Study design and setting
This single-center, retrospective cohort study was approved by the Institutional Review Board of the Taipei Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation on August 20, 2019 (Protocol No.: 08-X-068) and was conducted in accordance with the guidelines of the amended Declaration of Helsinki. Waiver of informed consent was reviewed and approved by the Institutional Review Board, and patients’ data anonymity was protected. The study was performed in a facility settled within a 1000-bed tertiary referral hospital containing 45 medical intensive care unit (MICU) beds. The nurse-to-patient and respiratory therapist-to-patient ratios in the MICU were 1:2 and 1:10, respectively. Experienced physicians served all patients with pulmonary and critical care medicine issues, and in-hospital night coverage was provided by resident physicians. Consultation services were provided mostly by medical and surgical physicians.
Selection of Participants
This study enrolled patients with pneumonia and suspected septic shock, who were treated at the MICU of the Taipei Tzu Chi Hospital from July 1, 2013 to June 30, 2019. The ED electronic medical records were screened. Especially, the first three codes of the international classification of diseases (ICD)-9-CM codes 785.52, 995.91, 995.92, and the ICD-10-CM codes A41, R65.20, R65.21 were screened for sepsis and septic shock. Additionally, the ICD-9-CM codes 481 to 483, 485, 486, and the ICD-10-CM codes J13 to J18 were screened for pneumonia.
Pneumonia classifications were based on risks of MDROs. HCAP was defined based on the patients' medical history of hospitalization for ≥2 days within 90 days prior to the diagnosis of infection or showing any of the following events within 30 days prior to the diagnosis of infection: attending to a healthcare institution, living in a nursing home (nursing home-associated pneumonia [NHAP]), or receiving medical treatments (wound care, chemotherapy, intravenous injections, or hemodialysis [HDAP]) [3, 30, 31]. CAP was defined based on the patients' history of not meeting any of the HCAP criteria. NHAP was further divided into the high-risk and low-risk groups (patients from the long-term care facilities with high and low prevalence of MDROs).
Screening of risks for the MDROs started from the time of the ED visit during the study periods. According to the MDROs risk phenotype classifications, pneumonia was further divided into the high-risk (high-risk HCAP and NHAP, and HDAP) and low-risk (low-risk NHAP and CAP) groups for the MDROs (Table 1) .
On July 1, 2016, the ED in the Taipei Tzu Chi Hospital advocated early screening of the MDROs and healthcare classifications for patients who were suspected of having sepsis, and changed the time of screening from 6 h to 1 h before taking the decision to administer antibiotics for the first time. Patients with MDROs whose screening was completed within 6 h from July 1, 2013 to June 30, 2016 and those whose screening was completed within 1 h from July 1, 2016 to June 30, 2019, were enrolled. The “within 6 h” group was treated according to the 2005 US pneumonia treatment, 2007 Taiwan, and 2012 Surviving Sepsis Campaign (SSC) guidelines (Table 2) [27,32,33] In contrast, the “within 1 h group” was treated based on the 2016 US, 2018 Taiwan, and 2016 SSC guidelines [3,4,21,34]
The study population was divided to four subgroups according to the MDROs risk phenotype, and the screening completion time within 1 or 6 h. The clinical outcomes were compared between these four groups (low-risk within 6 h, low-risk within 1 h, high-risk within 6 h, and high-risk within 1 h groups).
The inclusion criteria were symptoms of pneumonia and early sepsis encountered at the ED. Pneumonia was defined as early onset of fever, cough, or phlegm, and infiltrations or patches in chest X ray. Patient had suspected septic shock and an ED physician decided whether fluid resuscitation or inotropes should be applied to keep the mean arterial pressure ≥ 65mmHg .
Patients who had infection (tuberculosis, viral, or fungal infections only), transferred from the RCW with nosocomial infections, could not admit to the ICU within 24 h, or those for whom hospice palliative care was decided were excluded from the study.
