ORIGINAL STUDIES Simple Clinical Criteria to Identify Sepsis or Pneumonia in Neonates in the Community Needing Treatment or Referral

million neonates in developing countries; most deaths occur at home. Objectives: To develop simple clinical criteria, enabling health workers in communities to identify neonates with potentially fatal sepsis; and to identify the danger signs alerting mothers to seek care. Methods: In a field trial in 39 villages in Gadchiroli, India, trained health workers visited all neonates at home 8 times during the first 28 days of life, recording signs and outcome without interventions during 1995–1996 and with home-based management of sick neo-nates during 1996–1999. An independent neonatologist assigned the cause of death. We use the term “sepsis ” to include sepsis, menin-gitis and pneumonia. We evaluated 31 signs as predictors of 43 sepsis deaths among 3567 neonates. We also evaluated mothers’ observations as the danger signs to seek care. Results: Simultaneous presence of any 2 of 7 signs (reduced or stopped sucking; weak or no cry; limbs becoming limp; vomiting or abdominal distension; baby cold to touch; severe chest indrawing; umbilical infection) predicted sepsis death with sensitivity 100%, specificity 92%, positive predictive value 27.2 % and negative pre-dictive value 100 % in the nonintervention period. The criteria identified 10.6 % of the neonates in the community as suspected sepsis, at a mean of 5.4 days before death. The criteria remained valid in the postintervention period. Any 1 of the 5 maternally observed danger signs (reduced sucking, drowsy or unconscious, baby cold to touch, fast breathing and chest indrawing) gave 100% sensitivity and identified 23.9 % neonates for seeking care. Conclusion: These criteria identify neonates in the community who are at risk for dying of infection with excellent sensitivity, specificity and negative predictive value but a moderate positive predictive value. They can be used by health workers to select sick neonates for treatment or referral. One potentially fatal case would be treated per 4 presumptive cases treated