Objective: To examine the limits of the effectiveness of critical care through the study of patients for whom it was ineffective.
Design: We studied the relationship between resource use and long-term outcome (2-year follow-up) in 402 consecutively admitted critical care patients to develop a benchmark for ineffective applications of critical care. We defined an outcome called potentially ineffective care (PIC), developed and evaluated a model with an independent data set to predict PIC from a patient's response to treatment, and estimated the economic effects of limiting care after a prediction of PIC.
Setting: The combined medical and surgical intensive care unit at a 600-bed university teaching hospital.
Patients: Two groups of 402 consecutively admitted critical care patients, one from 1989, the other from 1991.
Main outcome measures and results: Based on observations from a two-dimensional plot of resource use vs benefit for 402 critical care patients, PIC was defined as resource use in the upper 25th percentile and survival for less than 100 days after discharge. Thirteen percent of the patients fell into the PIC category and used 32% of the resources. A product of the APACHE risk estimates on days 1 and 5 of at least 0.35 predicted 37% of PIC outcomes with a specificity of 98%. In a second data set, PIC outcome prediction had a sensitivity of 43% and a specificity of 94%, and a positive predictive value of 80%. For the hospital studied, reduction of intensity of treatment after a prediction of a PIC outcome would result in a reduction of hospital charges in the range of $1.8 million to $5 million per year.
Conclusion: Patients in the PIC category consumed a large portion of the resources devoted to critical care at an academic teaching hospital. We suggest a change in focus from assessment of the quality of critical care and risk-adjusted mortality to an assessment of ineffective care based on outcome and resource use and a patient's response to treatment over time.