Scoring systems designed to rate the severity of an illness are being used for comparison of hospital units to identify different standards of care and to allocate resources. One such scoring system is the Acute Physiology and Chronic Health Evaluation (APACHE) system which is designed to assess the severity of illness of patients in intensive care units (ICUs). It is widely assumed that different ICUs can be compared by the ratio of actual mortality to that predicted by the APACHE score. However, we suggest that the use of physiological data that can be influenced by medical and nursing intervention should not be used for audit. For example, by good care a patient may be made less severely ill and, therefore, may have a lower actual mortality while, at the same time, accumulating only a low APACHE score with low predicted mortality. This patient could have, therefore, the same mortality ratio as a patient treated inappropriately, who may have a higher actual mortality and a high APACHE score with greater predicted mortality. Paradoxically, the very accuracy of these scoring systems for assessing the severity of illness precludes their use for comparison and audit.