Reply: Paediatric hospitalacquired bacteraemia in developing countries

摘要

We congratulate Alexander Aiken and colleagues1 on their important study, which addresses two under-researched areas of hospital epidemiology: paediatrics and sub-Saharan Africa. The paper brings to light a subtle statistical issue that has been largely neglected in published studies on hospital infection, but which could lead to bias when assessing risk factors for nosocomial infection.In survival analyses, as presented by Aiken and colleagues, patients who do not have bacteraemia are often regarded as censored when they are discharged. However, hospital discharge is a competing event for nosocomial infection.23 Risk factors should then be assessed in two ways: an event-specific hazard regression (one for nosocomial bacteraemia and one for the hospital discharge), and a subdistribution approach.4 These approaches will esti mate different effects for risk factors than a simple survival analysis if the same risk factor has an effect on both infection and discharge.