9.2 Clinical Decision Support Types and Design

9.2 Clinical Decision Support Types and Design

Quick Answer: Design CDS with the five rights: right information, right person, right intervention format, right channel, and right time in workflow. Interruptive alerts spend attention; non-interruptive tools (infobuttons, defaults, displays) often do more work. Order checks and BPAs are different tools. More alerts are not safer.

Task A.6 is not only a maintenance calendar. CPHIMS also asks you to choose a mechanism clinicians can actually use. A correct fact delivered as the wrong interrupt still fails. This section is the design language: five rights, interruptive versus non-interruptive, infobuttons, order checks, and best-practice alerts (BPAs).

CPHIMS practice questionsPractice questions with detailed explanations

Why design type is a scored skill

Organizations reach for a modal alert because it is visible to leadership. Visibility is not value. The cheapest interrupt to build is often the most expensive attention tax. Good design asks what decision is being made, who can change it, and how little friction is required to make the safe action the easy action.

Exam stems usually give you a true clinical fact, then offer a loud interrupt. The correct answer is often a quieter type that still meets the five rights.

The five rights

HIMSS-style CDS teaching uses Osheroff’s five rights. Memorize them as design tests, not a poster on the informatics wall.

Right information is not enough. A potassium warning can be factually correct and still fail if it interrupts a nephrologist who already documented the plan, or if it fires after the dose is given. When a stem says “the information is right, so the alert is well designed,” check the other four rights.

Interruptive versus non-interruptive

Non-interruptive CDS stays in the user’s peripheral vision or in the default path: infobuttons, hover definitions, relevant-data displays (last creatinine next to a vancomycin order), order-set defaults, pick-list constraints, passive icons, and worklists. The user can continue without clicking through a modal.

Interruptive CDS stops the task: modal BPAs, hard stops, and many order checks at sign. Use it when harm is high, the user can act now, and a quieter method cannot prevent the action.

Interruptive CDS must earn the interruption. If the same click already has three modals, the fourth is not additive safety. It is a queue that trains dismissal.

Hard stop versus soft stop. A hard stop is appropriate only when proceeding is almost never correct and a documented bypass path exists for the rare exception (for example, a look-alike chemotherapy dose that needs a second authorized signature). A soft stop warns and allows continue. Turning every caution into a hard stop is how you create unsafe workarounds: paper orders, borrowed logins, and a culture of clicking through.

Major CDS types you must tell apart

Infobuttons. Context-aware knowledge links on a problem, medication, or lab. They deliver reference without stopping the order. Use them for rare tests, patient education, and “tell me more” needs. They are not a substitute for a dose hard stop.

Order checks. Drug–allergy, drug–drug, dose-range, duplicate therapy, formulary, and sometimes pregnancy or geriatric checks. They typically fire at order entry or sign. Severity must be tuned: information, warning, or hard stop. An un-tuned drug–drug pack is how a hospital generates thousands of interrupts nobody reads.

Order sets and protocols. These are choice-architecture CDS. Putting the bundle in the right sequence, with the right defaults, often prevents the error that a later BPA would only scold. CPHIMS will tempt you to “add an alert” to a bad catalog; the better design is often to fix the set.

Documentation prompts and flowsheet rows. They capture the data later CDS and measures need (sepsis screening, fall risk, last bowel movement before an opioid). They are CDS when they change what is noticed; they are clutter when they capture unused scores.

Best-practice advisories (BPAs) and similar alerts. Rule-based interrupts or banners: sepsis, VTE prophylaxis, isolation, duplicate imaging, missing problem on an anticoagulant. Design them as one purpose, one owner, one recommended action—not a newsletter pasted into a modal.

Design rules that survive the exam

• Prefer preventing the unsafe choice (catalog constraint, default, order set) over scolding after the click.
• Make the recommendation one action, not a paragraph.
• Give each rule one owner and one purpose.
• Do not stack multiple interrupts on the same sign-click.
• Match role and channel: pharmacy verify is the right channel for some medication issues; the bedside nurse is the right person for others.
• Test with the real role in the real workflow, including night shift and trainees.
• Measure firing and override before you congratulate yourself on “going live with CDS.”

More alerts are not safer

After a missed diagnosis, leadership will ask for more BPAs. More alerts ≠ safer. Alert volume is not a safety outcome. High firing plus high override trains clinicians to dismiss the one interrupt that mattered. That is a design failure, not a training opportunity you can fix with another e-learning module. The five rights are how you refuse a volume target and still improve the decision.

If you remember only one design sentence for task A.6: pick the quietest tool that puts the right fact in front of the person who can act, at the moment they can still change the plan.

Scenarios and exam traps

Scenario. A missed sepsis case produces a demand to double BPAs on every vital-sign file. Diagnose the miss first. If the screening row was never in the nursing workflow, a louder modal on the attending after admission is the wrong person and the wrong time.

Scenario. A resident needs reference on a rare endocrine assay while ordering. An infobutton or non-interruptive knowledge link fits. A hard stop that pages endocrinology on every order does not.

Scenario. Nephrology is drowning in potassium drug–drug interrupts for patients they are already managing. The information can be right and the design still wrong. Change format, person, or timing; do not add a second interrupt.

Watch these traps:

1. Equating more alerts with more safety.
2. Treating the five rights as “right information” only.
3. Using a BPA where an order set, default, or catalog constraint would prevent the choice.
4. Firing on the person who cannot change the order.
5. Collapsing infobuttons, order checks, and BPAs into one idea called “pop-ups.”
6. Hard-stopping everything “so nobody can ignore it,” which creates workarounds.

Design choices become governance metrics in 9.3. If you ship the wrong type, no committee dashboard will save you.