OBJECTIVE: To describe methods to assess the practical impact of risk adjustment for central line-days on the interpretation of central line-associated bloodstream infection (BSI) rates, because collecting these data is often burdensome. METHODS: We analyzed data from 247 hospitals that reported to the adult and pediatric intensive care unit component of the National Nosocomial Infections Surveillance System from 1995 through 2003. For each unit each year, we calculated the percentile error as the absolute value of the difference between the percentile based on a risk-adjusted or more-sophisticated measure (eg, the central line-day rate) and the percentile based on a crude or less-sophisticated measure (eg, the patient-day rate). Using rate per central line-day as the "gold standard," we calculated performance characteristics (eg, sensitivity and predictive values) of rate per patient-day for finding central line-associated BSI rates higher or lower than the mean. Greater impact of risk adjustment is indicated by higher values for percentile error and lower values for performance characteristics. RESULTS: The median percentile error was +/-7 (i.e., the percentile based on central line-days could be 7% higher or lower than the percentile based on patient-days). This error was less than 10 percentile points for 62% of the unit-years, was between 10 and 19 percentile points for 22% of the unit-years, and was 20 percentile points or more for 15% of the unit-years. Use of the rate based on patient-days had a sensitivity of 76% and a positive predictive value of 61% for detecting a significantly high or low central line-associated BSI rate. CONCLUSIONS: We found that risk adjustment for central line-days has an important impact on the calculated central line-associated BSI percentile for some units. Similar methods can be used to evaluate the impact of other risk adjustment methods. Our results support current recommendations to use central line-days for surveillance of central line-associated BSI when comparisons are made among facilities.
OBJECTIVE: To describe methods to assess the practical impact of risk adjustment for central line-days on the interpretation of central line-associated bloodstream infection (BSI) rates, because collecting these data is often burdensome. METHODS: We analyzed data from 247 hospitals that reported to the adult and pediatric intensive care unit component of the National Nosocomial Infections Surveillance System from 1995 through 2003. For each unit each year, we calculated the percentile error as the absolute value of the difference between the percentile based on a risk-adjusted or more-sophisticated measure (eg, the central line-day rate) and the percentile based on a crude or less-sophisticated measure (eg, the patient-day rate). Using rate per central line-day as the "gold standard," we calculated performance characteristics (eg, sensitivity and predictive values) of rate per patient-day for finding central line-associated BSI rates higher or lower than the mean. Greater impact of risk adjustment is indicated by higher values for percentile error and lower values for performance characteristics. RESULTS: The median percentile error was +/-7 (i.e., the percentile based on central line-days could be 7% higher or lower than the percentile based on patient-days). This error was less than 10 percentile points for 62% of the unit-years, was between 10 and 19 percentile points for 22% of the unit-years, and was 20 percentile points or more for 15% of the unit-years. Use of the rate based on patient-days had a sensitivity of 76% and a positive predictive value of 61% for detecting a significantly high or low central line-associated BSI rate. CONCLUSIONS: We found that risk adjustment for central line-days has an important impact on the calculated central line-associated BSI percentile for some units. Similar methods can be used to evaluate the impact of other risk adjustment methods. Our results support current recommendations to use central line-days for surveillance of central line-associated BSI when comparisons are made among facilities.