Incidence and impact of sepsis on long-term outcomes after subarachnoid hemorrhage: a prospective observational study

We prospectively included all adult patients (≥ 18 years) admitted to the neurological intensive care unit (ICU) of the Paulo Niemeyer State Brain Institute (Rio de Janeiro, Brazil) with aneurysmal SAH between April 2016 and May 2018. The Institute is a reference center for neurovascular diseases and receives approximately 70–100 SAH patients per year from a statewide public healthcare network in Brazil.

The present study was approved by the Ethics Review Board of the Evandro Chagas National Institute of Infectious Diseases—Oswaldo Cruz Foundation (Rio de Janeiro). The data were deidentified by assigning each patient a unique identification number. SAH was diagnosed by findings from the initial computed tomography (CT) scan or by xanthochromia in the cerebrospinal fluid if the findings from the CT scan were normal. Patients who were admitted 30 days or more after a hemorrhagic ictus were excluded. We also excluded patients who were pregnant or had a life expectancy of less than 48 h after admission to the ICU.

Demographic data, social and medical history, and clinical features at onset were obtained shortly after admission. Neurological status was assessed with the Glasgow Coma Scale (GCS) [11] and the World Federation of Neurological Surgeons (WFNS) Scale [12]. Systemic disease severity was assessed with the Acute Physiology and Chronic Health Evaluation II (APACHE II) score [13]. Admission and follow-up computed tomography (CT) scans during hospitalization were evaluated by using the modified Fisher scale [14] and for the presence of global cerebral edema and infarction. Daily clinical and laboratory follow-ups were analyzed during the first 14 days of hospitalization or up to the ICU discharge. Data were collected daily during the hospital stay by two investigators (TS and MP), using an electronic case report form (through Research Electronic Data Capture—REDCap). For all patients, the following daily evaluations were also performed: the presence of SIRS and the presence of sepsis or septic shock (according to the Sepsis-3 criteria) [7]; organ dysfunction using the sequential organ failure assessment (SOFA) [15]; and the presence of infection according to clinical and the Centers for Disease Control and Prevention (CDC) criteria. The clinical variables and the diagnosis of infection and sepsis were validated both by the research study team (BG, RT, PK and CR) and through adjudication by an independent infectious disease specialist.

The main outcome was the functional outcome, using the Modified Rankin Scale (mRS) score [16]. Scores were prospectively assessed at hospital discharge and 3, 6 and 12 months post-discharge. Long-term outcomes (at 3, 6 and 12 months) were assessed via a telephone interview with the validated mRS score [17]. The interviews were conducted three times during the course of the study; thus, patients included in the final months of the study only had the 3- or 6-month follow-up. Any patient without data available after hospital discharge was considered lost to follow-up. Functional outcome was dichotomized to a poor outcome (defined as mRS 4 to 6) and a good outcome (mRS 0 to 3). Hospital complications after SAH were diagnosed by the clinical team and adjudicated by the research study team (BG, RT, PK and CR) on a weekly basis. DCI was defined as an otherwise unexplained clinical deterioration (such as a new focal deficit, decrease in the level of consciousness or both) or a new infarct shown on the CT scan that was not visible in the admission or the immediate postoperative CT scan, or both, after the exclusion of other potential causes of clinical deterioration. Postoperative deterioration due to operative complications was defined as any neurological worsening or a new infarct within 48 h after the aneurysm repair procedure. Other complications, such as hydrocephalus (defined in this study as the need for cerebrospinal fluid drainage by external ventricular drain or lumbar puncture for that end), rebleeding (defined as a new brain hemorrhage on CT and not related to a surgical procedure), vasospasm (defined as arterial narrowing on cerebral angiogram or mean velocity higher than 120 cm/s and Lindegaard index higher than 3 on transcranial Doppler) and seizures, were also recorded. Aneurysm rebleeding was defined as an acute neurological deterioration with a new hemorrhage apparent on the CT scan.

