Study protocol Effects of Philips Visual Patient Avatar on vital sign deviations and audible alarm burden in perioperative care: a dual-centre, quasi-experimental pre-post big-data study protocol (NewYork-Presbyterian/Weill Cornell and University Hospital Zurich)

This study protocol outlines a dual-centre, quasi-experimental pre-post big-data investigation designed to systematically quantify the impact of the Philips Visual Patient Avatar on vital sign deviations and audible alarm burden in perioperative care, aiming to determine if this avatar-based visualization improves situation awareness and reduces cognitive load compared to conventional monitoring.

The Gist

Imagine you are the captain of a ship navigating through a storm. Your dashboard is covered in dozens of separate gauges: one for speed, one for fuel, one for engine temperature, and another for wind direction. To understand if your ship is safe, you have to constantly look at each number, remember what it means, and mentally combine them all to get a full picture. If one gauge starts beeping, you have to stop what you're doing to figure out which one it is and why. This is exactly what anesthesiologists face in an operating room, but with a patient's heart rate, blood pressure, and oxygen levels instead of ship gauges.

This research paper is a study plan (a protocol) for a project designed to see if a new kind of "dashboard" can make this job easier and safer.

The Problem: The "Beeping" and the "Brain Drain"

Currently, patient monitors show numbers and squiggly lines. When something goes wrong, the machine makes a loud noise (an alarm).

• The Mental Load: Doctors have to read many separate numbers and put them together in their heads to understand the patient's condition. This is like trying to solve a math problem while someone is shouting at you.
• The Alarm Fatigue: There are so many beeps that doctors can get tired of them, much like a driver who gets used to a car's "check engine" light and starts ignoring it. This is dangerous because they might miss a real emergency.

The Solution: The "Living Avatar"

The researchers are testing a new technology called the Philips Visual Patient Avatar (VPA).

• The Metaphor: Instead of just showing numbers, imagine a small, animated human figure on the screen.
  • If the patient's oxygen is low, the avatar's skin turns blue.
  • If the heart is beating too fast, the avatar's chest pulses rapidly.
  • If blood pressure is dropping, the avatar might look pale or slump.
• The Goal: This allows the doctor to see the "whole picture" instantly, just by glancing at the avatar's face or body, without having to mentally calculate the numbers. It's the difference between reading a weather report full of data points and simply looking out the window to see if it's raining.

The Experiment: A "Before and After" Test

This study isn't a lab experiment with actors; it's a "big data" look at real hospitals.

• The Locations: They are comparing two huge hospitals: one in New York (USA) and one in Zurich (Switzerland).
• The Method: They will look at thousands of past and future surgery cases.
  1. Phase 1 (Before): They look at data from when the hospitals used the old, number-heavy screens.
  2. Phase 2 (The "Sink-in"): They skip the messy transition period where doctors are just learning how to use the new avatar screens.
  3. Phase 3 (After): They look at data after the new screens have been fully installed and doctors are used to them.

What Are They Measuring?

They aren't measuring if patients lived or died (that's too hard to pin down to just one screen). Instead, they are measuring two specific things:

1. Time in the Danger Zone: How long did the patient's vital signs stay outside the "safe zone"? (e.g., How many minutes was the blood pressure too low?) The hope is that with the avatar, doctors will spot these problems faster and fix them sooner, reducing this time.
2. The Noise Level: How many alarms went off, and how long did they last? The hope is that the avatar helps prevent small issues from becoming big, noisy emergencies, or helps doctors react so quickly that the alarms stop sooner.

The Rules of the Game

• It's a "Quasi-Experiment": They can't randomly assign some patients to the old screen and some to the new one because the whole hospital switches at once. So, they compare the "Before" time to the "After" time.
• The "Sink-in" Phase: Just like when you buy a new car, the first few weeks are awkward while you learn where the buttons are. The researchers are ignoring that awkward period to make sure they only measure the results once everyone is comfortable with the new system.
• Two Different Worlds: Since the two hospitals are in different countries with different rules, they will analyze them separately to see if the avatar works in both environments.

The Bottom Line

This paper is a roadmap for a study that asks: "If we replace a dashboard full of confusing numbers with a friendly, animated character that shows us how a patient is feeling, will doctors catch problems faster and hear fewer annoying beeps?"

The study is currently in the data-gathering phase. They haven't found the answer yet; they are just setting up the rules to find out. If the results are good, it could prove that a simple change in how we look at data can make surgery safer for everyone.

