Hidden hospital mortality in patients with sepsis discharged from the intensive care unit

Objective To evaluate the impact of the presence of sepsis on in-hospital mortality after intensive care unit discharge. Methods Retrospective, observational, single-center study. All consecutive patients discharged alive from the intensive care unit of Hospital Vila Franca de Xira (Portugal) from January 1 to December 31, 2015 (N = 473) were included and followed until death or hospital discharge. In-hospital mortality after intensive care unit discharge was calculated for septic and non-septic patients. Results A total of 61 patients (12.9%) died in the hospital after being discharged alive from the intensive care unit. This rate was higher among the patients with sepsis on admission, 21.4%, whereas the in-hospital, post-intensive care unit mortality rate for the remaining patients was nearly half that, 9.3% (p < 0.001). Other patient characteristics associated with mortality were advanced age (p = 0.02), male sex (p < 0.001), lower body mass index (p = 0.02), end-stage renal disease (p = 0.04) and high Simplified Acute Physiology Score II (SAPS II) at intensive care unit admission (p < 0.001), the presence of shock (p < 0.001) and medical admission (p < 0.001). We developed a logistic regression model and identified the independent predictors of in-hospital mortality after intensive care unit discharge. Conclusion Admission to the intensive care unit with a sepsis diagnosis is associated with an increased risk of dying in the hospital, not only in the intensive care unit but also after resolution of the acute process and discharge from the intensive care unit.

that, 9.3% (p < 0.001). Other patient characteristics associated with mortality were advanced age (p = 0.02), male sex (p < 0.001), lower body mass index (p = 0.02), end-stage renal disease (p = 0.04) and high Simplified Acute Physiology Score II (SAPS II) at intensive care unit admission (p < 0.001), the presence of shock (p < 0.001) and medical admission (p < 0.001). We developed a logistic regression model and identified the independent predictors of inhospital mortality after intensive care unit discharge.
Conclusion: Admission to the intensive care unit with a sepsis diagnosis is associated with an increased risk of dying in the hospital, not only in the intensive care unit but also after resolution of the acute process and discharge from the intensive care unit. had limitations of care due to the severity of their clinical condition, whether it was acute or chronic. (5) However, a large number died unexpectedly, which seems to be related to both comorbidities and the acute clinical insult. (2,6) The decision to discharge a patient to the ward is often based on clinical criteria, namely, "stabilization", which is often poorly validated. The use of such criteria may be especially important in populations at risk of dying in the ward after ICU discharge. With a careful approach to the data presented by some epidemiological studies, (2,6) it is possible to observe a significant excess of mortality in the ward among patients admitted to the ICU with sepsis. This excess may be related to premature discharge and suboptimal therapeutic approaches and care on the ward. (7,8) Identifying these high-risk populations should be of paramount importance to improve the process of care, (9) namely by creating objective criteria for discharge or for the use of progressive transition. (10) In this study, we evaluated the mortality and hospital length of stay of patients discharged from the ICU without an indication for limitation of care. We measured the impact of the presence of sepsis on admission to the ICU on in-hospital mortality (i.e., mortality that occurred after being discharged alive from the ICU). We also evaluated potential risk factors for in-hospital, post-ICU mortality, focusing on those that were identifiable during hospitalization in the ICU.

METHODS
We performed a single-center, retrospective, observational study to assess in-hospital mortality after ICU discharge. The Hospital Vila Franca de Xira Research and Ethics Committee approved the design of the study. Informed consent was waived due to the observational and retrospective nature of the study.
All consecutive patients discharged alive from the ICU of Hospital Vila Franca de Xira (Portugal) for one year, from January 1 to December 31, 2015, were included. Patients for whom there was a decision to limit of care or to preclude readmission to the ICU (N = 18) were excluded from further analysis.
Hospital Vila Franca de Xira is a small hospital near Lisbon (260 beds). As expected, the ICU receives mainly medical and surgical patients (roughly 75% -25%), with a very low rate of trauma patients. The mean Simplified Acute Physiology Score (SAPS) II score of patients is usually high (46.4 in 2015). The ICU has 8 level III beds. An intermediate care unit with 12 beds is associated with the same department and has common nursing and medical staff.
In this study, we collected demographic data, including sex, age, type of admission (either surgical or medical) and location (emergency room, operating room, ward). Laboratory data on admission (lactate) and clinical information (the presence of shock, comorbidities, such as hypertension, diabetes mellitus, chronic renal disease, heart failure and active neoplasia) were recorded. Disease severity on admission to the ICU was assessed by the calculation of the SAPS II. Sepsis was defined as the presence of infection (suspected or microbiologically documented) associated with systemic manifestations, including organ failure, (1) which led to the provision of antibiotic therapy, according to the physician in charge. The infection was categorized as acquired in the community or in the hospital, according to commonly used criteria. (1) When a patient experienced more than one infectious episode, each was recorded and accounted for. Each patient was included only once (first episode of hospitalization in the ICU).
All data were recorded in a file specially created for this study.

