To explore immunologic risk factors for death within 90 days of highly active antiretroviral therapy (HAART) initiation, CD4(+) and CD8(+) T cell subsets were measured by flow cytometry and characterized by logistic regression in 149 Zambian children between 9 months and 10 years of age enrolled in a prospective, observational study of the impact of HAART on measles immunity. Of 21 children who died during follow-up, 17 (81%) had known dates of death and 16 (76%) died within 90 days of initiating HAART. Young age and low weight-for-age z-scores were associated with increased risks of mortality within 90 days of starting HAART, whereas CD4(+) T cell percentage was not associated with mortality. After adjusting for these factors, each 10% increase in CD8(+) effector T cells increased the odds of overall mortality [OR=1.43 (95% CI: 1.08, 1.90)] and was marginally associated with early mortality [OR=1.29 (95% CI: 0.97, 1.72)]. Conversely, each 10% increase in CD4(+) central memory T cells decreased the odds of overall [OR=0.06 (95% CI: 0.01, 0.59)] and early mortality [OR=0.09 (95% CI: 0.01, 0.97)]. Logistic regression prediction models demonstrated areas under the receiver-operator characteristic curves of ≥85% for early and overall mortality, with bootstrapped sensitivities of 82-85% upon validation, supporting the predictive accuracy of the models. CD4(+) and CD8(+) T cell subsets may be more accurate predictors of early mortality than CD4(+) T cell percentages and could be used to identify children who would benefit from more frequent clinical monitoring after initiating HAART.