• Pneumonia
• respiratory infection

We thank Dr Challen for her letter regarding our article1 2 in which she highlights the limitations of CURB65 and PSI for guiding ICU admission. This is an important point which a number of authors including ourselves have made previously.3 Our meta-analysis demonstrates that CURB65 and PSI predict 30-day mortality, the end-point for which these scores were originally derived. Unfortunately, 30-day mortality risk does not translate directly into management decisions and so it is important to establish whether severity scores can impact positively in clinical practice. This ‘impact analysis’ is a critical part of the validation of all prognostic tools.4

Guidelines based on severity scores significantly increase the proportion of low-risk patients treated in the community without compromising patient safety or satisfaction5, and we have recently shown that guidance of antibiotic prescribing using CURB65 can safely reduce broad-spectrum antibiotic use.6 For critical care admission, however, the role of severity scores is not established. The major indications for critical care unit admission are requirement for mechanical ventilation or vasopressor support. As others have said, these patients are generally not difficult to identify7 and there are established guidelines such as surviving sepsis for the identification and management of these critically ill patients. There is little evidence that simply being managed on an intensive care unit for a patient not requiring mechanical ventilation or vasopressors improves outcome. Use of scoring systems such as CURB65/PSI or other recently proposed scores to admit these patients to critical care lacks evidence of benefit and may be impractical.

Studies suggest that less than 10% of hospitalised patients with CAP are currently admitted to ICU's. Implementing scoring systems would require a massive expansion of scarce ICU resources. Admitting all patients with CURB65 ≥3 (17–42% of patients), PSI class V (average of 20.9% of patients), SMART-COP score ≥3 (43.3% of patients in the derivation study) or all patients with three or more IDSA-ATS criteria (26% of patients based on the study of Phua et al)8 is not going to be possible without a huge expansion of critical care beds that could not be justified without evidence of benefit to patients. It should be noted that most of the studies reported by Dr Challen failed to exclude patients with do not attempt resuscitation orders or with directives not to be admitted to the ICU. Therefore, these percentages and the pooled performance characteristics do not necessarily reflect their ‘real life’ clinical utility.

Expanding on Dr Challen's statement that application of these tools ‘should be with caution’, we would suggest that severity scores should only be used to predict outcomes for which they have been validated and, as suggested by McGinn,4 scoring systems should have their impact assessed in clinical studies before being applied to guide clinical decisions.