Implementation Trial of Predictive Modeling to Enhance Diagnosis and Improve Treatment in Pediatric Septic Shock

Implementation Trial of Predictive Modeling to Enhance Diagnosis and Improve Treatment in Pediatric Septic Shock

Implement and PREDICT Shock: An Implementation Trial of Predictive Modeling to Enhance Diagnosis and Improve Critical Treatment in Pediatric Septic Shock

This study is a prospective, stepped-wedge implementation trial to test the effects of implementing a Clinical Decision Support (CDS) tool for prediction of septic shock in four Emergency Departments within a pediatric healthcare network. The primary outcome will be the proportion of sepsis patients who receive guideline-concordant septic shock care after implementation of the CDS, and the secondary outcome will be time-to-antibiotic after sepsis recognition.

Septic shock is a leading global cause of pediatric death. In the US, the in-hospital mortality rate for children with sepsis is 5-20%. Septic shock is a state of critical infection that requires advanced and resource-intensive resuscitation, and morbidity-free survival depends on timely diagnosis. Critical care delivered in a delayed fashion, after a child is in hypotensive shock, is less effective; for each hour of unrecognized shock the odds of death more than double. Advances have been made in timely sepsis treatment, but improving diagnosis of septic shock in children remains elusive. Improved early diagnosis would accelerate treatment and improve outcomes. Tools that have been deployed to improve diagnosis in pediatric sepsis either diagnose it after organ dysfunction criteria have been met, or depend heavily on subspecialty physician judgment and have not been tested outside of tertiary pediatric hospitals. Thus, the evidence-based 2020 Surviving Sepsis Children's Guidelines for pediatric sepsis stated that "high-quality trials on pediatric sepsis recognition are lacking, and data are not sufficient to suggest any particular screening tool," and identified pediatric sepsis recognition trials as an important research need. Despite this, the guidelines weakly recommended screening patients "who present as acutely unwell" for septic shock, citing very low quality evidence. The guidelines also stated that there is no evidence for the effectiveness of any existing pediatric sepsis screening tools. This study addresses the gap in knowledge about the effectiveness of pediatric sepsis prediction tools. The study team has developed and retrospectively validated early diagnostic models that leverage clinical data in the Electronic Health Record (EHR) to predict septic shock in children the emergency setting [1, 2]. In order to address concerns about alert fatigue and antibiotic overuse, these predictive models were designed to identify patients at high risk for shock among patients in whom clinicians initially had some suspicion for sepsis. In this pilot implementation trial, Clinical Decision Support based on septic shock prediction models will be introduced to 4 pediatric Emergency Departments (EDs) within the Children's Hospital Colorado care system in a stepped wedge design, in addition to routine clinical care. Routine clinical care at all study sites includes existing sepsis pathways, order sets, and quality metrics that are aligned with the national Improving Pediatric Sepsis Outcomes Collaborative and the Pediatric Surviving Sepsis guidelines. The primary outcome will be the proportion of sepsis patients who receive guideline-concordant septic shock care after implementation of the CDS, and the secondary outcome will be time-to-antibiotic after sepsis recognition.

This will be a stepped wedge cluster randomized trial, with the site as the unit of randomization. There will be five study periods and four sites. In the first period no sites will receive the intervention. During each subsequent period, one site will crossover to receive the intervention.

Emergency Department sites (unit of randomization) cannot be blinded; however patients who are subjects of the intervention are not aware of the arm of the site in the clinical trial.

Emergency Department sites in this arm will have Clinical Decision Support (CDS) alerts active in the Electronic Health Record during clinical ED care of patients with suspected sepsis, in addition to following usual institutional standard of care for sepsis. The CDS will alert providers to patients at high risk for developing septic shock.

Emergency Department sites in this arm will follow the institutional standard for sepsis care without Clinical Decision Support. Standard care includes clinical diagnosis of sepsis supported by institutional sepsis education, a sepsis pathway and orderset.

