MethodsGroups of HIV-infected and HIV-uninfected infants younger than 18 months (mainly younger than 6 months) were compared to identify clinical features that could differentiate the two groups. The HIV-infected group also was compared with HIV-infected children older than 18 months. Recruitment was as follows for the group younger than 18 months: 708 children admitted with sepsis and clinical features suggestive of HIV infection were screened for HIV1 and HIV2 by HIV enzyme-linked immunosorbent assay (ELISA), and polymerase chain reaction (PCR) was undertaken on all ELISA-seropositive blood samples (270). HIV infection was confirmed in 136 (19.2%), 438 (61.9%) were HIV-seronegative, 27 (3.8%) were HIV seroreverters, 36 (5.1%) were HIV-seropositive but PCR negative (uninfected), and 71 (10.0%) were indeterminate. One hundred thirty-six HIV-infected children were compared with 501 uninfected children. Confirmed HIV-infected children older than 18 months attending the pediatric HIV clinic were compared with the 136 HIV-infected children younger than 18 months.ResultsUnder 18 months, the median age of HIV-infected children (n= 136) was 4.0 months (range, 3d–18mo) and the median age of the uninfected children (n= 501) was 1.0 month (range, 3d–18mo). HIV-infected children were more likely to have had injections, chloroquine, and nystatin, and to have attended a health center or hospital (p< .001). In the HIV-infected group, theZscore for weight-for-age was −1.75, length-for-age −0.78, and weight-for-length 1.86, significantly lower scores than those of the uninfected group, which were −0.60, −0.23, and 3.05, respectively (p< .05). The mean head circumference was below the third percentile in 40% of HIV-infected compared with 22% of uninfected children (p< .001). Overall, 56 (8%) children had marasmus, 6 (0.8%) kwashiorkor, and 3 (0.4%) marasmic kwashiorkor. Sixteen percent of the HIV-infected and 7% of uninfected children had marasmus (p< .05). The 1989 revised World Health Organization clinical criteria for diagnosis of AIDS had sensitivity, specificity, and positive predictive values of 28%, 98%, and 93%, respectively. Older than 18 months (n= 109), the median age was 24 months (range, 18–60mo). The following were significantly more common in HIV-infected children older than 18 months than in those younger than 18 months: bacille Calmette-Guérin vaccination scar, parotid enlargement, nonspecific generalized dermatitis, and chronic diarrhea (p< .001). Oral candidiasis was more common in the group younger than 18 months (p< .001). In infants examined in the hospital for infective conditions, oropharyngeal candidiasis, ear discharge, dermatologic disorders, generalized lymphadenopathy, lobar consolidation, hepatosplenomegaly, and failure to thrive, especially marasmus, were important indicators of HIV infection.