PARTICIPANTS: Dr. Deborah Cook, Dr. William Sibbald, Dr. Gordon Rubenfeld, Dr. Derek Angus, and Dr. Sean Keenan. Dr. Michael Breslow: The intensive care unit is probably the highest mortality center in the average hospital and the highest cost center. Even though it accounts for about 10% of beds, it represents about 25% of acute care costs. There has been considerable attention focused on the consequences of the inefficiencies of our critical care units in terms of emergency department diverts and operating room inefficiencies. Solucient, LLC, provides mission critical intelligence to the health care industry by transforming multiple streams of health care data into actionable information and knowledge. Solucient performs the 100 Top Hospitals: Benchmarks for Success studies to achieve 3 goals: 1. To create useful benchmarks for measuring hospital performance nationally. 2. To highlight areas of wide variation in hospital performance in high-risk, high-cost, and highvolume services. 3. To focus attention on and encourage performance improvement by publishing benchmark scores by hospital type and the names of the 100 hospitals that have set the benchmark scores. VISICU, Inc. was approached to advise them on methodology. A major motivation for our participation in the project was the recognition that hospital administrators and the public are largely ignorant of the important role served by our ICUs, and the considerable variability in ICU performance that exists across hospitals. In addition, though administrative interventions in the ICU environment have not generally succeeded, focused clinical interventions have had impressive efficacy, with some programs showing mortality reductions of as large as 50%, and financial savings of up to 25%. In designing the study, we had to balance Solucient’s desire to have a national sample of hospitals across the United States with the reality that many hospitals have too few ICU patients to accurately assess quality of care. We elected to use Medicare data for analysis (as opposed to some of the state databases in which there might be more detail) because detailed databases are not available for most states. Performance was measured in 3 discrete populations of ICU patients: (1) patients admitted with a medical diagnoses, (2) patients admitted after a surgical procedure, and (3) patients that required mechanical ventilation for more than 96 hours (our presumption of the truly critically ill). To create the first 2 groups, we identified the most frequently encountered ICD-9-CM admitting diagnoses for ICU patients and those ICD-9-CM diagnoses that most often had an ICU stay. Although we were interested in identifying diagnoses with the largest number of patients admitted to the ICU, we did not wish to include diagnoses in which only a small subset went to the ICU. The outcome variables examined for the ICU patients of each hospital were hospital mortality, hospital length of stay (LOS), complications, and daily ICU ancillary costs. We used hospital mortality as an outcome variable because we did not have the ability to differentiate who died in the ICU and who died external to the ICU. Medicare data only provides total ancillary costs for the entire hospitalization without differentiation of what is consumed in the ICU and what is consumed outside the ICU. Because nationwide ancillary costs per day in the ICUs are about 3 times the ancillary cost per day on the floor, we estimated daily ICU ancillary costs from the following: the number of ICU days, the number of floor days, and the total ancillary expenditures for the hospitalization, by using a 3:1 cost ratio. We chose cost per day as opposed to total ancillary cost per ICU stay because total ICU ancillary costs primarily reflect ICU LOS, rather than discretionary use of ancillary resources. Complications were only measured in the surgical population based on the work of Lawthers et al, which suggests that complications cannot be accurately determined from administrative data in medical patients.
Read full abstract