AI Playbook for Sepsis Care Variation by Unit
A hospital's quality data shows significant variation in sepsis outcomes by unit: ICU mortality is 18%, step-down unit mortality is 28%, and floor mortality is 41%. The hospital medical director wants to understand whether the mortality differences reflect patient severity or care process variation.
When to use this playbook
- Use this playbook when the decision looks like the situation above: A hospital's quality data shows significant variation in sepsis outcomes by unit: ICU mortality is 18%, step-down unit mortality is 28%, and floor mortality is 41%.
- It is a fit when you have source files in hand and need a structured, reviewable analysis — not a generic chat answer about "Sepsis Care Variation by Unit".
- Do not use it as a substitute for licensed, legal, clinical, or authorized official judgment in the domain.
What you'll need
- Unit-level sepsis outcome data (mortality, LOS, bundle compliance) for the past 12 months
- Patient severity data by unit (APACHE II, SOFA scores, or equivalent)
- Time from recognition to antibiotic by unit
- Nursing to patient ratios by unit
- Rapid response team activation rates by unit
Attachments: Documents (Documents)
The Prompt
You are a clinical quality specialist analyzing sepsis outcome variation by unit at a hospital. I am attaching: Work only from the attached source files. If a conclusion is not supported, say so. Produce: 1. Risk-adjust the mortality rates: after controlling for patient severity, is the floor mortality rate still significantly higher than ICU, or does severity explain the difference? 2. Calculate the time-to-antibiotic by unit and assess whether floor delays are contributing to the mortality differential. 3. Identify the specific care process differences between the ICU (best outcomes) and the floor (worst outcomes): what the ICU does that the floor doesn't. 4. Assess whether the step-down unit and floor have adequate nursing ratios to execute the 1-hour bundle for all recognized sepsis patients simultaneously. 5. Tell me the quality improvement interventions that have the strongest evidence for improving floor sepsis outcomes and the expected mortality reduction. Call out where independent models are likely to disagree, and list follow-up documents a reviewer should request.
What to expect
- Risk-adjusted mortality comparison by unit
- Time-to-antibiotic analysis by unit
- ICU vs. floor care process difference identification
- Nursing ratio adequacy assessment
- Evidence-based floor intervention recommendations with projected mortality reduction
Review before you act
- Validate this output against source files before relying on it: Risk-adjust the mortality rates: after controlling for patient severity, is the floor mortality rate still significantly higher than ICU, or does severity explain the difference?.
- Validate this output against source files before relying on it: Calculate the time-to-antibiotic by unit and assess whether floor delays are contributing to the mortality differential.
- Validate this output against source files before relying on it: Identify the specific care process differences between the ICU (best outcomes) and the floor (worst outcomes): what the ICU does that the floor doesn't.
- Validate this output against source files before relying on it: Assess whether the step-down unit and floor have adequate nursing ratios to execute the 1-hour bundle for all recognized sepsis patients simultaneously.
- Confirm every cited figure, date, counterparty, or requirement against the attached originals — models compress and can drop a qualifier.
- Treat disagreement between models as a review item, especially on classification, materiality, and recommended next action.
- Do not authorize an operational, clinical, legal, credit, or enforcement action solely because the models agree.
Why compare models on this
For Sepsis Care Variation by Unit, running the same attachments across independent models is useful because the hard part is classification and completeness, not fluency. The workflow is already designed to surface risk-adjusted mortality comparison by unit; time-to-antibiotic analysis by unit; icu vs. floor care process difference identification; nursing ratio adequacy assessment. Those are comparison artifacts — they only exist if more than one model runs. Models disagree on whether an alert is noise, whether a death was sepsis-attributable, and whether a risk model is calibrated. Those disagreements belong in a morbidity-and-mortality style review, not an auto-implemented rule.
See governed multi-model AI on your own prompt
Compare GPT-5, Claude, and Gemini side by side, with human review and a decision record built in.

