CPIA Module 5, Section 4: Usability and Interface Optimization
MODULE 5: CLINICAL DECISION SUPPORT (CDS) FUNDAMENTALS

Section 5.4: Usability and Interface Optimization

An exploration of the user experience (UX) of clinical decision support. An alert can be clinically perfect but functionally useless if it’s poorly designed.

SECTION 5.4

Usability and Interface Optimization

Designing CDS That Works for Clinicians, Not Against Them.

5.4.1 The “Why”: The Critical Last Mile of CDS Effectiveness

We have reached the final, crucial stage in our journey of CDS design. We have established the core philosophy of the Five Rights, mastered the IF-THEN logic of rule building, and confronted the existential threat of alert fatigue. We have built a clinically perfect rule, targeting the right patients with the right information at the right time. And yet, all of this meticulous work can be instantly undone in the “last mile” of delivery: the user interface. An alert that is clinically brilliant but visually confusing, textually ambiguous, or functionally awkward is a failed alert. It creates friction, wastes cognitive energy, and ultimately invites the user to override it, regardless of the validity of its content.

This section is dedicated to the art and science of that last mile: Usability and User Experience (UX) Design. If the previous sections were about what the CDS should say, this section is about how it should say it. We will explore the principles of human-computer interaction and apply them directly to the unique challenges of the clinical environment. You will learn that the choice of words, the use of color, the placement of a button, and the structure of a sentence are not minor aesthetic details; they are powerful determinants of whether your CDS will succeed or fail.

As an informatics pharmacist, you are uniquely positioned to be a master of clinical UX design. You have spent your entire career on the receiving end of poorly designed information systems. You know the frustration of a confusing prescription label, an ambiguous order, or a clunky software interface. You have a deep, intuitive understanding of what constitutes clarity versus confusion in a high-stakes environment. This section will give you the formal language and the evidence-based principles to transform that intuition into a repeatable design methodology. You will learn to be a ruthless editor of your own work, to champion the user’s perspective above all else, and to design interventions that are not just clinically sound, but are also clear, concise, actionable, and even welcome. This is how we build systems that don’t just provide decision support, but actively reduce clinician burden and make the right thing the easiest thing to do.

Retail Pharmacist Analogy: The Perfectly Designed Prescription Label

Imagine two different prescription labels for the same medication: a complex prednisone taper for a patient with a severe asthma exacerbation.

Label A (Poor UX): The directions field contains a single, massive block of text: “TAKE THREE TABLETS BY MOUTH DAILY FOR THREE DAYS THEN TAKE TWO TABLETS BY MOUTH DAILY FOR THREE DAYS THEN TAKE ONE TABLET BY MOUTH DAILY FOR THREE DAYS THEN TAKE ONE TABLET EVERY OTHER DAY FOR THREE DOSES THEN STOP.” While technically correct, this is a cognitive nightmare for a patient who is likely already feeling unwell and anxious. It’s difficult to read, impossible to remember, and is a recipe for a medication error.

Label B (Excellent UX): The pharmacist recognizes the complexity and uses their software’s “structured directions” feature. They augment the label with clear, simple formatting and auxiliary labels.
The directions read: Prednisone 10mg – Taper Schedule

  • Days 1-3: Take 3 tablets (30mg) every morning.
  • Days 4-6: Take 2 tablets (20mg) every morning.
  • Days 7-9: Take 1 tablet (10mg) every morning.
  • After Day 9: Take 1 tablet every OTHER morning for 3 doses, then stop.
A bright yellow “Take with food” sticker is also applied to the bottle.

Both labels contain the same core clinical information. But Label B is infinitely more usable. It uses whitespace, bullet points, and clear headings to break down complex information into manageable chunks. It anticipates the patient’s questions and provides clear, unambiguous instructions. It is designed with empathy for the end-user. This is the essence of usability and interface optimization. Our goal in CDS design is to create the “Label B” of clinical alerts—to present even the most complex clinical logic in a format that is effortlessly simple, intuitive, and safe for the busy clinician to act upon.

5.4.2 The Foundation: Nielsen’s 10 Usability Heuristics for Health IT

In the field of user experience design, there is a set of foundational principles that have stood the test of time. Developed by usability expert Jakob Nielsen, these 10 “heuristics” (or general rules of thumb) provide a powerful framework for evaluating and designing user interfaces. While they were created for general software design, they are profoundly applicable to the world of clinical decision support. As an informatics pharmacist, internalizing these principles will provide you with a robust checklist for ensuring your CDS is not just clinically correct, but also functionally excellent.

