Despite advances in the quality of health care, deaths can occur after a patient has survived treatment for a serious illness in an intensive care unit (ICU). Post-ICU mortality rates described in the literature can range from 5% to 27% [1-5]. Mortality rates after leaving the ICU can be considered indicators of health care quality. Several factors have been described associated with death in the short-term after ICU discharge. Most of these risk factors are associated with age, disease severity, and length of stay in the ICU, in addition to infections and the need for unplanned readmissions [6-8]. Health care-associated infections (HAIs) are worrisome complications during an intensive care stay and may be associated with increased mortality in ICU survivors [7,9]. Another risk factor that is associated with mortality after an intensive care stay is readmission to the ICU, which can increase intra-hospital mortality by ten times [8]. It is estimated that for every 100 patients discharged from the ICU, 4 to 6 will be readmitted to this sector, noting that between 3 and 7 of these patients will die before hospital discharge, thus demonstrating the high risk in this transition of care [10]. It is estimated that the hospital mortality rate for patients who require readmission to the ICU is 34.3% [9].

Decisions to discharge patients from the ICU should take into consideraticon possible risk factors for hospital mortality present at that moment, to allow safer transition of care. Prognostic scores are widely used, and these tools can contribute as an indicator of the performance of care provided by the units and hospitals through severity-adjusted mortality rates. A prognostic score that is widely used and with good performance is the Simplified Acute Physiology Score (SAPS) III, which corresponds to the clinical analysis of the patient on admission, and is scored according to demographic, clinical, and physiological variables [11,12]. The current study aims to describe the hospital mortality rate and assess risk factors for death in surviving patients discharged from an ICU at a university hospital.

During the study period, 1,596 patients were admitted to the ICU and of these, 537 patients died in this unit. Of the 1,059 surviving ICU patients, 33 were excluded for incomplete data and 1 for being under 18 years of age (Figure 1). Data were collected from 1,025 medical records. The median age was 62 years (IQR, 46.0–73.5 years), with a predominance of males, 573 (55.9%). The length of hospital stay of these patients had a median of 17 days (IQR, 8–30 days) and the SAPS III severity score presented a median of 47 (IQR, 35.0–62.5).

Of the total number of patients studied, 314 (30.6%) had some type of infection at ICU discharge, 92 (9%) were readmitted to the unit, and 212 (20.7%) died. Considering the patients who developed some type of infection, in 103 (32.8%) it was not possible to identify the microorganism, 78 (24.8%) were sensitive microorganisms, 81 (25.8%) were XDR, 50 (15, 9%) MDR, and 2 (0.6%) PDR. The most frequently found infectious foci were pulmonary, with 253 cases (77%), followed by the urinary tract with 63 (18.1%) and the surgical site with 24 (6.4%).

There were 588 (57.4%) unplanned ICU admissions and 437 (42.6%) planned admissions in the study period. The most frequent admission diagnoses were 153 lower limb osteosynthesis (14.9%), 128 digestive tract surgeries (12.5%), 84 intracranial hemorrhages (8.20%), 81 sepsis (7.9%), 52 neurosurgeries (5.1%), 44 peripheral vascular surgeries (4.3%) and 44 head traumas (4.3%).

When assessing vital status at hospital discharge, the median age of non-survivors (67.5 years) was higher compared to survivors (60 years, P<0.001). There was no association between sex and hospital mortality (P=0.103). The length of hospital stay and the SAPS III were higher among non-survivors (P<0.001). The presence of infection and readmission to the ICU were associated with hospital death (P<0.001). When analyzing the 212 cases of death, 117 (55.2%) occurred in infected patients; 28 (13.2%) by sensitive microorganisms, 30 (14.2%) by unidentified microorganisms, and 59 (27.8%) by resistant microorganisms (Table 1).

In the logistic regression, the vital status on hospital discharge was considered as the dependent variable, and age, length of hospital stay, SAPS III, presence of infection, readmission to the ICU, and microorganisms were considered as covariates. The variables that remained in the regression model were age (OR, 1.029; 95% CI, 1.01–1.04; P<0.001), length of hospital stay (OR, 1.011; 95% CI, 1.003–1.019; P=0.004), SAPS III (OR, 1.042; 95% CI, 1.031–1.054; P<0.001), presence of infection (OR, 1.251; 95% CI, 1.081–1.447; P=0.003), and readmission to the ICU (OR, 11.659; 95% CI, 6.522–20.842; P<0.001) (Table 2). Kaplan-Meier analysis shows a median 49,000 days (95% CI, 42.987–55.013 days) after ICU discharge and 200 death events in 90 days (Figure 2).

