Background: In-hospital patient deterioration, often unpredictable and multifaceted, presents a significant challenge in hospital medicine. Despite existing measures like illness severity scoring systems and rapid response teams (RRT), patient outcomes remain suboptimal. Delays in recognizing and treating worsening conditions lead to adverse effects and increased healthcare costs.
Purpose: In our large healthcare system, covering two academic and six community hospitals, we observed limitations in existing systems, particularly in identifying non-sepsis deterioration. To address this, we implemented the Epic Deterioration Index (DTI), an early warning score that uses patient data to predict real-time deterioration risk within 12-38 hours.
Description: Our approach utilizes the FOCUS-PDCA (Find, Organize, Clarify, Understand, Select, Plan, Do, Check, Act) methodology from Six Sigma for clinical intervention design. We aim to assess the impact of DTI-based interventions on patient outcomes through a before-and-after interrupted time series analysis. The pilot study, which went live on a 30-bed intermediate care unit on 11/25/2023, comprises three core intervention elements:1) Custom system-level patient list. For all patients on the unit, the patient list displays DTI scores, changes in scores, and whether the patient has reached a high-risk threshold. (Figure 1)2) Charge nurse workflow ownership. Based on stakeholder input, the unit charge nurse is responsible for monitoring the custom patient list and identifying patients on their unit who enter elevated risk categories.3) Team-based clinical reasoning: When a patient’s deterioration risk becomes elevated, the charge nurse contacts and performs a bedside assessment with the primary nurse. Based on their combined assessment, nursing decides whether to contact the primary team, activate a Rapid Response, or both. (Figure 2)Our primary outcome, “Provider Action,” is defined as the first appropriate care escalation after a patient reaches an elevated-risk deterioration threshold. Secondary outcomes include in-hospital mortality, length of stay, readmission rate, and ICU transfers per 100,000 patient days. Acceptance and sustainability will be evaluated through clinician surveys and interviews, focusing on the Technology Acceptance Model (TAM) constructs.
Conclusions: This project underscores the potential of integrating advanced data analytics into clinical practice to enhance the early detection and management of in-hospital patient deterioration. The combination of system-level patient list customization, charge nurse workflow ownership, and team-based clinical reasoning represents a novel approach to patient care – one that does not contribute to alert fatigue through low-yield “Best Practice Advisories.” Our model empowers nursing staff with real-time data and fosters collaborative decision-making, ensuring timely interventions. The ongoing analysis of this pilot study will provide valuable insights into the effectiveness of predictive analytics workflow in enhancing patient safety and healthcare efficiency.
