, Question 1
A nursing informatics team is selecting an AI-powered sepsis alert for the
EHR. The vendor claims 95% sensitivity but provides no AUROC, calibration
curve, or subgroup performance data. Which evaluation principle is MOST
critically violated?
A. The alert lacks real-time interoperability with the pharmacy system.
B. The vendor omitted measures of discrimination, calibration, and
fairness necessary to judge clinical utility.
C. The alert does not comply with HL7 FHIR R4 standards.
D. The sensitivity threshold exceeds the FDA's 90% requirement for
SaMD.
Correct Answer: B - The vendor omitted measures of
discrimination, calibration, and fairness necessary to judge
clinical utility.
RATIONALE
Sensitivity alone is insufficient; discrimination (AUROC), calibration,
and subgroup fairness are essential to assess whether an AI tool will
generalize safely. The other options reference standards not mandated
in this context or misstate FDA requirements.
Question 2
Which regulatory pathway classifies a nursing-facing AI chatbot that triages
patient symptoms and recommends home care as a medical device requiring
FDA clearance?
A. Clinical Decision Support Software (CDSS) exemption under 21st
Century Cures Act
B. Software as a Medical Device (SaMD) under FDA oversight
C. Health IT Module Certification under ONC
D. General Wellness Product exemption
Correct Answer: B - Software as a Medical Device (SaMD) under
Page 2
,FDA oversight
RATIONALE
A chatbot that triages symptoms and recommends care meets the
definition of SaMD because it informs clinical management. CDSS
exemptions apply only when the clinician can independently review
the basis, which is not the case here.
Question 3
In evaluating an AI-powered clinical documentation tool, which metric BEST
reflects the tool's ability to reduce nursing documentation burden without
introducing errors?
A. Word error rate (WER) alone
B. Time saved per note combined with clinically significant error rate
C. Number of notes generated per shift
D. User satisfaction score from a single survey
Correct Answer: B - Time saved per note combined with
clinically significant error rate
RATIONALE
True burden reduction must balance efficiency gains with safety; time
saved is meaningful only if clinically significant errors remain low.
Other metrics are incomplete or misleading in isolation.
Question 4
A nurse manager must choose between two AI triage tools. Tool A has higher
accuracy but requires cloud data transfer; Tool B has slightly lower accuracy
but runs on-premises. Which framework BEST guides this decision?
A. Accuracy-first model selection
B. Total cost of ownership analysis
C. Privacy, security, and regulatory compliance risk assessment
D. Vendor market share and brand reputation
Page 3
, Correct Answer: C - Privacy, security, and regulatory compliance
risk assessment
RATIONALE
Cloud data transfer introduces privacy, security, and compliance risks
(e.g., HIPAA, state laws) that may outweigh marginal accuracy gains.
A structured risk assessment is essential in healthcare AI selection.
Question 5
Which statement BEST characterizes the concept of 'human-in-the-loop' in
nursing AI deployment?
A. The AI system operates autonomously and alerts nurses only for
failures.
B. Nurses retain authority to override or confirm AI recommendations
based on clinical judgment.
C. Nurses must follow all AI recommendations to ensure consistency.
D. The AI system replaces the nurse's decision-making in routine cases.
Correct Answer: B - Nurses retain authority to override or
confirm AI recommendations based on clinical judgment.
RATIONALE
Human-in-the-loop means clinicians maintain decision authority and
can override AI, preserving accountability and contextual judgment.
Full autonomy or mandatory compliance undermines safe practice.
Question 6
An AI tool for fall risk prediction performs well in the development hospital
but poorly when deployed in a community hospital. Which phenomenon
MOST likely explains this?
A. Overfitting to the development site's data distribution
B. Insufficient training data volume
Page 4
A nursing informatics team is selecting an AI-powered sepsis alert for the
EHR. The vendor claims 95% sensitivity but provides no AUROC, calibration
curve, or subgroup performance data. Which evaluation principle is MOST
critically violated?
A. The alert lacks real-time interoperability with the pharmacy system.
B. The vendor omitted measures of discrimination, calibration, and
fairness necessary to judge clinical utility.
C. The alert does not comply with HL7 FHIR R4 standards.
D. The sensitivity threshold exceeds the FDA's 90% requirement for
SaMD.
Correct Answer: B - The vendor omitted measures of
discrimination, calibration, and fairness necessary to judge
clinical utility.
RATIONALE
Sensitivity alone is insufficient; discrimination (AUROC), calibration,
and subgroup fairness are essential to assess whether an AI tool will
generalize safely. The other options reference standards not mandated
in this context or misstate FDA requirements.
Question 2
Which regulatory pathway classifies a nursing-facing AI chatbot that triages
patient symptoms and recommends home care as a medical device requiring
FDA clearance?
A. Clinical Decision Support Software (CDSS) exemption under 21st
Century Cures Act
B. Software as a Medical Device (SaMD) under FDA oversight
C. Health IT Module Certification under ONC
D. General Wellness Product exemption
Correct Answer: B - Software as a Medical Device (SaMD) under
Page 2
,FDA oversight
RATIONALE
A chatbot that triages symptoms and recommends care meets the
definition of SaMD because it informs clinical management. CDSS
exemptions apply only when the clinician can independently review
the basis, which is not the case here.
Question 3
In evaluating an AI-powered clinical documentation tool, which metric BEST
reflects the tool's ability to reduce nursing documentation burden without
introducing errors?
A. Word error rate (WER) alone
B. Time saved per note combined with clinically significant error rate
C. Number of notes generated per shift
D. User satisfaction score from a single survey
Correct Answer: B - Time saved per note combined with
clinically significant error rate
RATIONALE
True burden reduction must balance efficiency gains with safety; time
saved is meaningful only if clinically significant errors remain low.
Other metrics are incomplete or misleading in isolation.
Question 4
A nurse manager must choose between two AI triage tools. Tool A has higher
accuracy but requires cloud data transfer; Tool B has slightly lower accuracy
but runs on-premises. Which framework BEST guides this decision?
A. Accuracy-first model selection
B. Total cost of ownership analysis
C. Privacy, security, and regulatory compliance risk assessment
D. Vendor market share and brand reputation
Page 3
, Correct Answer: C - Privacy, security, and regulatory compliance
risk assessment
RATIONALE
Cloud data transfer introduces privacy, security, and compliance risks
(e.g., HIPAA, state laws) that may outweigh marginal accuracy gains.
A structured risk assessment is essential in healthcare AI selection.
Question 5
Which statement BEST characterizes the concept of 'human-in-the-loop' in
nursing AI deployment?
A. The AI system operates autonomously and alerts nurses only for
failures.
B. Nurses retain authority to override or confirm AI recommendations
based on clinical judgment.
C. Nurses must follow all AI recommendations to ensure consistency.
D. The AI system replaces the nurse's decision-making in routine cases.
Correct Answer: B - Nurses retain authority to override or
confirm AI recommendations based on clinical judgment.
RATIONALE
Human-in-the-loop means clinicians maintain decision authority and
can override AI, preserving accountability and contextual judgment.
Full autonomy or mandatory compliance undermines safe practice.
Question 6
An AI tool for fall risk prediction performs well in the development hospital
but poorly when deployed in a community hospital. Which phenomenon
MOST likely explains this?
A. Overfitting to the development site's data distribution
B. Insufficient training data volume
Page 4
