top of page
Search

How to write Living Algorithms that actually help clinicians

  • Apr 11
  • 3 min read

Updated: 3 days ago


Good treatment algorithms don't just summarize information. They guide decisions. The best algorithms are specific, patient-centered and structured around real clinical workflows so clinicians and trainees can quickly understand what to do next at the point of care.

Why most algorithms fall short

Treatment algorithms are everywhere in oncology, but many are difficult to use in practice.

They are often:

  • Too broad

  • Too abstract

  • Not aligned with real clinical workflows

As a result, clinicians still need to interpret, adapt and fill in gaps before making a decision. This defeats the purpose.

What a great algorithm actually does

A strong algorithm answers a simple question: what should I do next for this patient?

In order to do that, it needs to:

  • Start with the patient

  • Follow a clear pathway

  • Highlight key decisions

  • Provide just enough context

Not more information. Better structure.

Core principles for writing better algorithms

1. Be specific

Specificity is what makes an algorithm usable. Instead of broad topics like breast cancer or lung cancer, define the exact clinical scenario:

  • Stage

  • Subtype

  • Biomarker

For example:

  • Stage IV NSCLC (EGFR-mutant)

  • Early-stage HER2+ breast cancer

This immediately tells the clinician: "I'm in the right place."

It also improves discoverability in search and LLMs.

2. Optimize for clinical decision making

Algorithms should follow how clinicians actually think, not how documents are structured.

A typical workflow might look like:

  1. Diagnosis

  2. Staging

  3. Biomarker testing

  4. First line treatment

  5. Subsequent lines

Or more detailed:

  • Imaging → staging → tissue confirmation → ECOG → treatment selection

The key is to prioritize decisions and action steps, not background information.

3. Think like a trainee

Many users are fellows, residents or community oncologists covering multiple diseases. This means they're synthesizing large amounts of information quickly.

Your algorithm should help them:

  • Orient themselves

  • Understand the logic

  • Avoid common mistakes

Where helpful, be sure to include:

  • Brief context for decisions

  • Short rationale (1-2 lines)

  • Common pitfalls or exceptions

Be detailed, but concise.

4. Incorporate real clinical pathways

Clinical care is not linear, and good algorithms reflect that. Be sure to include sub-pathways when they matter:

  • Biomarker-driven decisions

  • Surgical vs systemic vs radiation approaches

  • Pathology workflows

  • Supportive care considerations

This is especially important in complex cancers where cross-disciplinary specialties are involved.

Best practices for writing content

Titles and structure

Your title is critical. It should be:

  • Specific

  • Searchable

  • Clinically meaningful

Be sure to include:

  • Cancer type

  • Stage

  • Key biomarker, when relevant

This improves usability, as well as discoverability in ChatGPT, Claude, Gemini and other tools.

References: quality over quantity

References build trust and improve discoverability. Focus on:

  • NCCN, ASCO, ESMO

  • Landmark trials (NEJM, Lancet)

Add references where they matter most:

  • Key decision points

  • Pivotal treatment choices

You don't need references for every node. Clinical judgment still plays a role.

Visual design matters

Remember that clinicians think visually. The right images and video can:

  • Improve understanding

  • Speed up decision making

  • Support patient discussions

Useful visuals can include:

  • Survival curves

  • Treatment comparison tables

  • Risk stratification summaries

  • Imaging or pathology examples

Be sure to keep your images and video simple and clinically relevant.

How to structure your Living Algorithm

In medicine, clarity is everything. Each node should be:

  • Concise

  • Actionable

  • Unambiguous

For example:

  • "If EGFR mutation" → osimertinib

  • "If no actionable biomarker and PD-L1 ≥50%" → pembrolizumab monotherapy

Avoid vague language and be sure to specify:

  • Thresholds

  • Criteria

  • Exact conditions

This is what separates a poorly-designed algorithm from a valuable decision tool.

What you should avoid

Common pitfalls include:

  1. Overly broad scope

  2. Too much text per node

  3. Lack of clear decision points

  4. Missing thresholds or criteria

  5. Trying to include everything

Remember: an algorithm is not a textbook, it's a pathway.

From information to action

The goal is not to capture all knowledge, it's to:

  • Guide decisions

  • Reduce uncertainty

  • Save time

A well-designed algorithm should allow a clinician to:

  • Find their scenario quickly

  • Follow a clear path

  • Act with confidence

A real world example

Algorithms developed by leading clinicians demonstrate what's possible. For example, early-stage HER2+ breast cancer pathways clearly map:

  • Risk stratification

  • Treatment sequencing

  • Decision points

See this example from Dr. Ilana Schlam at Dana-Farber Cancer Institute in HER2+ Breast Cancer:



This is the level of clarity that makes algorithms extremely useful in practice.

Bottom line

Writing a good Living Algorithm is not about adding more detail, it's about making the decisions clear. Be sure to:

  • Be specific

  • Follow real workflows

  • Focus on action

  • Keep it concise

When done well, algorithms become more than references. They become tools that clinicians can rely on every day.

Want to explore more?

If you're interested in how treatment algorithms can help real-world decision making, explore Living Algorithms on Open Medicine.

You may find they change how quickly you move from uncertainty to clarity.

 
 

Open Medicine is where leading doctors post Living Algorithms to share their expertise. Instead of static diagrams in PDFs, Living Algorithms are mobile-first, interactive and updated instantly as new clinical evidence emerges.
 

We make expert medical knowledge easy to access so clinicians can offer the best treatment for their patients.

135 Commonwealth Drive

Menlo Park, California 94025 USA

     

    © 2025 Open Medicine

     

    bottom of page