Paediatric intensive care and haematological units are ideal sites for the development of nosocomial infections. These infections remain a significant source of mortality and morbidity and increase length of stay and costs. Selective digestive decontamination (SDD) includes topical antibiotics during the entire intensive care unit (ICU) stay, parenteral antibiotic administered for three to five days, hand hygiene and surveillance cultures of throat and rectum. Its use is based on the observation that resistant bacteria are often imported by the patients themselves, and the fact that transmission via the hands of carers could be responsible only for infections occurring after one week. In adult patients, seven meta-analyses have demonstrated that SDD reduces the odds ratio for lower airway infections, and sometimes mortality (particularly in surgical and trauma patients). The main criticism against SDD is the possible emergence of antibiotic resistant bacteria, which is a growing problem in Europe and United States of America. Only four studies on SDD in children have been reported in the literature: due to methodological weaknesses and small size of samples, definitive conclusion cannot be drawn. However, one study in a 20 bed paediatric intensive care unit has demonstrated that SDD prevent both infections and the emergence of resistant bacteria. Furthermore, it has been demonstrated that more than 50% of children carrying resistant bacteria are detected within 24 hours of admission, suggesting that they import the resistant strains onto the intensive care unit. Factors that predict facility, administration of i.v. antibiotics within the past 12 months, previous intensive care unit admission and hospitalization of a household contact within the past 12 months. As suggested by several authors, the term selective should mean selection of appropriate patient groups (those at high risk of nosocomial infection, e.g. patients mechanically ventilated for at least 48 hours) and units (excluding those where multiresistance is endemic). Obviously, surveillance of patient and unit bacterial ecology and improvement of antibiotic policy must be reinforced.