Content area
Full Text
Introduction
Sepsis is a common and highly morbid clinical syndrome that affects more than 850 000 patients in the United States each year [1] . Although hospital survival rates have been improving over the past decade [2,3] , mortality remains high, and sepsis continues to be the most common reason for death in noncardiac intensive care units (ICUs) [4] .
To date, clinical and epidemiologic studies in septic populations have typically reported outcomes in terms of crude, all-cause mortality rates [5] . This practice, however, may obscure important heterogeneity in reasons for death among septic patients. Even when reasons for death are reported, there is no standard methodology, and results cannot be easily compared across trials [6] . Understanding the proximal cause of death in patients with sepsis not only is essential to directing the focus of clinical care but also is critical in the design and interpretation of interventional clinical trials. In a study geared toward improving hemodynamic function, for instance, patients who die of hypoxemia should perhaps be interpreted in a different light than those who die as a result of refractory shock. Furthermore, enriching trial cohorts with patients more likely to have a certain reason for death (eg, refractory shock) may limit the administration of a study drug to patients who are unlikely to benefit while allowing for a smaller overall sample size.
In the present study, we developed a methodology for classifying the reason for death in patients with sepsis and septic shock. We then applied this methodology to a retrospective cohort of patients with sepsis or septic shock who died during their hospitalization. Finally, we sought to identify patient characteristics associated with certain reason for death categories.
Materials and methods
Study population
This was a retrospective, single-center observational cohort study of consecutive patients who were diagnosed with sepsis or septic shock and ultimately died in a tertiary care center between March 2013 and June 2014. Patients were included if they had evidence of systemic hypoperfusion (systolic blood pressure <90, need for vasopressors, and/or lactate >=4), infection was identified as the primary cause of hypoperfusion (determined via retrospective review of the clinical chart), and the patient died during their hospitalization. Patients were excluded if infection was not the primary cause of...