Performance of the integrated management of childhood illness algorithm for diagnosis of HIV-1 infection among African infants

Abstract

Early infant HIV-1 diagnosis and treatment substantially improve survival. Where virologic HIV-1 testing is unavailable, integrated management of childhood illness (IMCI) clinical algorithms may be used for infant HIV-1 screening. We evaluated the performance of the 2008 WHO IMCI HIV algorithm in a cohort of HIV-exposed Kenyan infants. From 1999 to 2003, 444 infants had monthly clinical assessments and quarterly virologic HIV-1 testing. Using archived clinical data, IMCI sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were calculated using virologic testing as a gold standard. Linear regression and survival analyses were used to determine the effect of age on IMCI performance and timing of diagnosis. Overall IMCI sensitivity, specificity, PPV, and NPV value were 58, 87, 52, and 90%, respectively. Sensitivity (1.4%) and PPV (14%) were lowest at 1 month of age, when 81% of HIV infections already had occurred. Sensitivity increased with age (P < 0.0001), but remained low throughout infancy (range 1.4-35%). Specificity (range 97-100%) was high at each time point and was not associated with age. Fifty-eight percent of HIV-1-infected infants (50 of 86) were eventually diagnosed by IMCI, and use of IMCI was estimated to delay diagnosis in HIV-infected infants by a median of 5.9 months (P < 0.0001). IMCI had low sensitivity during the first month of life, when the majority of HIV-1 infections had already occurred and initiation of treatment is most critical. Although sensitivity increased with age, the substantial delay in HIV-1 diagnosis using IMCI limits its utility in early infant HIV-1 diagnosis.