Our analysis showed that progression from infection to sepsis, severe sepsis or septic shock is strongly determined by the source of infection, independent of the initial values of SOFA and APACHE II scores. Similarly, this progression between clinical stages jointly with severity scores, age and the respiratory source of infection were the main determinants of mortality during the first week. We use specialized statistical models in order to fully explore the timeline of progression of the clinical stages of sepsis. Previously, Clermont et al. worked on dynamic microsimulation models to predict the temporal patterns of multiple outcomes in critically ill patients . However, such approaches address mainly issues of prediction and forecasting and are not suitable enough for identification of individual covariates.
On the other hand, it has been reported in the literature that mortality has differential behavior according to the complexity and the source of the infection; with mortality rates usually higher at the stages of severe sepsis and septic shock, as well as in patients with pneumonia [1, 3, 5, 8, 24–27]. Our results confirm that septic shock as initial clinical stage, in addition to several sources of infection, increases the 28-day hazard of death. However, when stage shifts over the first week are assessed, it is clear that progression to severe sepsis and/or to septic shock are also independent determinants of mortality and the respiratory source of infection remained as a significant prognostic factor.
In order to compare our findings with those of other investigations, it is necessary to take into account all the differences of our study population. Our cohort consists of patients with community- and hospital-acquired infections, hospitalized in ICU or in general wards, whose mean age shows a young population, 33% of whose individuals have no comorbidities. In terms of the quantification of severity by SOFA and APACHE II scores, our data are comparatively low in comparison with those of other populations, even in the septic shock group [7–9, 28–30].
Alberti et al.  published a multicenter study that enrolled 3443 ICU patients diagnosed with infection, 1531 of whom did not present with severe sepsis or shock. Eleven percent (n = 167) and 13% (n = 201) of these patients progressed to severe sepsis and septic shock, respectively, at some moment of their ICU stay before day 30. One of the chief factors of that progression to severe sepsis or septic shock was the source of infection; the risk was highest in cases of bacteremia (HR = 1,81; 95% CI =1,18-2,76), followed by peritonitis (HR = 1,51; 95% CI = 1,07-2,13) and pneumonia (HR = 1,47; 95% CI = 1,18-1,82). These results are in keeping with our own. Besides including only ICU patients and not estimating the effect of clinical progression on mortality, that study carried out an analysis based only on the time elapsed between admission and the appearance of severe sepsis or septic shock using a Cox model with competing risks. Such a model is not able enough to capture daily changes in clinical stages, which are outcomes that can appear repeatedly in a given patient during his/her hospital stay. Moreover, the authors found that growth of Gram-positive cocci and Gram-negative bacilli in blood increases the risk of progression to severe sepsis or septic shock during the first 30 days of ICU. In our study population, in which infections caused by Gram-negative bacilli were predominant in contradistinction to other current series across the world [11, 31], the type of microorganism was not associated with progression to more advanced clinical stages. The apparent “protective association” of Gram-negative bacilli in blood with first-week mortality in our cohort is explained by a survival bias, given by a higher mortality in patients with Gram-negative bacteremia beginning during the second week. This was confirmed by the analysis with 28-day mortality, which showed no association with this type of microorganism (HR = 1,02; 95% CI = 0,75-1,40).
Dremsizov et al.  carried out a study in patients presenting to the emergency room with pneumonia, aimed at determining the onset and timing of severe sepsis and the ability of the Systemic Inflammatory Response Syndrome (SIRS) criteria and the Pneumonia Severity Index (PSI) to predict its development. Of 1339 patients with pneumonia, 882 did not have severe sepsis on admission, and 20,6% (n = 182) of these patients progressed to that stage at some moment of their hospital stay. The authors found that using 2 or 3 SIRS criteria does not predict the risk of progression to severe sepsis or septic shock or 30-day mortality, in contradistinction to PSI, which was associated with the development of severe sepsis. Besides considering only a specific type of infection and a specific type of risk score, that study did not analyze the factors from the perspective of their change over time, neither for exposure variables nor for outcome variables. Glickman et al.  carried out an investigation in patients presenting to the emergency room in order to determine the incidence, the mortality, and the factors associated with progression over the first 72 hours from sepsis to septic shock. The study included 472 patients without shock at the moment of evaluation in the emergency room, whose median age was 52 years and median APACHE II score was 9 points. The most common source of infection in the whole population was the respiratory, and the factors associated with progression to septic shock were catheter-related infection, age, female sex, temperature, and chronic lung disease. Besides the relative lack of ability of the logistic regression analysis to explore the behavior of variables that change over time, the study might have limitations related with the sample size necessary for a multivariable analysis.
The strengths of our study include the great diversity of clinical and epidemiologic aspects of its population, the quality of the data and the reproducibility of the definitions. Our study showed that severity scores not only increase with higher complexity of the infection [24–26] and predict mortality or organ dysfunction [13, 14], which is what they were designed for, but also determine the risk of progression to more severe clinical stages. The latter suggest an additional advantage of severity scores, as potential tools for exploring the complexity of sepsis physiopathology. The limitations of our study include the lack of assessment of the impact of adequate treatment and resuscitation strategies on progression from one clinical stage to another [33, 34], and the absence of biochemical markers used in clinical practice such as C-reactive protein, procalcitonin, or lactate.