The predominant driver of cost in the intensive care unit (ICU) is length of stay. The most simplistic method for analyzing cost has been to simply assign a dollar value for one day of ICU stay and assume that it is constant. This oversimplification, however, causes significant loss of accuracy for determining the cost of an intervention. Beyond this, a simple total quantity can be misleading, as true costs for hospital stay include direct variable costs (e.g. medications) that can vary from day to day, direct fixed costs (e.g. nursing labor costs) which can only be changed over the course of months, and indirect costs (e.g. depreciation of the building) which are not going to change with a change in clinical practice except over the course of years. Through our research, we aim to address these problems by developing a methodology for modeling costs using data available from rapid chart review and in large databases. Specifically, we will perform a micro-costing analysis by retrospectively extracting costs from hospital accounting data for a large number of patients who have a pre-specified range of medical conditions at two large academic centers. We will then use multivariate regression to generate a model for costs using only data elements that are present in the clinical record (e.g. ICU length of stay, diagnostic code, volume status, insurance status, hospital location, etc.) Through developing and validating our model, we aim to be able to accurately and rapidly estimate costs for both future and past studies, and ultimately aid in guiding where a health system should invest resources for the most benefit.
National Institutes of Health