1401: ANTIPYRETIC THERAPY IN CRITICALLY ILL SEPTIC PATIENTS: A SYSTEMATIC REVIEW AND META-ANALYSIS

Learning Objectives: Sepsis has poor outcome, and the mortality rate is 30%50% in patients with septic shock. Early diagnosis and treatment of sepsis are important for improving the prognosis. Sepsis is the most common disease associated with coagulopathy. Approximately 20%-40% of all sepsis patients are complicated with disseminated intravascular coagulation (DIC). Therefore, inflammation and coagulation are closely interrelated in sepsis. In 2016, new sepsis definition named SEPSIS-3 was published. We aimed to identify the most predictive marker of 28-day mortality using inflammation and coagulation markers. Methods: This was a single-center retrospective study and included 186 patients with sepsis from January 2013 to September 2015. The definition of sepsis was used according to SEPSIS-3. Inflammatory markers [C-reactive protein (CRP), procalcitonin (PCT), and presepsin (PSEP)] and coagulation markers [platelet count, prothrombin time-international normalized ratio (PT-INR), activated partial thromboplastin time (APTT), antithrombin (AT), D-dimer, thrombin-antithrombin complex (TAT), plamin-2 plasmin inhibitor complex (PIC), protein C (PC), soluble fibrin (SF), and plasminogen activator inhibitor (PAI)-1] were assayed on ED admission. Univariate and multivariate logistic regression analyses were performed to identify independent predictive markers of 28-day mortality. The area under the curve (AUC) and the optimal cut-off value of the most predictive marker were examined. Results:The 28-day mortality rate was 19% (36/186). PSEP, TAT, PC, SF, and PAI-1 were found to be significant predictive markers of 28-day mortality according to univariate analysis. Subsequently, we performed multivariate logistic regression analysis using these five markers as explanatory variables. PAI-1 was found to be the only independent predictive marker of 28-day mortality (P <0.05). AUC of PAI-1 was 0.72, and the optimal cut-off value was 83 ng/ml, with sensitivity and specificity of 75% and 61%, respectively. Conclusions: In conclusion, PAI-1 may be a useful predictive marker of 28-day mortality for sepsis.

Learning Objectives: Sepsis is among the most common diagnosis for hospital admission. Severe sepsis and septic shock can require rescue with extracorporeal membrane oxygenator (ECMO) support. Little is known about outcomes of patients with sepsis and septic shock requiring ECMO support, particularly in the pediatric patient population. Methods: This is a descriptive analysis of pediatric patients presenting with sepsis necessitating ECMO support during 1990 to 2015. Patients were identified from the ELSO database based on assignment of diagnostic ICD9 codes associated with sepsis and septic shock. The following codes were included: Septic shock (785.52); Severe sepsis (995.92); Sepsis (995.91), SIRS (995.9); Toxic Shock Syndrome (040.82), Shock without mention of trauma (785.5) associated with infection, and Septicemia from any organism (038.09). Only the first ECMO run was included for each patient. Patient demographics are described using descriptive statistics with a comparison of survivors and nonsurvivors. Results: A total of 1846 patients presented with one of the above diagnostic codes: 407 patients with septic shock; 168 with severe sepsis; 126 with sepsis; 23 with SIRS; 13 with toxic shock syndrome; 59 shock (other), and 1050 with organism specific septicemia. 707 patients survived to discharge (38.2%). Survivors tended to have a shorter time between intubation and ECMO cannulation, shorter duration of ECMO, and had a higher preECMO pH. Percent survival trended higher between the years of 1990 and 2015. The number of cases per year increased between 1990 and 2015. Conclusions: An increasing number of patients with severe sepsis or septic shock are being placed on ECMO. The survival to discharge amongst these patients has improved over the last two decades. However, these patients continue to have an increased risk of mortality. Contributory factors to survival vs. nonsurvival in this population may include shorter intubation to ECMO cannulation time, shorter duration of ECMO, and pH status preECMO. Further study is required and warranted to identify additional mortality risk factors.