Statistical analysis
Continuous variables are expressed as the mean ± standard deviation according to their distribution; discrete variables are expressed as the total number (percentage).
Univariate analysis was performed using the chisquare test (for discrete variables) or Student's T or Mann-Whitney tests (for continuous variables) according to the data distribution. The variables that were identified in the univariate analysis as potentially associated with in-hospital mortality after ICU discharge (considering a p < 0.1) were included in a logistic regression model. All these variables were assessed 2 to 2 for the exclusion of significant correlations. A coefficient r < 0.3 was considered sufficient to exclude this correlation. In cases where a correlation was found, the variable considered to be of the greatest clinical importance was included in the model. For the logistic regression model, the results are expressed as odds ratios (OR) with 95% confidence intervals (95%CI).
As the SAPS II score is predictive of in-ward mortality, and we believe that the inclusion of this score can compensate for possible imbalances that may exist in patients' admission characteristics. In fact, the SAPS II score accounts for age, type of admission (either surgical or medical), presence of significant comorbidities and clinical severity on admission to the ICU.
All statistical tests were two-tailed, and a value of p < 0.05 was considered statistically significant. Statistical analysis was performed using the Statistical Package for Social Science (SPSS) version21 (SAS Institute, Cary, NC) software package.

RESULTS
During the study period, 473 patients were discharged alive from the ICU without a decision to limit care or to preclude ICU readmission. All of these patients were retrospectively screened for the presence of infection and sepsis at ICU admission (N = 140, 29.6%).
The clinical characteristics of patients with or without infection upon admission to the ICU are presented in table 1.
Infected patients more often had a severe disease on admission (significantly higher SAPS II and the presence of shock). Intensive care unit-acquired infections were also more frequent in these patients (15% versus 4.8%, p = 0.001). In contrast, comorbidities were more commonly reported in noninfected patients (Table 1).
After discharge from the ICU, 61 (12.9%) patients died in the hospital, usually on the ward. As expected, the SAPS II was lower in patients who were discharged alive from the hospital (37.1 ± 15.7 versus 51.5 ± 16.4, p < 0.001) ( Table 2).
The patients who died were significantly older (71.2 ± 15.5 versus 66.2 ± 14.6; p = 0.02), more frequently (68.9%) were male (p < 0.001), more often were admitted to the ICU for a medical reason (88.5%) and had non -clinically relevant higher median lactate levels on Again, the presence of sepsis at admission to the ICU was associated with the risk of dying in the hospital, after   ICU discharge (21.1% versus 10.7%, p = 0.001), but not with ICU readmission (Table 1). In our cohort, patients discharged after an episode of bacteremia had a higher mortality risk (Table 3). Having had a respiratory or abdominal focus of infection was also associated with an increased risk of dying on the ward. Interestingly, of the other studied comorbidities, namely, active cancer, HIV infection, chronic heart failure, arterial hypertension or diabetes mellitus, none were associated with in-hospital mortality after discharge from the ICU (Table 2).
Accordingly, we developed a logistic regression model to evaluate the independent associations with the risk of dying in the hospital, after ICU discharge. We found that the presence of sepsis on admission to the ICU (OR 2.32, 95%CI 1.28 -4.24), male sex (OR 2.26, 95%CI 1.21 -4.24) and the SAPS II score (OR per point 1.05, 95%CI 1.03 -1.07) were independent predictors of in-hospital death (Table 4). This association of sepsis with mortality in the ICU is well known (6) and may persist even after hospital discharge, (11) eventually becoming associated with a high incidence or worsening of a previously existent cardiovascular disease. (12) Although the relationship of cardiovascular disease risk with sepsis decreases with time, it may persist for several years. (12) In this study, we focused only on mortality in the hospital after ICU discharge. We believe that this risk factor, sepsis on admission to the ICU, should be taken into account when allocating resources, namely, in the admission of patients to functional units with greater availability of human resources, the use of progressive stepdown or hand-over programs or even the prolongation of hospitalization in the ICU, to help reduce morbidity and mortality. Our data suggests that such measures are especially important for patients admitted with bacteremia or an abdominal or respiratory infection.
Some patients die soon after discharge from the ICU, and most of these deaths are associated with a plan to limit care and are to be expected. In our study, we identified and excluded 18 patients in this situation. However, other patients die unexpectedly on the ward, and these death could potentially be avoided with a better process of care. (13) Several patients are discharged from the ICU to an intermediate care unit to receive a higher standard of care than can be provided on the ward. Nevertheless, a study of 690 patients discharged from the ICU failed to demonstrate an association between ICU discharge to an intermediate care unit and a better outcome. (14) Of note, another multicenter study showed a survival benefit for the overall population when intermediate care units were available in the hospital (adjusted OR 0.63, 95%CI 0.45 to 0.88), although this benefit was not necessarily related to admission to those units. (15)

