A Strategy to Improve Implementation of LTVV for Patients w/ ARDS

Overview

The primary objective of this study is to evaluate whether a multi-component implementation strategy/quality improvement intervention comprised of 1) clinical decision support that couples a natural language processing (NLP) acute respiratory distress syndrome (ARDS) recognition tool with a clinician alert system, and 2) audit and feedback improves the implementation of low tidal volume ventilation (LTVV) for patients with the acute respiratory distress syndrome (ARDS). This will be accomplished with a cluster randomized controlled trial comparing the implementation strategy to usual care

Setting: This is a multi-ICU, single-center cluster randomized trial that will be conducted among the patients and their clinicians on four adult ICU services at Northwestern Memorial Hospital. Each ICU service will act as a cluster. Each ICU service cares for patients in one or two physical ICU locations: * MICU service (Medical ICU location) * NCC service (Neuroscience/spine ICU location) * ACC service (Cardiothoracic/transplant ICU and neuroscience/spine ICU locations) * SCC service (Cardiothoracic/transplant ICU and surgical ICU locations) Overall Study Design: The study will consist of a pre-trial period of baseline data collection followed by a cluster randomized controlled trial. During the pre-trial period, the investigators will retrospectively determine the proportion of patients with ARDS on each ICU service who received at least one ventilator setting adherent to LTVV in the previous one year (baseline LTVV rate). At the end of the pre-trial period, the investigators will randomize the four ICUs services into an intervention group (ICUs A and B) and usual care group (ICUs C and D). Based on pre-trial data, the investigators will calculate the difference in baseline LTVV rates between ICUs A+B and C+D, which will be used to calculate the sample size necessary for the trial period. Following the pre-trial period, ICUs A and B will begin to receive the implementation strategy while ICUs C and D will continue with usual care as a comparator group. All mechanically ventilated, non-ARDS patients on all four ICU services during the pre-trial and trial periods will also be included as a separate concurrent comparator cohort. This will allow us to assess for temporal changes in LTVV use for mechanically ventilated patients in general, and potential spill-over of LTVV use for non-ARDS patients in ICUs A and B. Phase I: Pre-trial cohort study The study will begin with a pre-trial cohort study. The investigators will conduct a retrospective chart review of all mechanically ventilated patients admitted to one of the four participating ICUs within the previous one year. All patients who meet the inclusion criteria as either ARDS or non-ARDS patients will be included. The main purpose of the pre-trial study is to establish a baseline rate of LTVV use in each study ICU, which will be used as part of the primary endpoint for the clinical trial described below. All secondary endpoints listed in the Appendix will be collected on all patients in the pre-trial study. Phase II: Randomization At the end of the pre-trial study, the investigators will randomize the four ICU services to the intervention and usual care groups. The study statistician will conduct a simple randomization approach using computer-generated random numbers to allocate each of the four ICUs to the intervention (2 ICUs) and usual care groups (2 ICUs), ensuring that there are two intervention and two usual care ICUs. The intervention ICU services will be labeled ICU A and ICU B, and the usual care ICU services will be labeled ICU C and ICU D. The allocation of each ICU will not be revealed to other study personnel until after the calculation of the baseline LTVV rates for each ICU service. After randomization, the investigators will calculate the baseline LTVV rate in each ICU (and therefore for the intervention and usual care groups). Based on these baseline LTVV rates, the investigators will finalize the sample size calculation for the clinical trial. Phase III: Cluster randomized controlled trial After the randomization process is completed, the person conducting ICU randomization will reveal the allocation of ICUs to other study co-investigators. Immediately following this unblinding, the cluster randomized controlled trial will begin. The two ICUs randomized to receive usual care (ICUs C and D) will not be exposed to an intervention during the trial period. The two ICUs randomized to the intervention group (ICUs A and B) will receive the intervention for the duration of the trial period. Implementation strategy/Intervention (ICUs A and B) The multi-component implementation strategy/quality improvement intervention that ICUs A and B will receive during the clinical trial will consist of three parts: orientation, audit and feedback, and clinical decision support that combines an NLP-driven ARDS diagnostic screening tool, EHR-based tasks for clinicians to complete, and a clinician reminder system. Implementation strategy component A: Orientation The attending and fellow physicians and ICU nurse managers of ICUs A and B will be sent an email with information orienting them to the study. Respiratory therapists will not be sent the orientation email since they rotate between the different ICUs. The text of the email is shown in the Appendix (final wording may be modified). Implementation strategy component B: Audit and feedback Audit and feedback will be directed to ICU attendings and fellows. At the end of every ICU shift/rotation of at least five days, each physician who has cared for at least one ARDS patient in ICU A or B during the previous shift/rotation will receive an anonymous email from a study co-investigator providing their specific rate of LTVV (number of patients to whom they gave LTVV / number of ARDS patients managed). Implementation strategy component C: Clinical decision support The clinical decision support (CDS) intervention combines 1) an ARDS screening tool using a natural language processing (NLP) algorithm, 2) EHR-based tasks for clinicians to complete, and 3) a clinician reminder system. 1. Based on our prior research, the investigators developed a computer application that automatically screens the EHR in real-time for initial diagnostic criteria for ARDS. If a patient is flagged as meeting these initial criteria, the application will download all data necessary to make a diagnosis of ARDS according to the Berlin Definition. Part of the CDS intervention includes a natural language processing (NLP) program. The NLP program will analyze downloaded data and determine whether the patient has ARDS according to the Berlin Definition. 2. If a patient in ICU A or B is identified as having ARDS, the NLP program will trigger the EHR-based tasks for clinicians to complete and the clinician reminder system. The NLP will send an alert to at least one member of the study team that a patient has been identified as having ARDS. 1. The study team member will create an order set for low tidal volume ventilation for that patient. This order set will not be signed, but will be directed to the attending and/or fellow physician to sign. The order set will include information on the patient's current tidal volume, target tidal volume for low tidal volume ventilation, and an order to initiate low tidal volume ventilation. In addition, the co-investigator will also enter a task for the patient's nurse and RT. The task will alert the nurse and RT that the patient has ARDS, and that they should discuss initiating low tidal volume ventilation with the patient's physicians. 2. The study team member who receives the NLP alert will also send a text page to the pagers of the attending, fellow, and resident physicians currently taking care of a newly identified ARDS patient. The page will contain text suggesting that the physician consider a diagnosis of ARDS and initiate LTVV. For all patients in ICUs A and B identified as having ARDS and reported to the co-investigator, who are subsequently not started on LTVV within 24 hours of the clinical decision support intervention described above, both the EHR-based tasks and the web pages will be repeated every 24 hours.

Conditions

Interventions

Eligibility

Locations (1)

Outcomes