostic groups. Classification tre e analysis uses selected independent variables to group patients accor ding to a dependent variable in a way that reduces variation. In this study, the influence of three easily identified patient attributes on their length of intensive care unit stay was explored using classifica tion analysis. Two thousand five hundred and forty-five critically ill patients from three hospitals were classified into groups so that the variation in length of stay within each group was minimised. In 23 ou t of 39 terminal groups, the interquartile range of the length of stay was less than or equal to 3 days.