Article Text

Download PDFPDF

Severity scores for CAP. ‘Much workload for the next bias’
  1. Santiago Ewig1,
  2. Antoni Torres2
  1. 1Thoraxzentrum Ruhrgebiet, Kliniken für Pneumologie und Infektiologie, EVK Herne und Augusta-Kranken-Anstalt Bochum, Germany
  2. 2Servei de Pneumologia, Institut Clinic del Tórax, Hospital Clinic de Barcelona, Facultad de Medicina, Universitat de Barcelona, Idibaps, Ciber de Enfermedades Respiratorias (CIBERES), Spain
  1. Correspondence to Prof. Dr. Santiago Ewig, Thoraxzentrum Ruhrgebiet, Kliniken für Pneumologie und Infektiologie, EVK Herne und Augusta-Kranken-Anstalt Bochum, Bergstrasse 26, D-44791 Bochum, Germany; ewig{at}

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Tools for the assessment of severity of patients with community-acquired pneumonia (CAP) have attracted much interest in the recent past. In this issue of Thorax (see page 878) two systematic reviews and meta-analyses address the value of such tools including no less than 401 and 23 studies.2 Despite different inclusion strategies and statistical approaches, both studies report two main and similar conclusions. First, both the most extensively investigated tools (Pneumonia Severity Index (PSI) and CURB-65/CRB-65) have remarkably favourable power to predict mortality. Secondly, whereas PSI is somewhat better in predicting patients at low risk, the reverse is true for CURB-65/CRB-65. However, these differences are of questionable clinical relevance. Thus, both tools can be regarded as equivalent. CRB-65 is the simplest tool and can easily be remembered and also applied in the outpatient setting. These straightforward conclusions are the result of a decade of intensive and successful work to establish clinically useful tools of severity assessment. So far success is impressive; clinicians now can use a very simple tool as an adjunct to clinical judgement, and studies on patients with CAP can rely on a validated tool for severity stratification.3

At this point, it is worthwhile having a look at the perspectives beyond: where do we have to go from here? Actually, there are many issues that still must be addressed concerning validation in intervention studies and still insufficently recognised ambiguities inherent to the severity scores.

PSI has been prospectively validated in independent populations as a tool to guide site of treatment decisions, and the use of the PSI was associated with a larger proportion of patients in PSI risk classes I and II who were treated in the outpatient environment without compromising their safety.4 No such studies have been performed for CURB-65/CRB-65, and obviously these studies need to be performed.

On the other hand, it is difficult to imagine that such a complex tool as the PSI will ever be implemented in routine clinical practice, particularly in the outpatient setting. CURB-65, and particularly CRB-65, is clearly superior in terms of simplicity, and it is its simplicity that makes it preferable even if it provides slightly inferior predictions. However, the four-variable CRB-65 score is a delicate tool and vulnerable to ambiguities of its components. First, the respiratory rate may be difficult to assess in patients with rapid shallow breathing. In fact, no rules have been established regarding how to assess the respiratory rate precisely. Secondly, pneumonia-associated confusion is difficult if not impossible to distinguish from premorbid mental deficiencies in elderly and disabled patients. Evidently, pre-existing and pneumonia-related confusion cannot be confidently differentiated in every case. As a result, confusion cannot be regarded exclusively as a parameter reflecting only severe sepsis but must also be recognised as a parameter possibly reflecting pre-existing mild central nervous system co-morbidity. The substrate of the parameter ‘confusion’ remains somewhat vague. Thirdly, if CURB-65 is used, blood urea nitrogen may be a confounder in patients with pre-existing renal insufficiency and in the elderly. In fact, some have found that CURB-65 works less well in elderly patients. In one recent study of hospitalised elderly patients, the area under the curve (AUC) was significantly higher for the <65-year cohort in comparison with older patients (0.93 vs 0.7).5 Others have suggested that oxygenation instead of confusion and blood urea nitrogen might work at least as well.6 There is evidence that performance status is an independent predictor for short- and long-term mortality in hospitalised elderly patients.7 Accordingly, combining the modified PSI with performance status led to better predictions being obtained.8 Fourthly, the threshold of age (65 years) seems arbitrary and there is some evidence that higher cut-offs might work better.9 Splitting age into decades might be even more promising. However, the rule would clearly lose much of the main strength of simplicity if stratification by age were to form a part of it.

