Early recognition of sepsis is crucial to prevent unnecessary deaths as a consequence of acute infections. General practitioners (GPs) have an important task to assess which patients should be referred directly to the hospital for possible sepsis. In this thesis, the current recognition and management of sepsis in adults in the primary care setting is investigated. The main aim of this thesis is to develop an optimal approach to guide GPs when to refer patients to the hospital for possible sepsis. In chapter 2, a questionnaire study among GPs is presented, exploring the clinical decision making process in patients with suspected severe infections. In chapter 3, the prevalence of systemic inflammatory response syndrome (SIRS) criteria at out-of-hours GP cooperatives is investigated, and how these criteria are associated with hospital referral. In chapter 4, a retrospective study of patients admitted to the intensive care unit due to community onset sepsis is described. About half of these patients were assessed at an out-of-hours GP cooperative prior to the hospital admission. One in three patients was not directly referred after the first contact. In chapter 5, the accuracy and feasibility of the respiratory rate measurement by GPs is assessed. GPs can accurately count the respiratory rate, but it is not counted in most patients, and the rate is often underestimated when estimated. The methods of the TeSD-IT study are presented in chapter 6. In this study, clinical information and biomarkers that can be assessed at the bedside (lactate, C-reactive protein and procalcitonin) were prospectively collected in acutely ill adult patients at four out-of-hours GP cooperatives to develop a new sepsis prediction model. Chapter 7 describes the main findings of the TeSD-IT study. Data was used from 357 patients. The primary outcome sepsis within 72 hours of inclusion was met by 42% of the patients. The developed model consists of a count of one point for the following six variables: age>65 years, temperature>38Ā°C, systolic blood pressureā‰¤110 mmHg, heart rate>110/min, saturationā‰¤95% and altered mental status. The respiratory rate as well as the tested biomarkers did not improve the performance of the model. This model resulted in a C-statistic of 0.80. The performance of the model was compared to existing models and externally validated in two emergency department populations. In chapter 8, results of additional testing of biomarkers from the obtained blood samples is presented. Chapter 9 describes an early economic evaluation of the new model. Guidance for referral of adult patients with suspected sepsis in the primary care setting using any cut-off point of the sepsis prediction model is not likely to save costs. In the general discussion in chapter 10, it is concluded that the diagnostic accuracy combined with simplicity makes the new developed model more suited for implementation in primary care than existing sepsis scores. However, the model should not replace clinical judgement of the GP, and it is important to show improvement compared to usual care before formulating new GP guidelines for suspected sepsis.