Article Text

Download PDFPDF

Lung infection: a multi-faceted problem
S129 Predicting outcome from HIV-associated pneumocystis pneumonia
  1. D Armstrong-James1,
  2. A J Copas2,
  3. P D Walzer3,
  4. S G Edwards4,
  5. R F Miller2
  1. 1Hospital for Tropical Diseases, London, UK
  2. 2University College London, London, UK
  3. 3University of Cincinnati, Cincinnati, USA
  4. 4Camden Provider Services PCT, London, UK


Background The presentation of Pneumocystis jirovecii pneumonia (PCP) ranges from mild to severe. The former responds to antimicrobial therapy, the latter has a high mortality rate despite treatment. Several studies have described clinical and laboratory factors that are predictive of death from PCP. Our objectives were to create a prognostic scoring model to aid the clinician in predicting outcome from HIV-associated PCP.

Methods A prognostic scoring model was built using risk factors identifiable at/soon after hospitalisation, that is, ‘by the bedside’— which have previously been identified as being associated with mortality from PCP (a repeat episode of PCP, the patient's age, their haemoglobin (Hb), PaO2 (breathing room air), both on admission, the presence of co-morbidity (Comorb), such as lymphoma or pregnancy, and the presence of pulmonary Kaposi sarcoma (PKS) (Walzer PD, et al CID 2008;46:625–33). The model was built from data concerning 592 consecutive episodes of PCP that had occurred among 540 patients presenting to a specialist inpatient HIV treatment centre.

Results The prognostic scoring model was: [25.5 + (age in years/10) + 2 (if a repeat episode of PCP) + 4 (if PKS detected) + 4 (if Comorb present)−PaO2 – Hb]. The prognostic model produced scores ranging from 0 to 20; median (IQR)=9 (7–11). Testing interactions between risk factors and time showed the model to be applicable across all time periods. Patients were divided into five groups according to prognostic score: 0–3.9=group 1, 4–7.9=group 2, 8–11.9=group 3, 12–15.9=group 4, 16 or greater =group 5. Abstract S129 Table 1 shows mortality rates among the 540 patients with PCP, grouped according to their prognostic scores.

Conclusions While this prognostic scoring model requires further validation in patient cohorts from other healthcare institutions, it is potentially a simple ‘by the bedside’ method of identifying patients early in their hospital admission who are at high and low risk of in-hospital death from PCP and so may aid the clinician in assessing the severity of illness and in deciding on treatment strategies.

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.