Identifying sick children requiring referral to hospital in Bangladesh.

The object of this study was to evaluate and improve the guidelines for the Integrated Management of Childhood Illness (IMCI) with respect to identifying young infants and children requiring referral to hospital in an area of low malaria prevalence. A total of 234 young infants (aged 1 week to 2 months) and 668 children (aged 2 months to 5 years) were prospectively sampled from patients presenting at a children's hospital in Dhaka, Bangladesh. The study paediatricians obtained a standardized history and carried out a physical examination, including items in the IMCI guidelines developed by WHO and UNICEF. The paediatricians made a provisional diagnosis and judged whether each patient needed hospital admission. Using the paediatrician's assessment of a need for admission as the standard, the sensitivity and specificity of the current and modified IMCI guidelines for correctly referring patients to hospital were examined. The IMCI's sensitivity for a paediatrician's assessment in favour of hospital admission was 84% (95% confidence interval (CI): 75-90) for young infants and 86% (95% CI: 81-90) for children, and the specificity was, respectively, 54% (95% CI: 45-63) and 64% (95% CI: 59-69). One fourth or more in each group had a provisional diagnosis of pneumonia, and the IMCI's specificity was increased without lowering sensitivity by modifying the respiratory signs calling for referral. These results show that the IMCI has good sensitivity for correctly referring young infants and children requiring hospital admission in a developing country setting with a low prevalence of malaria. The guidelines' moderate specificity will result in considerable over-referral of patients not needing admission, thereby decreasing opportunities for successful treatment of patients at first-level health facilities. The impact of the IMCI guidelines on children's health and the health care system must be judged in the light of current treatment practices, health outcomes and referral patterns.