AGREEMENT BETWEEN DIFFERENT ULTRASOUND APPROACHES FOR THE ASSESSMENT OF FLUID RESPONSIVENESS
Sara Crager, Ricky Amii, Caleb Canders, Daniel Weingrow, Stephanie Tseeng, Alan Chiem Learning Objectives: Both inadequate and excessive fluid resuscitation are associated with increased morbidity and mortality. Point-of-care tools, such as ultrasound (US), are increasingly being used by critical care physicians in the assessment of fluid responsiveness (FRes). Controversy remains as to the optimal US parameter for FRes assessment. This study evaluated the consistency of multiple US parameters currently used to predict FRes. Methods: Medical intensive care unit patients for whom FRes determination was required were enrolled in the study (n=29). Fellowship trained sonographers, blinded to clinical context, obtained four US measurements in conjunction with passive leg raise: left ventricular outflow tract velocity time integral (LVOTVTI), carotid Doppler maximal velocity (cVmax), femoral Doppler maximal velocity (fVmax), and inferior vena cava collapsibility index (IVC-CI). The critical care physician caring for the patient was also asked to predict FRes based on clinical judgment. Kappa coefficients (κ) were calculated to determine agreement between the tested parameters. Results: Overall, we found poor agreement between the different approaches to FRes assessment. The only measures that showed fair agreement were LVOTVTI and cVmax (κ=0.34), and fVmax and cVmax (κ=0.26). IVC-CI showed poor agreement with all other measures studied (κ=0 to -0.01). Critical care physicians rated their certainty in their clinical assessment of FRes as high or moderate 80% of the time, however clinician judgment showed poor agreement with all US measures studied (κ= -0.15 to 0.18). Conclusions: Our data suggest that different US approaches may offer variable predictions of FRes in a given patient, and future studies are required to identify the most accurate parameter. Ultimately, a more nuanced approach that integrates clinical context with a comprehensive US assessment including variables such as cardiac contractility, extravascular lung water, and venous return in addition to FRes may be preferable to a single US parameter that provides a cutoff value used to dictate decisions about volume administration. and ClinicalTrials.gov through February 2016. Randomized or observational studies evaluating the effect of antipyretic therapy on mortality in septic patients were eligible for inclusion. Studies were excluded if they included pediatric patients, healthy volunteers, or patients with neurological injury. Two reviewers independently extracted data. The primary outcome was 28-day/hospital mortality. Randomized and observational studies were analyzed separately. Pooled relative risks (RR), odds ratios (OR), or mean differences were calculated using a randomeffects model. Simulation-based sample size analysis was performed to assess the influence of a future study on the meta-analysis results. Results: Of 1125 studies initially identified, 8 randomized studies (1507 patients) and 8 observational studies (17432 patients) were included in the analysis. Antipyretic therapy did not reduce 28-day/hospital mortality in the randomized studies (RR 0.93 (95% CI 0.77, 1.13, I2 = 0.0%)) or observational studies (OR 0.90 (95% CI 0.54, 1.51, I2 = 76.1%)). Shock reversal (RR 1.13 (95% CI 0.68, 1.90, I2 = 51.6%)) and acquisition of nosocomial infections (RR 1.13 (95% CI 0.61, 2.09, I2 = 61.0%)) were also unchanged. Antipyretic therapy decreased body temperature (mean difference -0.38 °C, (95% CI -0.63, -0.13, I2 = 84.0%)), but not heart rate or minute ventilation. Simulation analysis showed that an additional randomized study would require 29,000 patients to achieve 80% power to change the meta-analysis results for mortality. Conclusions: Antipyretic treatment does not significantly improve 28-day or hospital mortality in adult patients with sepsis.

CURB-65 SCORE CORRELATES WITH ICU ADMISSION AND IN-HOSPITAL MORTALITY IN PATIENTS WITH SEPSIS
Huthayfa Ateeli, Faraz Jaffer, Khushboo Goel, See-Wei Low, Bhupinder Natt Learning Objectives: Severity scores in patients with Sepsis are useful for triaging and mortality estimates. A simple bedside score is more useful than complicated calculations for clinicians in the emergency department and medical wards. Mortality in Emergency Department Sepsis (MEDS) score is validated in patients with sepsis in the emergency department. CURB-65 is validated in patients with community-acquired pneumonia but not in sepsis. CURB-65 is a simple bedside tool that has many common elements with new sepsis identification score-qSOFA. Methods: This study was approved by the Institutional Review Board of the University of Arizona. Clinical data was collected to calculate the MEDS and the CURB-65 scores on adult patients with sepsis syndromes admitted to the Banner-University Medical Center between November,2013 and July, 2015. Patients were identified using ICD-9 scores (995.91, 995.92, 785.52). These patients were studied using CURB-65 scores and their ICU admission and in-hospital mortality was ascertained. Results:A total of 853 patients were identified with Sepsis Syndromes. 62 patients had a CURB-65 score of 0, corresponding mean MEDS of 4.09, 27% ICU admission and 4% in-hospital mortality. 152 patients with CURB-65 score of 1, mean MEDS of 5.64 had 56% ICU admission and 6.5% in-hospital mortality. 209 patients were identified with a CURB-65 score of 2, mean MEDS of 8.63 of which 69% had ICU admission had 23% in-hospital mortality. 237 patients had a CURB-65 score of 3, mean MEDS of 10.9-84% were admitted to the ICU and had 26% mortality. 158 patients had a CURB-65 score of 4, mean MEDS of 12.9-90% had ICU admission and 40 % mortality. 35 patients had CURB-65 of 5, mean MEDS of 15.4-95% ICU admission and 63% mortality. Conclusions: A simple bedside score, the CURB-65 score, can predict higher mortality corresponding to a higher CURB-65 score in patients with sepsis due to any cause. Further validating studies are required on this topic.