Disease severity indexes of sepsis and pneumonia were collected from the ED or MICU including the pneumonia severity (CURB-65), acute physiology and chronic health evaluation II (APACHE II) , quick sequential organ failure assessment (qSOFA), and SOFA scores, and the mechanical ventilation and lactate levels. CURB-65 and qSOFA scores were measured as previously described . Additionally, underlying comorbidities in patients were recorded and used to derive the Charlson comorbidity index (CCI) . Previous antibiotic exposure was defined as at least one course of antibiotic therapy within 90 days before the current ED admission. Previous cultures were collected within 90 days of ED admission, where available.
CURB-65 was measured based on confusion (newly disoriented in person, place, or time), blood urea nitrogen (BUN) >20 mg/dl, respiratory rate ≥30 breaths/min, blood pressure (systolic [SBP]) <90 mmHg, or diastolic blood pressure ≤60 mmHg), and age ≥65 years  Additionally, underlying comorbidities (disabling neurologic conditions, chronic obstructive or restrictive lung disease, coronary atherosclerotic disease, congestive heart failure, liver cirrhosis, end-stage renal disease, diabetes, metastatic cancer, and acquired immunodeficiency syndrome) in patients were recorded. The comorbidities were used to derive the CCI . The qSOFA score was calculated by adding one point for any of the following criteria: altered mental status (Glasgow coma scale <15), high respiratory rate (≥22/min), or low blood pressure (SBP ≤100 mmHg).
Sputum cultures were collected within 3 days of admission to the ED by expectoration, secretion suction, or endotracheal suction. The sputum quality was determined by the presence of squamous epithelial cells on Gram stain (≤10/low power field), and the first collection specimen with positive result and good quality was chosen.
Antimicrobial susceptibility test was performed according to the Clinical and Laboratory Standards Institute guidelines . α- and β-streptococcus, and Haemophilus influenzae were subjected to the BBL™ Sensi-Disc™ susceptibility tests (Becton Dickinson and Company, Franklin Lakes, NJ, USA), and other bacteria were subjected to the minimum inhibitory concentration tests by using the VITEK®2 automated system (bioMérieux, Lyon, France). Antibiotic resistance was manually examined by the BBL Sensi-Disc test or automatically determined by the VITEK®2 automated system. Infection with polymicrobial pathogens in the sputum and more than two infection sources other than those developed in the lung, were recorded. MDROs were defined as vancomycin-resistant Enterococci, oxacillin-resistant Staphylococcus aureus, or multidrug-resistant Gram-negative bacilli (GNB), with acquired non-susceptibility to at least one agent in three or more antimicrobial categories 
The appropriate antibiotic treatment was defined as the final cultured microorganisms that were susceptible to the antibiotics used. Multidrug therapy was defined as a combined therapy of more than one antibiotic to broaden the antimicrobial coverage or accelerate pathogen clearance; this included combination therapies .
The clinical primary outcomes examined were the ICU and in-hospital mortality rates, length-of-stay in the ICU, and ventilator-use days. The secondary outcomes were the rates of the appropriate antibiotic treatment within 1 h and the rates of multidrug therapy in the ED and the ICU.
The initial admission time since the ED admission was measured. The ICU mortality was determined as the number of patients who died during the ICU admission divided by the total number of patients hospitalized. In-hospital mortality was determined as the number of patients who died during this admission divided by the total number of patients hospitalized. The length of the ICU stay was measured as the time from the ED admission to the end of the ICU care. Ventilator-use days were measured as the total number of days from the ED admission until the patient was weaned off the ventilators (including the non-invasive positive airway pressure and invasive mechanical ventilators).
Continuous data are expressed as the mean ± standard deviation. Categorical data are expressed as frequencies and percentages. The demographic and clinical characteristics were compared using the Student’s t-test between the “within 1 h” and “within 6 h” groups. The analysis of variance (ANOVA) was used for comparisons among the four subgroups. The chi-square test was used for comparing categorical variables between the subgroups. Additionally, we performed post-hoc analysis for continuous variables that showed significant differences. The Kaplan-Meier survival curve was used to examine mortality among different levels of risk for the MDROs. Multivariate analysis using the Cox regression analysis was performed to determine the factors associated with ICU mortality after adjusting for other confounding factors and those associated with multidrug therapy in the ICU. The data were analyzed using SPSS version 24.0 (IBM Corp, Armonk, NY, US) and a p-value of <0.05 was considered statistically significant.