To determine how routine inpatient provider-initiated HIV testing differs from traditional community-based client-initiated testing with respect to clinical characteristics of children identified and outcomes of outpatient HIV care. Prospective observational cohort. Routine clinical data were collected from children identified as HIV-infected by either testing modality in Lilongwe, Malawi, in 2008. After 1 year of outpatient HIV care at the Baylor College of Medicine Clinical Center of Excellence, outcomes were assessed. Of 742 newly identified HIV-infected children enrolling into outpatient HIV care, 20.9% were identified by routine inpatient HIV testing. Compared with community-identified children, hospital-identified patients were younger (median 25.0 vs 53.5 months), with more severe disease (22.2% vs 7.8% WHO stage IV). Of 466 children with known outcomes, 15.0% died within the first year of HIV care; median time to death was 15.0 weeks for community-identified children vs 6.0 weeks for hospital-identified children. The strongest predictors of early mortality were severe malnutrition (hazard ratio, 4.3; 95% confidence interval, 2.2-8.3), moderate malnutrition (hazard ratio, 3.2; confidence interval, 1.6-6.6), age < 12 months (hazard ratio, 3.2; 95% confidence interval, 1.4-7.2), age 12 to 24 months (hazard ratio, 2.5; 95% confidence interval, 1.1-5.7), and WHO stage IV (hazard ratio, 2.2; 95% confidence interval, 1.1-4.6). After controlling for other variables, hospital identification did not independently predict mortality. Routine inpatient HIV testing identifies a subset of younger HIV-infected children with more severe, rapidly progressing disease that traditional community-based testing modalities are currently missing.