DOES CIRRHOSIS AFFECT THE PROGNOSTIC UTILITY OF LACTATE FOR SEVERE SEPSIS?
Stephanie Taylor, Anthony Roohollahi, Colleen Karvetski, Megan Templin, Brice Taylor Learning Objectives: Elevated serum lactate level and impaired lactate clearance are associated with increased morbidity and mortality in patients with severe sepsis and septic shock. However, serum lactate can be elevated due to other reasons, including impaired hepatic clearance. We investigated whether the initial lactate and lactate clearance (LC) at 6 hours was predictive of mortality in severe sepsis patients with cirrhosis. Methods: We retrospectively studied 3786 patients enrolled in a severe sepsis database between 2014 and 2016. Cirrhosis was determined by ICD-9 and ICD-10 codes. Lactate was measured at time of sepsis identification and 6 hours later. Lactate clearance (LC) was defined as the difference between initial and subsequent lactate divided by the initial lactate x 100%. Results: Cirrhosis was present in 445 (11.4%) patients. Patients with cirrhosis had a higher initial lactate and lower LC at 6 hours than in non-cirrhotic patients. In the subset of patients with cirrhosis, initial lactate and LC at 6 hours were strongly associated with mortality (P < 0.0001, P < .0001). The median lactate in cirrhotic survivors was lower than that of cirrhotic non-survivors (4.5 vs 6.8, P<.001). The median LC in 6h in cirrhotic survivors was higher than that of cirrhotic non-survivors (43% vs. 13%, P<.0001). After adjusting for confounding factors including severity of illness and renal failure, initial lactate and LC at 6h remained associated with hospital mortality but the effect was small (adjusted OR 0.91, 95% CI 0.86-0.95 for initial lactate; adjusted OR 1.005, 95% CI 1.001-1.009 at 6h). Conclusions: Initial lactate and LC at 6 hours is predictive of mortality in patients with cirrhosis but the adjusted effect is relatively small.

Alan Heffner, Colleen Karvetski
Learning Objectives: Patients requiring hospitalization for infection represent a large modern healthcare burden. We aimed to identify the incidence, timing, and outcomes of patients requiring ICU transfer following emergency department (ED) admission to a general ward for acute infection. Methods: Adult patients tracked in our surveillance of ED admissions for an infection composite index comprised of ICD9/10 diagnoses present on admission served as our data source. Patient discharged from one of five urban hospitals within our healthcare system between January 2014 and June 2016 were used for analysis. Outcomes of infected patients admitted to a general ward with subsequent ICU transfer were compared to patients directly admitted to an ICU. Results: Among 18,145 adult ED patients admitted for infection (44% male, mean age 65 ± 18years), 5,312 (29.3%) required ICU admission during their hospital stay. 713 (5.3%; 95% CI: 4.9-5.7%) were transferred to the ICU following initial admission to a general ward. Transfer to the ICU occurred at a mean of 60 hours and median of 32 hours (IQR 13-67) following hospital admission; 25% of patients were transferred within 13 hours, 50% within 33 hours, 75% within 67 hours, and 90% within 139 hours. Transfer to the ICU was not associated with increased mortality (15.0% vs 17.5%; p=0.08) or increased ICU admission APACHE IV Score (68.6 vs 70.1; p=0.21) but was associated with increased ICU LOS (4.1 vs 3.3 days; p<0.0001) and hospital LOS (11.3 vs 7.6 days; p<0.0001). Time from admission to ICU transfer was associated with increased hospital mortality (p=0.002). Odds ratio of death for 6 hour and 24 hour increments of time from admission to ICU transfer were 1.04 (95% CI:1.02-1.07) and 1.18 (95% CI:1.06-1.31) respectively. These associations persisted after adjusting for severity of illness (APACHE IV). Conclusions: A small but important fraction of patients hospitalized for acute infection deteriorate and require ICU care following initial admission to a general ward. Time from admission to ICU transfer is associated with adverse outcome.

SEASONAL VARIATIONS IN SEVERE SEPSIS HOSPITALIZATIONS
Ankit Sakhuja, Gagan Kumar, Saraschandra Vallabhajosyula, Kianoush Kashani Learning Objectives: Severe sepsis is a life-threatening clinical syndrome which is the third leading cause of death in the United States. Though seasonal variation in heart disease is well documented, there is a paucity of such contemporary studies for severe sepsis hospitalizations. Knowledge of seasonal variations is not only important from public health perspective, but also for policy making and future research. We therefore designed this study to look into seasonal variation in epidemiology and outcomes of severe sepsis hospitalizations. Methods: We used data from NIS database from years 2000 through 2013. We identified severe sepsis hospitalizations based on ICD-9 codes as per previous literature. Month of each hospitalization was identified. Independent effect of admission month was estimated using a multivariable logistic regression model that was adjusted for patient age, sex, race, primary payer, charlson's co-morbidity index, individual acute organ dysfunctions, hospital location, teaching status, region, bedsize and volume. Results: Of estimated 10,266,341 severe sepsis hospitalizations over the 14 years studied, the highest number of daily admissions was in winter months with