The proportion of HIV-negative LCWI pneumonia cases that would have been admitted to hospital and the proportion of all deaths that may have therefore occurred in hospital, under 8 different clinical management scenarios.

The line links ‘efficient’ scenarios that minimise potentially unnecessary hospitalisation of cases. Scenario C: If all children with malnutrition (low or very low mid-upper-arm circumference or weight-for-age scores) were admitted to hospital but well-nourished children were treated at home: 16% of the PERCH LCWI cases would have been admitted, including 53% of the deaths. Scenario G: If LCWI pneumonia cases with hypoxaemia, or HIV exposure, or low MUAC/WAZ were admitted: 40% of the LCWI cases would have been admitted, including 86% of the deaths. Scenario H: If LCWI pneumonia cases with low MUAC/WAZ or those HIV exposed were admitted: 23% of all LCWI cases would have been admitted, including 71% of the deaths. Scenario I: If LCWI pneumonia cases with any ‘underlying condition’ as per IMCI guidelines were admitted i.e. those severely malnourished (WFH z-score <-3 regardless of age, or MUAC <115 mm for children ≥ 6 months, or pedal oedema on admission, or admission diagnosis of kwashiorkor), or children who were HIV exposed or premature and <6 months of age or with a diagnosis on admission of: heart disease, developmental delays, or congenital abnormalities (e.g. Trisomy 21, glucose-6-phosphate dehydrogenase (G6PD) deficiency): 24% of all LCWI cases would have been admitted, including 66% of the deaths. See S6 Table in S1 File for the complete list of scenarios examined.