The simplicity of the CRB-65 rule is not without pitfalls. A score of CRB-65 = 1 already implies an increased risk of death and should prompt consideration of hospitalisation. However, when this score is applied in primary care, hospital referral would have to be considered in all patients just because of an an age older than 65 years. Although increasingly age >50 years is associated with increased risk of death, it is clearly impractical and inadequate to hospitalise all patients with CAP only because they are older than 65 years.10 Thus, it is important to exclude age alone as a criterion when CRB-65 is used as an aid for the decision to hospitalise.

Studies addressing severity assessment of CAP mainly refer to hospitalised patients, and clearly more data are needed for the outpatient setting. Obviously, the available tools do not measure the same things and are imperfect, and there has been insufficient attention so far paid to the reasons behind this. Several studies have incorporated biomarkers to improve mortality predictions, and as things stand today it appears that some of these (in particular C-reactive protein and11 12 procalcitonin,11 but also adrenomedullin)13–15 should be able to meet expectations. Limited data are available for specific important subgroups of patients with pneumonia such as patients with cancer, those receiving haemodialysis and those residing in nursing homes. Settings different from that of the derivation study may result in different predictions of the severity tools. Finally, as stated by Chalmers et al, impact analyses are lacking to investigate the impact of severity tools on predefined outcomes.1

These perspectives are challenging, and deserve much attention in the future. Nevertheless, rethinking the potentials and limitations of severity scores for pneumonia, there are two areas of uncertainty which need to be addressed. The first relates to the adequate use of severity scores as an aid for the decision to hospitalise. In younger and/or not severely disabled patients, things are quite straightforward: the more severe condition with which a patient with pneumonia presents, the clearer is the indication to hospitalise. This is quite different from the far more complex situation of elderly and disabled patients. It is well described that the functional status of patients with CAP may worsen after having been hospitalised.7 16 17 This worsening is associated with severity of pneumonia and premorbid performance status.7 Some of this decline may be ascribed to hospitalisation per se and may be preventable. In fact, ‘hospital at home care’ has been shown to be feasible and to be associated with fewer complications.18 In another study, patients treated with ‘hospital at home care’ experienced modest improvements in performance scores, whereas those treated in a traditional acute care hospital declined. Likewise, a greater proportion of ‘hospital at home care’ patients improved in function and smaller proportions declined or had no change in functional status.19 Finally, a considerable proportion of severely disabled patients may be candidates for restriction on treatment escalations and are better treated where they reside, provided adequate palliative care can be offered. All these considerations should lead to the investigatation of which elderly patients with CAP truly benefit from hospitalisation and to include the assessment of function before and during hospitalisation as an integral part of clinical evaluation of these patients. Moreover, models of ‘hospital at home care’ carry an enormous potential for improvements in the care of elderly and disabled patients and should be further evaluated in terms of efficacy and cost-effectiveness.18

The second important issue relates to the assessment of severe CAP. Neither PSI nor CURB-65/CRB-65 is an adequate predictor of the need for intensive care treatment, either in the intensive care unit or in intermediate care settings.19–21 Therefore, additional parameters and scores have been derivated and validated, and it appears that the modified American Thoracic Society (ATS) score as well as the most recent Infectious Diseases Society of America (IDSA)/ATS criteria are better predictors of the need for intensive care treatment.22–25 Nevertheless, they both remain imperfect, and additional severity scores seem impractical (Ewig S et al unpublished data 2010). Therefore, we need to look for parameters which can be incorporated into the CURB-65/CRB-65 scores and which improve the predictions of the need for intensive care treatment without compromising overall risk predictions (particularly low risk predictions) and the simplicity of these scores. Oxygenation as assessed by oximetry is a compelling variable which may bring about substantial independent predictive power to such a severity score. Again, biomarkers may be of help at least in the emergency department and/or hospital setting. As the German poet Bertolt Brecht said: ‘I have much workload, I am preparing my next bias’.


View Abstract


  • Linked articles 133280, 134072.

  • Competing interests None.

  • Provenance and peer review Commissioned; externally peer reviewed.

Linked Articles