DISCUSSION
In this study, we report an association between sepsis on admission to the ICU and hospital mortality that persists after resolution of the acute disease and discharge from the ICU (OR 2.32, 95%CI 1.28 -4.24) and is independent of comorbidities or age.
We clearly need to better understand the prognostic factors associated with this late mortality risk to identify patients who may benefit from this resource. In a British study, Daly et al. (16) suggested that prolonging the hospitalization of critically ill patients with sepsis in the ICU for an additional 48 hours could reduce hospital mortality by approximately 39%. In addition, even the hour and weekday of discharge may influence the outcome of these patients, (17) probably in relation to the varying availability of resources in the hospital. Discharge to an intermediate care unit often occurs at late hours, (18) which may jeopardize the potential benefits of a progressive step-down approach.
Lactate has also been proposed as a possible prognostic factor for patients admitted to the ICU, (19) especially the septic population. (20) However, there are multiple pitfalls in the interpretation of either absolute or relative lactate levels, (21) and their relationship with outcomes may not be straightforward, as we found in our study. Another potential prognostic factor is the C-reactive protein concentration at ICU discharge, which may help to identify patients at risk. (22,23) The in-hospital mortality rate of patients discharged from the ICU in our study was high, 12.9%; this was in sharp contrast with a meta-analysis published in 2014, where the mean hospital mortality after ICU discharge was only 6.8%. (4) However, these numbers are common in Portuguese ICUs. (2,22) We believe that these higher values were mostly related to the case mix, namely, the high prevalence of comorbidities in the Portuguese ICU population, (2) the predominance of patients admitted for medical reasons or sepsis and even the high prevalence of patients admitted from the ward.
On the other hand, a higher late mortality on the ward among patients admitted to the ICU with sepsis was evident in other studies, although it was rarely mentioned. In the International Study of Prevalence and Outcomes of Infection in Intensive Care Units (EPIC) II, (6) the in-hospital mortality after ICU discharge was 7.3%. This mortality rate was clearly higher in those patients who were deemed to be infected on the study day: 10.3% in infected patients versus 4.6% in others (estimated OR 2.4, p < 0.001). (6) In the Portuguese multicenter study Infection on Admission to the Intensive Care (INFAUCI), the inhospital mortality after ICU discharge was 11.5%, but it was higher (14.2%) in patients who had an infection on admission (9.6% in the others, OR 1.6, p < 0.001). (2) In another cohort of surgical patients, the mortality risk in the first year after hospital discharge was also increased in the septic population (OR 1.71, 95%CI 1.46 -2.00), particularly among the elderly (> 75 years) and those with the highest Charlson scores. (24) In fact, this persistent mortality risk among the septic population may be related to the continuation of the inflammatory state that persists even after the clinical resolution of sepsis, a risk that was unveiled in patients addmited with community-acquired pneumonia. (25) In a multicenter study, GenIMS, (25) high levels of IL-6 at hospital discharge were associated with a higher 3-month mortality that may be related to a high prevalence of new or worsening cardiovascular disease, which may persist for several years. (12,26) Consequently, we believe that a more aggressive approach to care for these patients, not only in the short term but also in the long term, may lead to a reduction in mortality and morbidity.
Our study has several limitations: it was a retrospective, single-center study. Sepsis was identified by the physician in charge (at ICU discharge), and the risk of misclassification was real. Moreover, our exclusion of patients with a limitation of care were based only on the decision made at ICU discharge and did not consider eventual additional decisions made later, on the ward.
We adjusted the mortality risk to the SAPS II score. However, these scores is determined at ICU admission and do not include some important predictors of mortality at ICU discharge, such as the level of consciousness, the presence of tracheostomy, the need for noninvasive ventilation, ICUacquired muscular weakness, and ulcerations.
However, our study focused specifically on the inhospital period after ICU discharge (usually "blind" in epidemiological studies), excluded patients with a plan to limit care (who naturally have higher mortality rates and contaminate the interpretation of the results), included consecutive patients throughout a whole year (thus eliminating seasonal variations that may exist) and evaluated data for different comorbidities. Furthermore, the identified mortality was relatively high, which allowed the impact of different risk factors to be discriminated.

CONCLUSION
In conclusion, we found an association between being admitted to the intensive care unit with an infectious disease and sepsis and an increased risk of dying in the hospital, even after discharge from the intensive care unit after the acute process was resolved. This association was independent of age or comorbidities.