Materials and Methods
Definitions and Terms
Sepsis. The co-occurrence of suspected infection and two or more of the systemic inflammatory response syndrome (SIRS) criteria: 1) temperature more than 38°C or less than 36°C, 2) heart rate more than 90 beats/min, 3) respiratory rate more than 20 breaths/min or PaCO2 less than 32 mm Hg, and 4) WBC count more than 12,000 cells/mm or less than 4,000 cells/mm or more than 10% immature (band) forms.
Modified SIRS Criteria. Two or more SIRS criteria met within a rolling 24-hour window, with at least one being abnormal temperature or WBC count.
Listening Application. An electronic tool that monitors patient data in real time, evaluates data against diagnostic and alerting rules to identify patients who newly meet modified SIRS criteria, and communicates with the alerting system to notify providers.
Alerting System. An electronic tool that receives information from the listening application on patients who have met modified SIRS criteria, notifies providers of the finding, and solicits an assessment to determine if the patient clinically meets criteria for sepsis.
Integrated Sepsis Assessment and Management Tool. A software program within the electronic medical record (EMR) designed to import, synthesize, and display sepsis-related data from different portions of the record, use logic rules to offer an up-to-date, individualized evaluation of sepsis severity and response to therapy, inform users about evidence-based guidelines, and facilitate rapid order entry.
Clinical Provider. ICU resident physician or nurse practitioner primarily responsible for patient management and order entry.
Screening and Enrollment
From April 1 to July 31, 2012, we conducted a pragmatic, open-label, parallel-group randomized trial in the medical ICU (MICU) and surgical ICU (SICU) of Vanderbilt University Hospital. This study was approved by the institutional review board with waiver of informed consent.
All patients admitted to a study ICU and assessed by their clinical provider as having sepsis were enrolled (Fig. 1). Patients assessed as septic via an automatic prompt offered at ICU admission were enrolled immediately. For all patients without sepsis at ICU admission, electronic monitoring was employed to facilitate early detection of sepsis development (Fig. S1, Supplemental Digital Content 1, http://links.lww.com/CCM/B296). A previously established listening application screened real-time data from the nursing documentation, electronic health record, and laboratory system for modified SIRS criteria. If modified SIRS criteria developed in an ICU patient who had not previously been assessed by a clinical provider as septic, an alerting system notified the clinical provider via text page and a flag appeared next to the patient's name on the clinical provider's electronic patient list. Using this flag, clinical providers recorded a revised sepsis assessment. If an assessment was not recorded within 1 hour, a reminder was sent. No management recommendations were given by the listening application or alerting system. If patients were assessed not to be septic, further alerts were suppressed for 48 hours, and then electronic monitoring resumed. If patients were assessed as septic, they were enrolled in the study and no further alerts were delivered for that hospitalization. Study enrollment occurred at the time the first prompt to which the clinical provider confirmed the presence of sepsis was offered. Patients were excluded if they were never assessed as septic by their clinical provider or had been previously enrolled during the same hospitalization. Patients with orders limiting resuscitation were not excluded.
(Enlarge Image)
Figure 1.
Enrollment and randomization. Of 475 patients assessed as septic, 68 had been enrolled in the study previously in the hospitalization and were excluded. The remaining 407 were randomized, followed, and included in the analysis.
Randomization and Masking
All ICU patients assessed by a provider as septic were randomized by a computerized algorithm without use of permuted blocks or stratification to usual care (control) or the availability of an integrated electronic sepsis assessment and management tool (intervention). Clinical providers were aware of group assignment. Separate study personnel remained blinded until the completion of data collection.
