Clinical decision support (CDS)

In one line

Clinical decision support delivers knowledge — alerts, reminders, order sets, risk scores, guidelines — to a clinician (or patient) at the point of care, to improve a decision. The art is the five rights: the right info, to the right person, in the right format, through the right channel, at the right time.

The forms it takes

Drug–drug and allergy alerts, dose checks, duplicate-order detection, preventive-care reminders, order sets and care-plan templates, diagnostic and risk calculators, and — increasingly — AI predictions (deterioration, sepsis, readmission). The CDS Hooks lab lets you author these as real cards.

The standards

Modern CDS is decoupled from the EHR: CDS Hooks fires a service at a workflow moment and returns cards; SMART on FHIR apps embed richer tools; shareable logic is written once and reused. That's how a rule written in one place can run in many EHRs.

Watch for — alert fatigue

The classic failure is too many low-value alerts: clinicians learn to dismiss them, and the one critical warning is dismissed too. Good CDS is specific, actionable, and sparing — and, when it uses AI, wrapped in guardrails and human review. Noise is not safety.

CDS that clinicians trust

Good CDS earns its interruption. The bar: it fires on a specific, evidence-based trigger; it arrives at the moment of the decision, not after; it offers an action (order this, stop that), not just a warning; and it can be measured and tuned post-launch. Most CDS failures aren't wrong logic — they're poorly targeted logic that trains clinicians to dismiss it.

Key takeaways

• CDS delivers the right knowledge to the right person at the right moment — the five rights.
• Forms: alerts, reminders, order sets, risk scores, guidelines, and (increasingly) AI predictions.
• Modern CDS is decoupled via CDS Hooks / SMART on FHIR — write once, run in many EHRs.
• The enemy is alert fatigue — be specific, actionable and sparing; with AI, add guardrails + review.