The present study describes factors associated with hospital mortality in ICU survivors. The presence of infection on discharge from the ICU and readmissions demonstrated the greatest weight for the risk of death for these patients. These results can support decisions on optimizing HAI prevention and treatment, in addition to the allocation of resources and the prioritization of patients at a higher risk for intermediate care units after leaving the ICU [14,15].

Previous studies have shown that several factors may be associated with worse outcomes after ICU discharge. The intensity of therapeutic interventions and nursing overload at ICU discharge were associated with an increase in hospital mortality [3]. A systematic review described a readmission score that predicts adverse events after ICU discharge. The Minimizing ICU Readmission score takes into account several factors for this prediction, such as the SAPS II prognostic score at ICU admission, the use of a central venous catheter during the ICU stay, the SOFA score at ICU discharge, the presence of SIRS in the last 2 days of ICU stay, and discharge from the ICU at night [16].

Although the prognostic scores present moderate performance, with limitations in their calibration in low and middle-income countries, the SAPS III remains the most applied prognostic assessment tool in several countries [17,18]. The SAPS III calculated at ICU discharge could better reflect the risk of hospital death, but it is not common for units to have this data in routine care furthermore, SAPS III was developed to predict hospital mortality, not ICU mortality. In our study, we chose the SAPS III at ICU admission because it is the score available in the data source and because it is adopted by the institutional policy to evaluate the unit's performance and, therefore, can be easily incorporated to identify patients at a high risk of an unfavorable outcome at ICU discharge.

In a recent study of patients affected by coronavirus disease 2019 (COVID-19), age was the factor identified as associated with increased mortality [19], similar to what was found in our sample, where each year of age increased the chance of post-ICU death by approximately 3%. On the other hand, Madotto et al. [20] describe that the presence of comorbidities and limitations of life support treatments resulted in worse outcomes in ICU survivors.

Our post-ICU discharge mortality can be considered high when compared with results from Europe and New Zealand [21,22]. Although the research institution has implemented follow-up of surviving ICU patients by a rapid response team led by an intensive care physician, there is a high demand for ICU beds that pressures early discharge and there is no availability of intermediate care beds. Reports from low- and middle-income countries show similar results [5].

The prevalence of death in infected patients who survived the ICU was high in our study. The increased risk of post ICU mortality in patients with infections has been pointed out by other authors [23-26]. It is estimated that infection by MDR increases hospital mortality by approximately 1.7 times [23,24], which is similar to the results of the current study.

Pneumonia is the most frequent focus of infection reported by other studies [27-29], and this result is consistent with our findings. Klebsiella species was the most commonly isolated microorganism in our patients. This bacterium is related to almost a third of infections caused by Gram-negative bacteria in general, and presents high resistance to antimicrobials [30,31]. Other Gram-negative microorganisms responsible for HAIs are Enterobacteriaceae, Pseudomonas aeruginosa, Acinetobacter baumannii, and Stenotrophomonas maltophilia. The frequency of pneumonia caused by Klebsiella pneumoniae resistant to carbapenems is above 25%, while the frequency for P. aeruginosa is between 20 and 40% and for A. baumannii, between 40% and 70% [32].

Readmission to the ICU is associated with higher mortality and longer hospital stays [33]. Patients surviving readmission should be monitored and, preferably, remain in intermediate care units. The worse prognosis of patients requiring readmission could be associated with decompensation in the ward, which may result in late intervention. The use of intermediate care units can meet the needs of these higher-risk patients who require surveillance [14]. The length of hospital stay, as a risk factor for death, was also demonstrated in the present study, corroborating studies that show that a period longer than 6 days exposes the patient to more infectious risks [34].

It is possible to reduce readmission rates and increase the safety of surviving ICU patients through the implementation of rapid response teams, in view of the follow-up performed by an intensive care physician in the wards of patients discharged from the ICU [35]. It is estimated that implementation of the rapid response team is associated with a 20.6% decrease in the mortality rate [36].

The study population composed exclusively of ICU survivors, in addition to the sample size, can be considered strengths of the present study. Limitations of the study are the fact that it is a retrospective and single-center analysis, which limits its external validity. Also, a limited number of variables were investigated as risk factors for mortality. Other factors that can affect mortality, such as underlying comorbidities, the use of mechanical ventilation, renal replacement treatment or vasoactive drugs were not included in the regression analysis.

In conclusion, a high mortality rate was observed in patients who survived hospitalization in the ICU and the factors associated with hospital mortality were age, SAPS III, presence of infection, readmission to the ICU, and length of hospital stay.