Pulse leverages the power of Anthuria and its AI agents to proactively identify potential clinical triggers requiring review. Currently, the co-pilot reads and reviews all clinical notes. The ability to read all PDF and documents is coming soon.
Understanding the AI Agents:
Unlike traditional word-matching tools, Pulse utilizes a sophisticated Large Language Model (LLM) specifically trained on SNF-specific semantics and context. This advanced AI understands the nuances of clinical documentation, going beyond simple keyword recognition. For example, while a word-matching tool might misinterpret "didn't fall" as a "fall," the LLM understands the contextual meaning and avoids false positives. Furthermore, the LLM can identify implicit information, such as recognizing "patient was found on the floor" as a fall even without the explicit word "fall."
How Pulse Works:
The AI agent continuously analyzes every note within your EHR, proactively identifying potential triggers that Directors of Nursing typically look for during census reviews. For each potential trigger, the AI agent:
Tags the note: The note is labeled with the specific trigger.
Provides an AI explanation: The AI explains why the note was tagged with the trigger, providing context and supporting evidence.
Displays the original note: You can review the complete note within Pulse.
Highlights relevant text: Text evidence highlighted within notes to allow you to quickly audit and take action.
Offers direct EHR access: A hyperlink to the original progress note in your EHR is provided for seamless access to the full record.
Navigating the Pulse Dashboard:
Trigger Summary: At the top of the Pulse page, pill buttons display the number of patients requiring review for each identified trigger.
Overall Trigger Percentage: The upper right corner shows the percentage of notes containing a potential trigger.
Trigger Information: An info icon on each trigger provides details about the specific contextual cues that prompt the AI agent to tag a note with that trigger.
Filtering by Trigger: Clicking a trigger button filters the patient list to show only those with that specific trigger.
Review Status: A blue dot next to a patient's name indicates that their notes have not yet been reviewed. Clicking into the patient and reviewing the note and AI explanation removes the blue dot.
Key Features and Functionality:
Progress Note Review: When selecting a patient, you'll see the full progress note, its ID (hyperlinked to your EHR), the AI explanation for the tag, and AI-generated categories.
AI Feedback: The thumbs up/down feedback option helps train the AI agent, continuously improving its trigger recognition accuracy.
Date Range Filtering: The revision date filter allows you to search across any desired date range.
Customizable Column View: The gear icon lets you customize your column view, adding relevant information like "created by" or "Progress Note ID."
Free Text Search: Use the search bar to find specific keywords (e.g., "antibiotics," "IV fluids") within the notes. The agent will highlight the specific text that matches your search.
Custom Triggers: Create your own triggers by creating a category name, the facilities the trigger should be applied to and a trigger word and/or specific keywords. It takes 20 minutes for keywords to start working and 24 hours for trigger words. Once the AI has had the necessary time to train and start implementing, a preview feature shows real notes where the AI detects the trigger. You can refine your settings at any time by selecting the info icon on the lower right of the trigger button. The AI agent will tag all clinical notes with the trigger category name.
Downloading Pulse Lists: You can download the data shown in your current Pulse list by clicking the download icon (shown below) located to the right of the search bar. The downloaded file will include all of your current settings and filters. This means:
If you've customized your columns, the download will reflect your customized view.
If you've used filters (e.g., by trigger or search phrase), the download will only contain the filtered results.
Creating Triggers
Definitions
Category Name: This is the name displayed on the button used to access the list of patients matching this trigger.
Trigger Word: Input the word that will signal the AI to find related patient data. The AI uses the meaning of the word to identify matching instances.
Facility: Select the facilities where the trigger will be active
Keywords: The AI will identify all variations of the keywords you enter. For example, entering 'flu' will also find 'influenza'.
Trigger Words vs. Keywords: When to Use Each
Trigger Words: Focusing on Meaning and AI Reasoning
Use trigger words when:
You want the AI to find information based on the meaning of a word, not just exact word matches.
You want to quickly identify broad categories or themes (e.g., "respiratory," "cardiac," "infection").
You want to leverage the AI's ability to logically connect related concepts.
Example:
Instead of using separate keywords for "pneumonia," "bronchitis," and "asthma," use the trigger word "respiratory." The AI will understand that these conditions are related to respiratory issues and identify them.
Key points:
Only one trigger word is allowed per custom trigger.
The AI uses the trigger word to understand the meaning of what you are looking for.
The AI will take about 24 hours to start implementing trigger words.
Keywords: Focusing on Specific Word Variations and Nuances
Use keywords when:
You need to find specific word variations or related terms (e.g., "gait," "ambulation," "instability").
You are looking for nuanced categories where the meaning is complex and hard to define with a single word (e.g., "mobility changes").
You want to capture specific details that might be missed by a broad trigger word.
Note, keywords match on word variations, and not on meaning. Specifically, words are matched by snowball stemming which maps different forms of the same word to a common "stem."
Caution:
Keywords can produce false positives when used with negative statements (e.g., "no COVID" will trigger on the keyword "COVID").
Example:
To track mobility changes, use keywords like "balance," "transfer," and "weakness." This will capture a wider range of observations than a single trigger word.
Key points:
You can use multiple keywords per custom trigger.
Keywords match word variations.
The AI will take about 20 minutes to start implementing keywords.
Combining Trigger Words and Keywords:
You can use both a trigger word and keywords in the same trigger.
The AI will identify records that match either the trigger word's meaning or any of the keywords. This provides a broader search.
Not sure what configuration to use for your custom trigger? Create separate triggers to test results for 3 to 4 days and then iterate based on which performs best for your use case.
How to edit or Delete: Custom triggers can be edited and deleted should you want to remove a trigger or refine your keywords or category name.
Click on the info icon in the lower right corner of the trigger pill
Update Category Name, Facilities and Keywords as desired
The preview will update accordingly. Review the preview results to make sure the trigger will tag as desired
Submit to save changes
Delete to remove the custom trigger completely
Benefits of Pulse:
Proactive Risk Identification: Identify potential clinical issues early, allowing for timely intervention.
Reduced Review Time: The AI agent automates the initial review process, saving clinicians valuable time.
Improved Accuracy: The LLM's contextual understanding minimizes false positives and ensures more accurate trigger identification.
Enhanced Clinical Decision-Making: Provides clinicians with the information they need to make informed decisions. Quick identification of all triggers for each patient provides clinical staff with the full story so that they can take the right actions.