Technical Summary

Technical Summary: Effects of Philips Visual Patient Avatar on Vital Sign Deviations and Audible Alarm Burden

Problem Statement
Perioperative patient monitoring requires clinicians to integrate multiple physiological data streams (e.g., heart rate, blood pressure, oxygen saturation, end-tidal CO2) under high time pressure and frequent interruptions. Conventional monitors present these data as separate numerical values and waveforms, necessitating sequential interpretation and mental integration, which imposes substantial cognitive demands. While audible alarms are designed to enhance safety, they frequently contribute to alarm fatigue, increased workload, and reduced responsiveness to clinically relevant events. Furthermore, time spent outside predefined safe ranges for vital signs is associated with adverse outcomes. Although avatar-based monitoring has shown promise in laboratory and simulation studies for improving situation awareness and reducing perceived workload, its impact on objective monitoring outcomes in routine clinical practice has not been systematically quantified.

Methodology
This study employs a dual-centre, quasi-experimental pre–post design to evaluate the clinical introduction of the Philips Visual Patient Avatar (VPA). The study is conducted at NewYork-Presbyterian Hospital/Weill Cornell Medical Center (New York, USA) and University Hospital Zurich (Zurich, Switzerland).

• Data Source: The study utilizes large volumes of routinely collected perioperative monitoring data (big-data approach).
• Study Phases: Data are collected across three distinct phases at each centre:
  1. Pre-implementation: Baseline data prior to VPA introduction.
  2. Sink-in (Adaptation): A transition period during staged installation and training, excluded from the final analysis to mitigate learning effects.
  3. Post-implementation: Data collected after the completion of installation and training.
• Intervention: The VPA is a software update for Philips IntelliVue patient monitors (MX450, MX550, MX750) that displays an animated virtual patient integrating multiple vital signs and sensor states into a single visual representation (using cues like skin color for saturation and pulsation for pulse rate) alongside conventional numerical displays.
• Unit of Analysis: Individual anaesthesia cases.
• Primary Outcome: Vital sign deviation time, normalized to 60 minutes of effective measurement time. This is defined as the cumulative duration a vital parameter exceeds predefined upper or lower thresholds.
• Secondary Outcomes:
  • VitalSign AUC: An area-under-the-curve metric capturing both the magnitude and duration of deviations from the admissible range.
  • Audible Alarm Burden: Normalized to 60 minutes, including alarm count, cumulative alarm time, and mean alarm reaction time.
• Statistical Analysis: Multivariable regression models will estimate associations between the implementation phase (post vs. pre) and outcomes, adjusting for patient- and procedure-related covariates. Analyses are performed separately for each centre. The study excludes the sink-in phase from the primary comparison.

Key Contributions

• Real-World Quantification: This protocol outlines the first systematic quantification of the VPA's impact on objective monitoring outcomes in routine perioperative care, moving beyond simulation and laboratory settings.
• Dual-Centre Validation: By leveraging data from two independent tertiary academic hospitals on different continents with diverse workflows and alarm configurations, the study aims to enhance the external validity and generalizability of the findings.
• Process-Oriented Metrics: The study focuses on process-level metrics (vital sign stability and alarm burden) rather than hard patient outcomes, providing a clinically meaningful basis for evaluating the intervention's effect on safety indicators.
• Methodological Rigor: The inclusion of a dedicated "sink-in" phase addresses the confounding effects of learning and adaptation during the introduction of new technology, a common challenge in pre-post implementation studies.

Results
Note: As this document is a study protocol submitted prior to the completion of data analysis, no empirical results are reported.
At the time of submission, data collection was complete at University Hospital Zurich and ongoing at NewYork-Presbyterian Hospital. No statistical analyses had been performed. The paper outlines the planned analytical framework but does not present findings regarding the efficacy of the VPA.

Significance and Claims
The paper positions itself as a pragmatic evaluation of whether human-factors benefits demonstrated in controlled studies translate into measurable changes in real-world practice. The authors claim that if favorable associations are observed between VPA implementation and reduced vital sign deviations or alarm burden, these findings will provide the necessary evidence base for future investigations into patient-centred clinical outcomes.

The study acknowledges its limitations, specifically that the quasi-experimental design does not permit causal inference and that residual confounding by time and unmeasured factors cannot be fully excluded. However, the authors argue that the use of large-scale, routinely collected data across diverse clinical environments offers a robust assessment of the intervention's impact under varying real-world conditions. The ultimate goal is to support earlier detection of physiological deviations and improved situation awareness without increasing cognitive load, thereby addressing critical issues in perioperative patient safety.