Data are presented as means (standard deviations) or medians (interquartile ranges) for continuous variables and as absolute numbers and percentages for categorical variables. Univariate associations were tested by using the Chi-square or Fisher’s exact test for categorical variables, the two-tailed t test for normally distributed continuous variables and the Mann–Whitney U test for nonnormally distributed continuous variables. Logistic regression was used to test the association between sepsis and mortality and functional outcome using known predictors of poor outcome as covariates. An initial Multivariable analysis was performed to determine the relationship between premorbid demographic, admission clinical and radiographic variables, as well as complications and surgical treatment offered, with functional outcome at hospital discharge. We initially considered for the multivariable model variables associated with poor outcome with a p value lower than 0.1 in univariate analysis. Adjusted odds ratios for specific hospital complications, including sepsis, were calculated by sequentially adding each of these factors to the baseline model to evaluate their unique contribution. A Hosmer–Lemeshow test was used to test the goodness of fit of each model. The final model included age, sex and variables that remained significantly associated with the outcome variable. Potential multicollinearity between the parameters of the final regression model was assessed by calculating tolerance and variance inflation factor coefficients. A Kaplan–Meier survival curve was derived from the log-rank test for the septic and nonseptic groups. The significance was set to 0.05 for all analyses. All analyses were performed with commercially available statistical software (SPSS version 19.0; SPSS Inc., now part of IBM Corporation, Armonk, NY, USA).

In total, 149 patients were enrolled in the study. There were no patients admitted to the ICU who met the exclusion criteria during the period of the study. Age ranged from 22 to 79 years (median 55 years), and most patients were female (109–73%). The demographic and baseline characteristics of all patients are detailed in Table 1. In-hospital mortality was 17% (25 patients), and 69 patients (46%) had poor outcomes at hospital discharge (mRS of 4–6).

Fifty-six patients (38%) developed 60 infectious events (four patients had two distinct episodes of infection). The events were classified as follows: fourteen episodes of ventilator-associated pneumonias (VAP), thirteen nosocomial pneumonias, eleven tracheobronchitis, nine meningitis/ventriculitis, six urinary tract infections, two bloodstream infections, one infected pressure ulcer, one dental abscess and three indeterminate infections. Out of all episodes of infection, 27 (45%) were classified as pneumonias. The incidence of sepsis was 28% (41/149 patients), and among those, 18 patients developed septic shock (12% of all patients, 44% of the septic patients). Among septic patients, the initial infection was VAP in nine patients, nosocomial pneumonia in nine, tracheobronchitis in eight, meningitis/ventriculitis in three, urinary tract infections in six, bloodstream infections in two, infected pressure ulcer in one and infections of indeterminate origin in three patients. SIRS was present in 122 patients (82%). The median time between admission and the diagnosis of sepsis was 3 days (interval of 1–14 days). The characteristics of each group, septic and nonseptic, are detailed in Table 2.

In the univariate analysis, age over 55 years (p = 0.049), sepsis (p < 0.0001), poor grade SAH (WFNS scale of 4–5) (p < 0.0001), modified Fisher scale of 3–4 (p = 0.01), hydrocephalus (p < 0.0001), vasospasm (p = 0.004), postoperative neurological deterioration (p = 0.027) and DCI (p < 0.0001) were associated with a poor outcome. Multivariable logistic regression analysis, with outcomes at hospital discharge, revealed that death or functional dependence (mRS of 4 to 6) at hospital discharge were independently associated with sepsis (OR 3.4, 95% CI 1.16–9.96, p = 0.026) even after controlling for WFNS (OR 4.66, 95% CI 1.69–12.88, p = 0.003), hydrocephalus (OR 4.55, 95% CI 1.61–12.85, p = 0.004) and DCI (OR 3.86, 95% CI 1.39–10.74, p = 0.01) (Table 3). SIRS, however, was not independently associated with a poor functional outcome. The data of the multivariable analysis are presented in Fig. 1.

Follow-up data were acquired for 136 patients (13 patients were lost to follow-up—9%, 12 of those patients were in the nonseptic group). Data were available for 95 patients at 12 months (70%), 28 patients at 6 months (20.5%) and 13 patients at 3 months (9.5%). The long-term mortality rate (using the last follow-up available for each patient until 12 months) was 26% (36 out of 136 patients), and the poor outcome rate was 33% (46 out of 136 patients).

Mortality rates and functional outcomes for the long-term follow-up were significantly different in the septic and nonseptic groups, with higher rates in septic patients (log-rank test p < 0.0001). The long-term mortality rate of septic patients was 52.5% (21 out of 40 patients), and that of nonseptic patients was 16% (15 out of 96). Long-term poor outcome rates were 66% (26 out of 40) for the septic group and 21% (20 out of 96) for nonseptic patients. The Kaplan–Meier survival curve is depicted in Fig. 2, and Fig. 3 shows the long-term outcomes, divided by the mRS, in both the septic and nonseptic groups.