Advisory — clinical AI and the end of life
I have had the conversation your algorithm is about to trigger.
Deterioration and mortality models are live in health systems now. What almost none of them have decided is what happens next: who tells the patient, what the family is told about the model, in what order, in what words, and what the system must never say on a clinician's behalf. That is the work I do.
Nikki Patton · Advisory, clinical AI and the end of life
The gap
Governance frameworks stop at the bedside door.
In June 2026 the Joint Commission launched its Responsible Use of AI in Healthcare certification. Its standards are organized around five major areas: governance; effective data management; risk and bias reduction; monitoring, evaluating and validating safety, performance and responsible use; and transparency, education and training. CHAI's governance playbooks are the framework it was built from.
Four of those five areas govern the model. One governs what happens between people — and it carries transparency, education and training all at once. That single area holds the hardest problem in medicine, and it is close to empty. A model surfaces that a patient is likely in their final months. A human being then has to say something to a family.
Engineers cannot design that moment. Ethicists can describe it but rarely operationalize it. Most governance committees have never had to hold it. It is the single place where responsible AI stops being a policy question and becomes a person in a room.
And the clinician who inherits it is often unprepared — not through any failing of theirs, but because delivering a poor prognosis with transparency and grace is a skill almost nobody is formally taught.
The model is not the risk. The silence after it fires is the risk.
The work
Three ways I am engaged.
All work is scoped, time-bound, and delivered as artifacts your governance committee can actually file.
-
Governance advisory
Health systems & hospital leadershipBuilding the patient-transparency and staff-education layer of your AI governance program for prognosis-adjacent and deterioration models. Disclosure language, escalation sequencing, clinician scripting, documentation that survives audit, and the competency assessment behind it.
-
Product and language review
Health technology companiesPre-ship review of any feature that touches prognosis, decline, or end of life. What the interface says, what it defaults to, what it escalates, and the red lines it must not cross. Written by someone who has sat with the families your output lands on.
-
Expert seat
Committees, panels, standards workNamed participation on AI governance committees, advisory boards, standards workgroups, and public comment. Accountability requires a human signature. I put my name on the recommendation.
The artifact
The Human Handoff Standard
A governance artifact, not a course. What a health system does in the seventy-two hours after a model flags a patient as likely to die.
Disclosure
What a patient and family are told about the model that shaped a prognosis, in language a frightened person can hold.
Sequencing
Who is notified, in what order, within what window, and who is never the first to know.
Scripting
Clinician language for AI-informed prognosis conversations, built from what actually works in the room.
Documentation
What gets recorded, in what form, so the conversation is defensible a year later.
Competency
An assessment that satisfies a staff-education requirement with something more than an attestation.
Red lines
What the system is never permitted to communicate directly, and what must always route to a person.
Why me
Three things that rarely sit in one person.
Presence
Two decades across post-acute care, through vice president level, sitting with families at the moment a prognosis lands. Not studied — done. I have also been on the other side of it, as the family caregiver, five times over.
Operating pressure
I have carried the number. I know precisely how growth targets distort eligibility decisions, because I have led the teams making them. That is the failure mode your model will create, and most advisors cannot see it.
Technical fluency
I currently serve as Vice President of Sales at Curantis Solutions, a hospice technology company — disclosed on every engagement, and independent of the advisory work here. I know what a roadmap is, what ships, and what "the model flags it" means operationally.
- 20 yearsAcross post-acute and end-of-life care, through vice president level
- 175 → 550Average daily census growth led in 18 months, with full staff retention
- 900+ ADCConcurrent regional oversight across the Southwest, from start-up through multi-site expansion
- HarvardBusiness School Executive Education certificate in AI-Driven Healthcare Transformation
- 3 booksTwo published, one forthcoming — including a co-authored trade title
Writing & talks
The work behind the advisory.
Books, keynotes, and plain language for the conversations nobody is taught to have. This is where the method came from.
-
Book
Sacred Conversations
The method for serious illness conversation that underpins everything on this page.
- Book — forthcoming Say What Matters, While It Matters 20 lessons from my life inside hospice. Anticipated 2027. Publisher to be confirmed before listing.
- Speaking Keynotes & sessions Governance forums, clinical leadership meetings, and industry conferences. Signature keynote: Say What Matters, While It Matters.
- Consulting firm Always Human Separate practice serving hospice owners and operators. Distinct engagements, distinct terms.
Contact
If a model in your system can predict death, someone should decide what it says.
Advisory inquiries, committee invitations, and speaking requests all come here.