The effect of lead time bias on severity of illness scoring, mortality prediction and standardised mortality ratio in intensive care--a pilot study

Abstract

The effect of lead time bias on severity of illness scoring, mortality prediction and standardised mortality ratios was examined in a pilot study of 76 intensive care (ICU) patients using APACHE II, APACHE III and SAPS II scoring systems. The inclusion of data collected in the period prior to ICU admission increased severity of illness scores and estimated risk of hospital mortality significantly for all three scoring systems (p < 0.01) by up to 14 points and 42.7% (APACHE II), 50 points and 26.3% (APACHE III) and 23 points and 33.4% (SAPS II), respectively. Standardised mortality ratios fell from 0.99 to 0.79 (APACHE II), 0.96 to 0.84 (APACHE III) and 0.75 to 0.64 (SAPS II), but these changes failed to reach statistical significance. Lead time bias had most effect in medical patients and on emergency admissions, and least effect in patients admitted from the operating theatre. These trends suggest that mortality ratios may not necessarily reflect intensive care unit performance and indicate that a larger study of the effect of lead time bias, case mix, pre-ICU care or post-ICU management on standardised mortality ratios is indicated.