Setting: Mzuzu Central Hospital (MZCH), Mzuzu, and Chitipa District Hospital (CDH), Chitipa, Malawi. Objective: To compare management and outcomes of human immunodeficiency virus (HIV) exposed infants in early infant diagnosis (EID) programmes at MZCH, where DNA polymerase chain reaction (PCR) testing is performed on site, and CDH, where samples are sent to MZCH, between 2013 and 2014. Design: Retrospective cohort study. Results: Of infants enrolled at MZCH (n = 409) and CDH (n = 176), DNA PCR results were communicated to the children's guardians in respectively 56% and 51% of cases. The median time from sample collection to guardians receiving results was 34 days for MZCH and 56 days for CDH. In both hospitals, only half of the dried blood spot (DBS) samples were collected between 6 and 8 weeks. More guardians from MZCH than CDH received test results within 1 month of sample collection (25% vs. 10%). Among the HIV-positive infants, a higher proportion at MZCH (92%) started antiretroviral therapy than at CDH (46%). The relative risk (RR) of death was higher among infants with late DBS collection (RR 1.3, 95%CI 1.0-1.7) or no collection (RR 5.8, 95%CI 4.6-7.2), and when guardians did not receive test results (RR 8.3, 95%CI 5.7-11.9). Conclusion: EID programmes performed equally poorly at both hospitals, and might be helped by point-of-care DNA PCR testing. Better programme implementation and active follow-up might improve infant outcome and retention in care.