Peter M. C. Klein Klouwenberg David S. Y. Ong Marc J. M. Bonten Olaf L. Cremer Classification of sepsis, severe sepsis and septic shock: the impact of minor variations in data capture and definition of SIRS criteria Received: 14 October 2011 Accepted: 14 February 2012 Ó Copyright jointly held by Springer and ESICM 2012 P. M. C. Klein Klouwenberg ( ) ) D. S. Y. Ong O. L. Cremer Department of Intensive Care, University Medical Center Utrecht, Utrecht, The Netherlands e-mail: p.m.c.kleinklouwenberg@ umcutrecht.nl Tel.: ?31-887561124 P. M. C. Klein Klouwenberg D. S. Y. Ong M. J. M. Bonten Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, The Netherlands M. J. M. Bonten Department of Medical Microbiology, University Medical Center Utrecht, Utrecht, The Netherlands Abstract Purpose: To quantify the effects of minor variations in the definition and measurement of sys- temic inflammatory response syndrome (SIRS) criteria and organ failure on the observed incidences of sepsis, severe sepsis and septic shock. Methods: We conducted a prospec- tive, observational study in a tertiary intensive care unit in The Netherlands between January 2009 and October 2010. A total of 1,072 consecutive adults were included. We determined the upper and lower limits of the measured incidence of sepsis by evaluating the influence of the use of an automated versus a manual method of data collection, and variations in the number of SIRS criteria, concur- rency of SIRS criteria, and duration of abnormal values required to make a particular diagnosis. Results: The measured incidence of SIRS varied from 49 % (most restrictive setting) to 99 % (most liberal setting). Subsequently, the incidences of sep- sis, severe sepsis and septic shock ranged from 22 to 31 %, from 6 to 27 % and from 4 to 9 % for the most restrictive versus the most liberal measurement settings, respectively. In non-infected patients, 39–98 % of patients had SIRS, whereas still 17–6 % of patients without SIRS had an infection. Conclusions: The apparent incidence of sepsis heavily depends on minor variations in the definition of SIRS and mode of data recording. As a consequence, the current consensus criteria do not ensure uniform recruitment of patients into sepsis trials. Keywords Systemic inflammatory response syndrome Sepsis Intensive care unit Clinical trial Epidemiology Diagnosis Introduction The American College of Chest Physicians/Society of Crit- ical Care Medicine (ACCP/SCCM) consensus conference definitions for sepsis and severe sepsis are used in Intensive Care Units (ICUs) worldwide [1]. According to these defi- nitions, the diagnosis of sepsis requires clinical evidence of infection and the presence of a systemic inflammatory response syndrome (SIRS) that is characterized by specific physiological alterations, including aberrations in tempera- ture, white blood cell count, heart rate and respiratory rate. Despite the deceptive simplicity of this concept, the application of the SIRS criteria in diagnosing sepsis is not straightforward. Although SIRS symptoms are clearly associated with increased mortality and length of stay [2], their occurrence has proven to be neither specific for infection, nor robust across various clinical and research settings [3, 4]. In addition, the extent to which SIRS criteria are present depends on the time since the start of infection as well as resuscitation. Many researchers involved in sepsis trials have thus felt frustrated when a patient, who by clinical judgment clearly had sepsis, Intensive Care Med DOI 10.1007/s00134-012-2549-5 ORIGINAL Author's personal copy