Aims 8.8 million children under the age of 5 die worldwide each year and neonatal mortality contributes 4 million of these deaths. 99% of neonatal deaths occur in low and middle income countries. Identifying clinical signs which indicate severe illness in this age group is vital so that infants requiring urgent hospital referral and treatment are detected early. The WHO Integrated Management of Childhood Illness (IMCI) divides children into three age groups: 0–6 days, 7–59 days and 2–59 months. Current guidelines for infants less than 1 week of age are based on a small pool of evidence. This review was conducted focusing on all infants less than 2 months of age to help provide evidence for a referral checklist for sick neonates less than 1 week of age and allow existing IMCI guidelines to be improved.

Methods Electronic databases and relevant websites were searched using a specified search strategy. Citations which were eligible focused on clinical signs indicating severe illness in infants less than 2 months of age living in resource poor countries. Relevant papers were critically appraised and data was analysed.

Results 10 full text articles were applicable to the research question and predefined inclusion criteria. All studies were prospective cohorts, two of which were multicentre. Studies predominately focused on ‘severe illness’ as an outcome measure but mortality and infection were also used. Key clinical predictors of severe illness in infants less than 2 months of age were found to be: difficulty feeding, movement only when stimulated, temperature <35.5°C or ≥37.5°C, respiratory rate ≥60/min, severe chest indrawing and history of convulsions.

Conclusions An algorithm to identify severe illness could encompass all infants under 2 months of age, providing jaundice is included for those infants in the first week of life. Current IMCI guidelines may be simplified to accommodate these changes. Combining key clinical signs gives a sensitive and specific approach to allow early identification of seriously ill infants. This, used in conjunction with education and improved implementation of algorithms, could help tackle the Millennium Development Goal of reducing childhood mortality by two thirds of the 1990 level by 2015.