Masterclass Table: Applying Usability Heuristics to Pharmacy CDS
Heuristic & Core Principle Application in Pharmacy CDS Design Example of a Violation Example of Excellent Design
1. Visibility of System Status
The system should always keep users informed about what is going on.		 When a clinician acts on an alert, the system should provide immediate feedback that the action was successful. When a complex rule is running, the system should indicate that it is “thinking.” A provider accepts a renal dosing recommendation from an alert, but the dose in the ordering screen doesn’t visually update. The provider is left wondering if their click actually worked. After a pharmacist enters a vancomycin dose into a protocol calculator, the system displays a loading bar and the text “Calculating patient-specific regimen…” for a moment before showing the results.
2. Match Between System and the Real World
The system should speak the user’s language, with words and concepts familiar to the user.		 Alerts should use standard clinical terminology, not IT jargon or database field names. Dosing should be presented in familiar units (e.g., “mg,” not “0.001 grams”). An alert that reads: “Error: MED_ID 734 conflicts with ALLERGY_CLASS 12. Rule 401b triggered.” This is meaningless to a clinician. An alert that reads: “Allergy Alert: The order for Ceftriaxone may cross-react with the patient’s documented Penicillin allergy.”
3. User Control and Freedom
Users need a clearly marked “emergency exit” to leave an unwanted state. Support undo and redo.		 Soft-stop alerts must have a clear and obvious override button. If a user enters an order set by mistake, there should be an easy “Cancel” or “Back” button that discards all selections. A hard-stop alert for a non-critical issue with no override path, forcing the physician to abandon their entire order and start over to bypass it. A soft-stop renal dosing alert provides a clear “Accept Recommendation” button and an equally clear “Override – Manually Enter Dose” button, giving the user control.
4. Consistency and Standards
Users should not have to wonder whether different words or actions mean the same thing.		 All alerts across the entire health system should follow a consistent visual layout. The “Accept” button should always be in the same place, use the same color, and use the same language. The renal dosing alert has the “Accept” button on the left, but the drug interaction alert has it on the right. This inconsistency forces the user to stop and think, increasing cognitive load. The institution establishes a “CDS Style Guide” that mandates all alerts will have a bold headline, a brief explanation, and two buttons: “Accept Recommendation” (green) and “Close” (gray).
5. Error Prevention
Even better than good error messages is a careful design which prevents a problem from occurring in the first place.		 This is the core of passive CDS. Instead of letting a user order the wrong dose and then alerting them, guide them to the right dose from the start. Allowing a provider to type “500mg” into a dose field for a drug that is only available as a 250mg tablet, and then firing an error message. Using a structured dropdown menu for the dose, which only contains the available tablet strengths (e.g., “125mg,” “250mg”). This design makes it impossible to order the wrong dose.
6. Recognition Rather Than Recall
Minimize the user’s memory load by making objects, actions, and options visible.		 Instead of forcing a user to remember a complex protocol, present it to them in an order set. Instead of making them recall a patient’s last lab value, display it within the ordering screen. An alert that says “Adjust dose for renal function” but does not display the patient’s latest creatinine or calculated CrCl. A heparin ordering screen that clearly and persistently displays the patient’s current weight, aPTT, and the goal aPTT range, so the user doesn’t have to memorize them.
7. Flexibility and Efficiency of Use
Allow users to tailor frequent actions. Accelerators for expert users can speed up the interaction.		 For common overrides, allow experienced users to handle them efficiently. Create personalized order sets or “favorites” for specialists who order the same things repeatedly. Requiring a detailed, typed reason for overriding a very common, low-risk drug interaction alert every single time. For a common alert, provide pre-canned override reasons as clickable buttons (e.g., “Benefit outweighs risk,” “Monitoring plan in place”) in addition to a free-text option.
8. Aesthetic and Minimalist Design
Interfaces should not contain information which is irrelevant or rarely needed. Every extra unit of information competes with the relevant units.		 Alerts should be brutally concise. Remove every unnecessary word. Use whitespace to guide the eye. Don’t clutter an alert with academic information that isn’t needed for the immediate decision. A vancomycin dosing alert that includes a long paragraph on the drug’s mechanism of action, history, and spectrum of activity before getting to the dosing recommendation. An alert that reads: “Renal Dosing Alert. Pt eGFR: 22 mL/min. Rec: 1g IV Q24H.” It is minimalist, data-rich, and immediately understandable.
9. Help Users Recognize, Diagnose, and Recover from Errors
Error messages should be expressed in plain language, precisely indicate the problem, and constructively suggest a solution.		 This is the formula for a perfect alert text. State the problem clearly, explain why it’s a problem, and tell the user how to fix it. An error message that says “Invalid Dose.” An error message that says: “Invalid Dose: This medication cannot be dosed higher than 40mg/day. Please select a lower dose.”
10. Help and Documentation
It’s best if the system doesn’t need any additional explanation. However, it may be necessary to provide documentation to help users.		 For very complex CDS tools, like a pharmacist-driven anticoagulation management protocol, there should be an easily accessible infobutton that links to the full institutional policy document. A complex oncology order set with no guidance or links to the underlying treatment protocol, forcing the user to hunt for the documentation on the hospital’s intranet. An infobutton labeled “View Protocol Details” is placed at the top of a complex order set, which opens a concise summary of the evidence and guidelines the order set is based on.

