What is known and objective: Adverse drug events (ADE),common and underestimated in ICU patients, have direct consequences on length of stay, mortality and hospital costs. Critically ill patients with HIV/AIDS are at a high risk of ADE because of their need for multiple drug therapies. ADE can be prevented, especially by the identification of potentially harmful drug-drug interactions (DDIs). Electronic databases are useful tools for the investigation of DDIs to avoid potential ADEs, thereby increasing patient safety. The purpose of this study was to compare the classification and severity rating of potential adverse drug interactions seen in the prescriptions for patients with HIV/AIDS in two databases, one with free access (Drugs.comTM) and another requiring payment for access (Mi-cromedexâ). Methods: A cross-sectional retrospective study of the prescriptions issued for 40 ICU HIV/AIDS patients on mechanical ventilation, admitted for more than 48 h, in a referral hospital for infectious diseases in Rio de Janeiro, Brazil, was undertaken. One prescription was reviewed each week for each patient from the second day after admission. A list of all drug–drug interactions was generated for each patient using the two drug-drug interactions databases. The weighted kappa index was estimated to assess the agreement between the classifications of DDIs identified by both databases and qualitative assessment made of any discordant classification of recorded drug-drug interactions. Results and discussion: Of the 106 prescriptions analysed, Micromedex and Drugs.com identified 347 and 615 potential DDIs, respectively. A predominance of moderate interactions and pharmacokinetic interactions was observed. The agreement between the databases regarding the severity rating was only 68.3%. The weighted kappa of 0.44 is considered moderate. Better agreement (82.4%) was observed in the classification of mechanism of interaction, with a weighted kappa of 0.61. What is new and Conclusion: DDIs are common between the prescriptions of patients with HIV/AIDS admitted to the ICU. Although both databases were able to identify the clinically relevant DDIs, we observed a significant discrepancy in the classification of the severity of DDIs in the two bases. The free access database could serve as an alternative to the identificationof DDIs in resource-limited settings; however, there is a need for better evidence-based assessments for your use on clinical management of more serious DDIs.