Provider Trust and Workflow Challenges in AI Dosing

• Attitudes: Mixed optimism and caution; recognized potential benefits but significant reservations.

• Perceived Benefits: Improved dosing precision, reduced adverse drug events, lower workload.

• Key Concerns:

• Suitability for complex/atypical patients.

• Erosion of skills and critical thinking.

• Trust and transparency in “black box” algorithms.

• Liability and accountability for errors.

• Workflow integration challenges.

• Facilitators/Enablers:

• Human-in-the-loop oversight.

• Transparency & explainability (clear rationale, confidence levels).

• Education & training (understanding data sources, validation).

• Technical robustness and reliability.

• Seamless workflow integration into EHR and pharmacy systems.

• Role-Specific Insights:

• Physicians: valued cognitive relief but wary of autonomy/liability.

• Pharmacists: emphasized accuracy, data completeness, cross-checking.

• Nurses: focused on workflow fit, administration timing, safety.

• Precision Dosing Challenges: Tailoring medication doses to individual patient characteristics is complex and labor-intensive. Traditional one-size-fits-all dosing often misses the mark, especially for drugs with narrow therapeutic windows or variable pharmacokinetics (e.g., vancomycin, warfarin).

• Rise of AI in Dosing: AI-driven algorithms like CURATE.AI analyze patient-specific data to recommend personalized doses, promising improved safety and dynamic therapy adjustment.

• Need for Acceptance: Provider trust and usability are critical—poor CDS adoption history shows that without credibility and workflow fit, tools are ignored.

• Design: Qualitative interview study, analyzed via UTAUT framework.

• Participants: 16 HCPs (9 physicians, 4 nurses, 3 pharmacists) at Alexandra Hospital, Singapore.

• Intervention: CURATE.AI presented as a real-world example during semistructured interviews.

• Analysis: Deductive coding + thematic analysis; focused on attitudes, usefulness, trust, and workflow conditions.

• Pharmacists and AI Dosing (Vancomycin, Liu et al., 2023): Enthusiasm tempered by low acceptance in practice; distrust, transparency gaps, poor EHR integration hindered uptake.

• Physician Perspectives (Vijayakumar et al., 2023): Physicians want oversight, transparency, proof of benefit; adoption evolves from tolerance in early trials to demand for seamless integration.

• Systematic Review (Henzler et al., 2025): Catalogued 43 barriers and 49 facilitators; barriers include autonomy concerns, skill atrophy, liability, privacy; facilitators include efficiency, improved decision-making, and leadership endorsement.

Key Question: What are healthcare providers’ perspectives on AI-guided precision dosing tools, and what barriers or facilitators do they identify for integrating such AI tools into the medication-use process across the hospital?

• Single-site (Singapore) with 16 participants limits generalizability.

• Potential selection bias (participants possibly tech-enthusiastic).

• Hypothetical tool (CURATE.AI concept, not real-world experience).

• UTAUT framework may constrain analysis.

• Possible bias since researchers were linked to CURATE.AI development.

• Integration: Ensure AI dosing aligns across prescribing, verification, dispensing, administration, and monitoring.

• CDS Hooks & FHIR: Trigger AI in real time during order entry/modification.

• Advisories:

• OPAs for passive display of AI dose.

• BPAs for significant deviations with context provided.

• Visual Flags: Icons/flags to highlight discrepancies, enable feedback loops.

• Dispensing/Pyxis Integration: Adjust stock configurations for AI-suggested doses.

• Administration: Sync MAR, BCMA, and infusion pumps with AI adjustments.

• Monitoring Dashboards: Track override rates, therapeutic outcomes, pharmacovigilance.

• Governance: Establish update/validation workflows for AI models.

• Collaboration: Multidisciplinary review (IT, data science, pharmacy, nursing).

• Training: Education and quick-reference guides for all end-users.

Sumner J, Mohamed Ali J, Motani M, et al. Artificial intelligence guided dosing decisions: a qualitative study on health care provider perspectives. BMJ Health & Care Informatics. 2025;32(1):e101461. doi:10.1136/bmjhci-2025-101461