Initial selection and early deployment of appropriate/adequate empiric antimicrobial therapy is critical to minimize the significant morbidity and mortality associated with hospital‐ or healthcare‐associated infections (HAIs). Initial empiric therapy that inadequately covers the pathogen(s) causing a serious HAI has been associated with increased mortality, longer hospital stay, and elevated healthcare costs. Moreover, subsequent modification of initial inadequate therapy, later in the disease process when culture results become available, may not remedy the impact of the initial choice. Because of this, it is important that initial empiric therapy covers the most likely pathogens associated with infection in a particular patient, even if this initial regimen turns out to be unnecessarily broad, based on subsequent culture results. The current paradigm for management of serious HAIs is to initiate empiric therapy with a broad‐spectrum regimen covering likely pathogens, based on local surveillance and susceptibility data, and presence of risk factors for involvement of a resistant microorganism. Subsequent modification (de‐escalation) of the initial regimen becomes possible later, when culture results are available and clinical status can be better assessed, 2 to 4 days after initiation of empiric therapy. When possible, de‐escalation and other steps to modify antimicrobial exposure are important for minimizing risk of antimicrobial resistance development. This article examines the general process for selection of initial empiric antibiotic therapy for patients with HAIs, illustrated through 3 case studies dealing with healthcare‐associated pneumonia, complicated intra‐abdominal infection, and catheter‐associated bacteremia, respectively. Journal of Hospital Medicine 2012;7:S2–S12. © 2012 Society of Hospital Medicine