AI-powered Clinical Decision Support Systems (AI-CDSS) are no longer “future tech” in critical care. They’re increasingly embedded in how clinicians monitor risk, interpret trends, and anticipate deterioration. Yet the conversation often focuses on accuracy, workflow integration, and cost—while leaving a crucial human factor underexplored: how it feels to use AI when the stakes are high.

Journal Club Article

Awad, N. H. A., Aljohani, W., Yaseen, M. M., Awad, W. H. A., Abou Elala, R. A. S. A., & Ashour, H. M. A. (2025). When Machines Decide: Exploring How Trust in AI Shapes the Relationship Between Clinical Decision Support Systems and Nurses’ Decision Regret: A Cross‐Sectional Study. Nursing in Critical Care, 30(5), e70157.

A cross-sectional study where the authors examined the relationship between ICU nurses’ reliance on AI-CDSS, their decision regret, and—most importantly—trust in AI.

Background

Decision regret matters because it is not just an individual emotion; it can ripple into confidence, professional identity, moral distress, and willingness to engage with future decisions. In intensive care—where decisions are frequent, time-critical, and ethically complex—regret can become a quiet driver of burnout and hesitancy. This paper suggests that trust may shape whether AI support feels like a safety net or a second-guessing machine.

What the researchers did

The study surveyed 250 ICU nurses across three private tertiary hospitals in Egypt (62.5% response rate).

Nurses were eligible if they were full-time ICU nurses with at least six months’ ICU experience, and were excluded if they had no exposure to AI-CDSS or were in non-clinical roles.

Three validated measures were used:

• AI-CDSS reliance (Healthcare Systems Usability Scale, HSUS)
• Decision regret (Decision Regret Scale; scored 0–100)
• Trust in AI (Trust in AI scale; 4 items)

What they found?

Reliance is linked to less regret—especially when trust is high.

On average, nurses reported moderate levels of AI-CDSS reliance, trust in AI, and decision regret: reliance M = 78.6, regret M = 38.5, trust M = 13.9.

Three results stand out:

1. More reliance on AI-CDSS was associated with less decision regret (r = −0.42).
2. More trust in AI was associated with less regret (r = −0.33).
3. Reliance and trust were positively linked (r = 0.51).

Trust strengthened the relationship between reliance and lower regret, with a significant interaction term (p = 0.012). In other words, when nurses trusted AI more, relying on it was more strongly associated with reduced regret.

AI implementations often concentrate on what the system can do. This study argues you also need to manage how the system changes nurses’ emotional load.

If trust is low, AI may amplify second-guessing (“Did I miss something?”) or shift accountability anxiety (“If the AI said X, should I have done X?”). If trust is higher, AI support may function as reassurance—reducing decision regret after difficult calls. The take-home: trust isn’t a soft extra; it’s a determinant of experience and adoption.

Practical Implementation

1) Treat “trust in AI” as a measurable implementation outcome

2) Build transparency into training: not just “how to use it,” but “when not to”

3) Strengthen psychological safety around decision ownership

4) Make workflow fit visible: usability supports reliance

5) Align governance with bedside reality: clarify accountability early