The intervention will be activating a septic shock clinical decision support tool (CDS) in the Electronic Health Record at the site, making it available to trigger and alert Emergency Department providers during clinical care. Patients will be identified when providers suspect sepsis and initiate a sepsis evaluation, using the institutional, clinically-standard, sepsis pathway/orderset. After the CDS is triggered, it will use available Electronic Health Record data to calculate the risk of septic shock, using previously-published predictive models [Ref 1, 2]. The CDS will notify providers if the patient is at elevated risk of septic shock and prompt them to follow institutional standard care for septic shock (including close monitoring, complete laboratory evaluation for organ dysfunction, and immediate clinical involvement of an attending Pediatric Emergency Physician). All clinical care decisions will be determined by the providers.

Emergency Department sites in this arm will not have Clinical Decision Support for septic shock visible in the Electronic Health Record for providers at the site. Providers will follow usual institutional standards of sepsis care.

Treatment will be defined as concordant with Surviving Sepsis Campaign guidelines for shock if intravenous antibiotics are initiated within 60 minutes of sepsis recognition and an intravenous fluid bolus is initiated within 60 minutes of sepsis recognition. This will be a binary outcome. Sepsis recognition is defined as the earlier of: sepsis page sent, sepsis orderset use, or intravenous antibiotic order.

Time to antibiotics will be measured in minutes from the time of sepsis recognition to the start of intravenous antibiotic treatment. This will be a time-to-event outcome. Sepsis recognition is defined as the earlier of: sepsis page sent, sepsis orderset use, or intravenous antibiotic order.

The number of patients who experience an in-hospital death up to 30 days after Emergency Department arrival.

Septic shock will be defined as suspected infection and systolic hypotension and either vasoactive use or ≥30 ml/kg intravenous bolus fluid administration

Emergency Department Inclusion Criteria: Children's Hospital Colorado Emergency Care sites All providers (physicians, nurse practitioners, physician assistants) at Children's Hospital Colorado Emergency Care sites will be included Exclusion Criteria: - None Patients whose secondary data will be used to assess the outcomes of the intervention will be 60 days through 18 years old.

The study team guarantees that any and all data collected as part of this project will be released in accordance with standard data sharing policies and procedures to validate research findings if requested. Data will be made available in a timely manner to the broader scientific community, and will be complete, and as accurate as possible. All data released will be de-identified, with no information that could be linked to any study subjects, or participating study practices in order to ensure the confidentiality of all subjects and practices. If necessary, a data use agreement will be established.

The study team intends to share study data that may be requested from other research investigators in a data-sharing agreement provided at the individual study's end. The data-sharing agreement will include requirements to protect participants' privacy and data confidentiality. It will prohibit the recipient from transferring the data to other users and will require that the data's security be protected by standard means and be used for research purposes only. The method of distribution will be by request to Dr. Scott, the study PI. After a requestor completes the data-sharing agreement, we will upload data to a secure File Transfer Protocol (FTP) site with the limited dataset to the requestor, or email the data through our University secured email systems that require users to create an account and sign-in with a username and password in order to receive and download any type of sensitive data.

Scott HF, Colborn KL, Sevick CJ, Bajaj L, Kissoon N, Deakyne Davies SJ, Kempe A. Development and Validation of a Predictive Model of the Risk of Pediatric Septic Shock Using Data Known at the Time of Hospital Arrival. J Pediatr. 2020 Feb;217:145-151.e6. doi: 10.1016/j.jpeds.2019.09.079. Epub 2019 Nov 13.

Scott HF, Colborn KL, Sevick CJ, Bajaj L, Deakyne Davies SJ, Fairclough D, Kissoon N, Kempe A. Development and Validation of a Model to Predict Pediatric Septic Shock Using Data Known 2 Hours After Hospital Arrival. Pediatr Crit Care Med. 2021 Jan 1;22(1):16-26. doi: 10.1097/PCC.0000000000002589.