5.4.3 Masterclass in Microcopy: The Art of Writing Effective Alert Text

In the world of UX design, microcopy refers to the small, specific pieces of text that guide a user through an interface—the labels on buttons, the text in error messages, the instructions in a pop-up. In CDS, microcopy is everything. The few words you choose for your alert text can make the difference between an intervention that is acted upon and one that is immediately dismissed. Effective alert writing is a skill that combines clinical knowledge, empathy, and ruthless brevity.

The A-B-C Formula for Perfect Alert Text

When writing the text for any interruptive alert, strive to follow this simple, three-part formula to ensure maximum clarity and actionability.

A: Attention

State the core problem in a bold headline.

B: Because

Briefly explain the clinical reason or show the key data.

C: Course of Action

Provide a clear, specific, recommended next step.

Let’s apply this to a real-world example:

(A) Attention: Hyperkalemia Risk Alert

(B) Because: Patient’s potassium from 06:00 is 5.8 mmol/L. This order for Spironolactone will further increase potassium.

(C) Course of Action: Recommendation: Discontinue or hold Spironolactone. Consider treatment for hyperkalemia.

The Do’s and Don’ts of CDS Microcopy
Principle Don’t (The Common Mistake) Do (The Best Practice)
Use an Active Voice “The dose should be adjusted by the provider.” (Passive) “Adjust dose for renal function.” (Active, a direct command)
Be Specific and Data-Driven “High potassium detected.” (Vague) “Potassium is 5.8 mmol/L.” (Specific)
Front-load the Important Information “This is a drug interaction alert. According to studies, co-administration of Drug A and Drug B can increase the risk of a rare side effect…” (Buries the lede) “High Bleeding Risk: Avoid combining Warfarin and Amiodarone.” (Headline first)
Avoid Jargon and Acronyms “D/C med, re-eval in AM, check BMP.” (Insider shorthand) “Recommendation: Discontinue medication and re-evaluate in the morning. Check a Basic Metabolic Panel.” (Clear and unambiguous)
Write for Scannability A single, dense paragraph of text that is hard to read quickly. Use bolding for key terms, bullet points for lists, and whitespace to separate ideas.
Design Actionable Buttons Buttons labeled “OK” or “Cancel.” (What does “OK” mean? OK to ignore? OK to accept?) Buttons with clear actions: “Accept Recommended Dose,” “Override & Keep Current Dose,” “Close & Re-evaluate.”

5.4.4 Visual Design and Information Hierarchy

While the text of an alert is critical, the visual presentation can have an equally profound impact on its usability. Effective visual design is not about making things “pretty”; it’s about using principles of color, layout, and typography to reduce cognitive load and guide the user’s attention to the most important information. As an informatics pharmacist, you don’t need to be a graphic designer, but you do need to understand the fundamental principles of clear information display.

The Power and Peril of Color

Color is one of the fastest and most powerful ways to convey meaning, but it must be used consistently and sparingly.

  • Red: Should be reserved exclusively for critical warnings, errors, and “hard stop” conditions. Using red for routine, informational alerts will devalue its meaning and contribute to fatigue.
  • Yellow/Amber: Ideal for moderate warnings, “soft stops,” or situations that require caution and user attention.
  • Green/Blue: Good for informational messages, confirmations, or highlighting recommended actions.
  • Accessibility: Always ensure there is a secondary method to convey information besides color alone (e.g., an icon, bold text). A significant portion of the population has some form of color blindness, and relying on color alone will make your CDS unusable for them.
Layout and a “Good” vs “Bad” Alert Design

BAD DESIGN

ALERT

A drug-drug interaction was detected between the lisinopril you ordered and the patient’s home potassium chloride. The concurrent use of these agents may increase the risk for hyperkalemia. Please monitor the patient’s labs as appropriate. The patient’s most recent potassium was 4.1 on 10/17/25.

Why it fails:
  • Vague Title: “ALERT” provides no information.
  • Buried Lede: The key data (K=4.1) is at the very end.
  • Wall of Text: The single paragraph is hard to scan.
  • Passive Voice: “was detected” is weak.
  • Unactionable Advice: “monitor labs” is the standard of care, not helpful advice.
  • Ambiguous Button: “OK” is meaningless.

GOOD DESIGN

Hyperkalemia Risk Warning

Ordering Lisinopril for a patient on Potassium Chloride.

Most Recent Lab:
Potassium: 4.1 mmol/L (on 10/17/25)

Risk is low with normal renal function and current K+. Proceeding is acceptable with routine monitoring.

Why it works:
  • Clear Headline: States the specific clinical risk.
  • Data Up Front: The key lab value is highlighted and easy to find.
  • Minimalism: Uses whitespace and a container to break up information.
  • Actionable Recommendation: Clearly states that proceeding is acceptable, resolving the user’s question.
  • Clear Buttons: The button labels describe the exact action they will perform.