Bringing Goals-of-Care Conversations Forward at an Academic Medical Center
680-bed academic medical center · Palliative care & care management · hospital-wide census · 7 min read
An academic medical center used Prescient's advanced-illness identification to systematically surface patients who would benefit from a goals-of-care conversation, ranked by risk and paired with the factors driving each score. The result: more of those conversations happened proactively, while patients were stable, rather than during an acute decline.
At a Glance
- Organization
- 680-bed academic medical center
- Scope
- Hospital-wide census
- Teams involved
- Palliative care, care management, hospitalists
- EHR environment
- Single EHR, HL7 FHIR integration
- Pillar deployed
- Advanced Illness Identification
- Time to go-live
- ~9 weeks from discovery
Representative example. This is an illustrative case study built from an anonymized customer archetype to show the shape of a Prescient engagement and its reporting. The scenario and all figures are representative examples, not published results from a named customer.
Results at a Glance
What Changed
Illustrative figures — representative of the reporting structure, not published results.
The Challenge
Where they started
Across a large academic census, the patients most likely to benefit from advanced-illness planning were not consistently identified in time. Palliative referrals tended to be triggered reactively — often during an acute decline — rather than while a patient was stable enough for a considered goals-of-care conversation.
Care management staff spent meaningful time on manual chart review trying to find candidates, and palliative, care management, and hospitalist teams each worked from their own view of the census with no shared, prioritized worklist.
- Palliative referrals were often triggered late — during an acute decline — because candidates weren't systematically identified across a large, busy census.
- Care management teams spent significant time on manual chart review to find patients who might benefit from advanced-illness planning.
- There was no shared, prioritized worklist across palliative, care management, and hospitalist teams.
The Solution
A shared, risk-ranked worklist
Prescient scored the full census for advanced-illness risk using diagnosis and comorbidity burden, functional-decline trajectory, and utilization pattern — with thresholds calibrated against the center's own population and documented goals-of-care history.
The output was a single prioritized worklist that fed into the existing care-management workflow, each patient paired with the specific factors driving their score. Palliative, care management, and hospitalist teams finally worked from one view instead of three.
How data flowed in this deployment
Diagnosis burden, functional status, and utilization history from the EHR.
Trajectory and comorbidity signals combine into a risk band.
Prioritized, risk-ranked list reaches palliative and care teams.
Goals-of-care conversation happens ahead of a crisis.
The Approach
How the Deployment Unfolded
- Weeks 1–4Calibration with clinical leadership
Identification thresholds were calibrated with the palliative and care-management leadership against the center's own patient population and documented goals-of-care history.
- Weeks 5–8Shadow mode & worklist design
The risk-ranked worklist ran in shadow mode while teams validated that the prioritized patients matched clinical judgment.
- Week 9Go-live for care management
The prioritized advanced-illness worklist went live inside the existing care-management workflow, each entry paired with its contributing factors.
- OngoingProactive outreach loop
Teams used the worklist to prioritize outreach, and outcomes fed back into ongoing threshold review with clinical leadership.
The Results
Measured Against Their Own Baseline
Advanced-illness outreach — illustrative before/after
Illustrative example only, not a published result — measured against the center's own pre-deployment baseline.
Goals-of-care conversations held before acute decline
Illustrative adoption curve as the worklist was integrated into care-management routine.
Key Takeaways
What Made It Work
Systematic identification replaced ad-hoc manual chart review across a large census.
One shared worklist aligned palliative, care management, and hospitalist teams.
Prioritization is decision-support only — the conversation stays with the clinical team.
Thresholds were calibrated to the center's own population and goals-of-care history.
We weren't being asked to make a different decision — we were being handed the patients we would have wanted to see anyway, earlier, and with the context that made the conversation better.
Representative quote illustrating the kind of feedback this engagement is designed to produce — not attributed to a named individual.
The Platform Behind This
Advanced Illness Identification
Find the patients who need a goals-of-care conversation — before a crisis forces one.
Frequently Asked
Questions About This Scenario
More Case Studies