NR 587AI Week 7 AI Disclosure Brief Example

Disclosure Brief: What Patients and Nurses Are Told About the Deterioration Index, in Drafted Sentences, at Named Moments, Including What to Say When the Score Is Wrong

Student Name

College of Nursing, Chamberlain University

NR 587AI: Advanced Nursing Leadership in Artificial Intelligence-Integrated Healthcare Environments

Instructor Name

Month Day, Year

Disclosure Brief: What Patients and Nurses Are Told About the Deterioration Index, in Drafted Sentences, at Named Moments, Including What to Say When the Score Is Wrong

The Position

Patients on our four medical-surgical units should know that a computer score helps nurses watch for changes in their condition, and that a nurse, not the score, decides what happens. Nurses should know what feeds the score, where it falls short and which response the hospital expects from them. These are different messages and are written separately.

For Patients: The Sentences

Admission information sheet, under "How we keep watch over you":

"Our record system looks at your vital signs, blood tests and nursing notes every 15 minutes and gives your nurses a score that can show if your condition may be changing. It is one of several things your nurses use. Your nurses and doctors make all decisions about your care. If you or your family notice something is wrong, tell us, whatever any score says."

At the bedside, when a nurse comes because of a flag:

"I've come to check on you because the system flagged a change in some of your numbers. It's often nothing, but I want to see how you're doing."

If a patient asks whether a computer is deciding their care:

"No. The score helps us notice changes. I decide what to do, with your doctor."

For Patients: The Moments

The written sentence is given at admission, on the information sheet every patient receives.

The bedside sentence is said when a nurse assesses a patient because of a flag, if the patient is awake and asks or seems concerned.

Patients are not told their numerical score. A number without context could alarm or falsely reassure.

For Nurses: What They Need

What the score uses: vital signs, laboratory results, nursing assessments and age, updated only when data are entered.

What it misses: about a third of patients who deteriorate are not flagged in advance. Scores at night may rest on data several hours old. Patients with cancer and older patients often score high without being in danger.

What is expected: assess any patient whose score reaches 60, document a response within 60 minutes and escalate if your assessment, not the score, says so. A score below 60 never overrides your concern.

Who watches the tool: the nurse informaticist each week and the quality council each month, with a route to report concerns.

For Nurses: The Forum

A 10-minute segment at each unit's staff meeting in the month the brief is approved, delivered by the nurse informaticist with the unit's nurse manager. A one-page summary posted at each nurses' station. A short section in the online module that every nurse completes for the record system.

Three Months After Go-Live: The New Hire

A nurse hired next spring did not attend the staff meeting. The orientation checklist for every new nurse on the four units will include the one-page summary and a 15-minute conversation with a preceptor about the score. The nurse informaticist checks each quarter that new hires have completed both.

The Wrong Recommendation

When the score is wrong, two things matter: what the patient is told and what the nurse does.

If a nurse assesses a flagged patient and finds nothing wrong, nothing needs to be said beyond the bedside sentence. That is the score working as intended.

If a patient deteriorates without a flag, and the patient or family later asks whether the system should have caught it, the nurse or manager says:

"The score helps us notice changes, but it doesn't catch every one, which is why your nurses check on you regularly and why we listen when you tell us something is wrong. We are reviewing what happened."

This is not a full disclosure conversation about an error; if a review finds that a delay in care caused harm, the hospital's existing disclosure policy applies, as it would for any other cause.

For nurses, the message is that a missed flag is expected sometimes and is not a reason to trust their own assessment less.

Why Tell Patients at All

Most patients will never notice the score. But when a nurse arrives at the bedside at 05:30 because a computer flagged a change, the patient or family may ask why. Predictive tools also make mistakes in both directions; a widely used sepsis model missed many cases outside the hospitals where it was built (Wong et al., 2021). Patients who know a score is one input among several are less likely to be falsely reassured by its silence or alarmed by its alerts.

Questions Staff May Be Asked

The one-page summary includes three questions patients and families ask most often, with short answers: whether a computer is making decisions (no), whether the patient can see the score (not routinely; the nurse can explain what it shows) and whether the score is used for billing (no).

On Regulation

I am not in a position to state precisely what federal rules require hospitals to tell patients about predictive tools. As I understand it, recent federal certification rules for health information technology require developers to make information about predictive decision support available to the clinicians who use it. That supports the nurse-facing part of this brief. The patient-facing part rests on the principle that patients should understand how their care is decided, not on a specific rule. The hospital's legal counsel should confirm both before the information sheet is printed.

Evidence Behind the Approach

Telling nurses plainly what the tool misses addresses the risk of automation bias, which is common where clinicians rely on decision support (Goddard et al., 2012). Keeping patient wording short, and centered on the nurse's role, follows the principle in published proposals for model labels that users should know a model's purpose and limits (Sendak et al., 2020).

Who Delivers the Message

Nurses hear about the score from their own nurse manager and the nurse informaticist together, so that the clinical and technical answers are both in the room. Patients hear about it from the nurse at their bedside, not from a leaflet alone.

Approval and Review

The brief goes to the quality council in May 2027 and to patient experience and legal counsel before the information sheet changes. The wording is reviewed after six months, including by asking ten patients whether the sentence was clear.

References

Goddard, K., Roudsari, A., & Wyatt, J. C. (2012). Automation bias: A systematic review of frequency, effect mediators, and mitigators. Journal of the American Medical Informatics Association, 19(1), 121-127. https://doi.org/10.1136/amiajnl-2011-000089

Sendak, M. P., Gao, M., Brajer, N., & Balu, S. (2020). Presenting machine learning model information to clinical end users with model facts labels. npj Digital Medicine, 3, 41. https://doi.org/10.1038/s41746-020-0253-3

Wong, A., Otles, E., Donnelly, J. P., Krumm, A., McCullough, J., DeTroyer-Cooley, O., Pestrue, J., Phillips, M., Konye, J., Penoza, C., Ghous, M., & Singh, K. (2021). External validation of a widely implemented proprietary sepsis prediction model in hospitalized patients. JAMA Internal Medicine, 181(8), 1065-1070. https://doi.org/10.1001/jamainternmed.2021.2626