Intervention
In the intervention group, when a provider's assessment confirmed the presence of sepsis, the integrated sepsis assessment and management tool (Video 1, Supplemental Digital Content 2, http://links.lww.com/CCM/B297) opened automatically and remained available to all providers throughout the ICU admission. The tool resided directly within the EMR and contained a divided display. Half presented graphs of current value, trend, and goal range for temperature, heart rate, mean arterial pressure, central venous pressure (CVP), WBC count, hemoglobin, platelet count, prothrombin time, lactate, and Richmond Agitation and Sedation Score. The other half contained a set of tabs dedicated to evaluation of the patient condition and management. The assessment tab offered providers an evaluation of sepsis severity using an algorithm in which sepsis was defined as previous confirmation of sepsis by a clinical provider in response to a prompt, severe sepsis was sepsis plus mean arterial pressure less than or equal to 60 mm Hg or a lactate above the laboratory upper limit of normal, and septic shock was sepsis plus vasopressor use or either mean arterial pressure less than or equal to 60 mm Hg or a lactate greater than or equal to 4 mmol/L despite an order for greater than or equal to 1,000 mL of IV crystalloid. The provider could agree with or modify the severity assessment, which would then propagate into the management tabs allowing the tool to highlight interventions recommended for that specific patient. The management tabs were displayed in a workflow modeled on the SSC 6-hour Resuscitation Bundle beginning with a diagnostics panel addressing basic labs, cultures, and imaging for source control, followed by a therapeutics panel addressing antibiotics and goal-directed resuscitation including fluid boluses, CVP, vasopressors, serial lactate measurements, inotrope use, and blood transfusion, and concluded by a supportive care panel addressing lung-protective ventilation, adrenal insufficiency, glucose control, and venous thromboembolic and stress ulcer prophylaxis. Each management tab contained evidence-based information and "single-click" order entry enabling providers to place related orders while reviewing the guidelines. The tool was integrated with electronic order entry from all parts of the hospital so that providers could see via a color-coded system not only which orders were recommended but which orders had already been completed in the ICU or prior to ICU admission. The tool could be minimized at any time to review the medical record and could be closed and reopened any time throughout the admission. Technical aspects of the sepsis tool's development, architecture, and user interface have been published previously and are summarized in the online supplement (Supplemental Digital Content 1, http://links.lww.com/CCM/B296).
Data Collection
Patients were followed for 28 days or until hospital discharge, whichever occurred first. Demographic and baseline characteristics, vital signs and laboratory results, and clinical outcomes were collected by study personnel blinded to group assignment. The date and time at which each of the SSC 6-hour Resuscitation Bundle elements was completed underwent automated, electronic documentation in the EMR as a part of routine clinical care. These data were prospectively abstracted into a database by blinded study personnel. Listening application, alerting system, and tool usage data were collected automatically in a separate database. Retrospective review of the medical record by study physicians used all available information to adjudicate whether each patient had been septic at the time of enrollment.
The primary endpoint was time from enrollment until completion of all indicated elements of the SSC 6-hour Resuscitation Bundle—blood cultures drawn, broad-spectrum antibiotics ordered, CVP measured, IV fluid bolus administered if indicated, vasopressors administered if indicated, and lactate measured. Lactate clearance is used in place of venous oxygen saturation in our ICUs. Time to completion for each element was calculated as time from enrollment until the time that each bundle element was electronically documented as completed in the EMR. Secondary endpoints included time to completion of each individual SSC 6-hour Resuscitation Bundle element, ICU mortality, days alive and free from mechanical ventilation (ventilator-free days [VFDs]), days alive and out of the ICU (ICU-free days), and days alive and free from vasopressor administration (vasopressor-free days), all to study day 28.
Statistical Analysis
Based on prior data from the same setting, planned enrollment of 400 patients provided 80% statistical power to detect a 1-hour decrease in the primary endpoint of time to completion of all SCC 6-hour Resuscitation Bundle elements with a type I error rate of 0.05. Demographics and baseline characteristics were summarized by median and interquartile range (IQR) or mean and SD for continuous variables and as numbers and percentages for categorical variables across intervention and control groups. To analyze the time to completion of all SSC 6-hour Resuscitation Bundle elements and each element individually, the cumulative event probabilities were estimated and compared using Kaplan-Meier method with log rank testing and Cox proportional-hazards regression. A logistic regression model with prespecified covariates was fit to assess what factors impacted the use of the tool to enter orders. All analyses were performed using the statistics software R version 3.0.1 (http://www.r-project.org).