Evidence-Based Medicine · EBM Tool

Cardiology
Trial Match

Browser-based educational tool that matches de-identified patient profiles against 39 landmark cardiology trials across heart failure, atrial fibrillation, and valvular disease. Understand how well your patient resembles the population that produced the evidence you're applying.

Trial library · v1.0

39 trials · 3 topics · 8 groups

How it works

Patient input model

The patient form accepts only de-identified clinical variables: NYHA class, LVEF, NT-proBNP, eGFR, systolic BP, heart rate, QRS duration, electrolytes, rhythm, comorbidities, recent hospitalizations, and current GDMT.

Any field left blank is treated as unknown: not as failing. This means a partial record still produces useful results; it just reduces certainty.

Criterion evaluators

Each trial criterion is a structured object referencing a named evaluator function. Evaluators return one of three values: met, not_met, or unknown (missing input).

Reusable evaluators handle common checks (ageGte, lvefLte, egfrLt, sbpLt, onMed, etc.); trial-specific evaluators handle composite biomarker rules like the PARADIGM-HF NT-proBNP threshold logic.

Eligible

All criteria met

Every inclusion criterion met; no exclusion triggered; no unknowns.

Partial match

Some criteria met

Some inclusion criteria met or unknown; no exclusion triggered.

Excluded

Exclusion triggered

At least one exclusion criterion met, or every inclusion criterion explicitly not met.

Insufficient data

Too many unknowns

Too many criteria are unknown to determine eligibility. Enter more variables.

NNT caveat: When a trial is anything other than 100% eligible, the trial card notes that “NNT increases when trial results are applied to lower-risk populations.” The published NNT applies to the trial population; patients who differ, especially toward lower baseline risk, typically have higher NNTs in practice.

Trial library

39 trials · 3 topics · 8 groups

  • HF

    Heart failure: 17 trials, 4 groups

    HFrEF foundational therapy (CONSENSUS, MERIT-HF, RALES, CHARM…), SGLT2i / novel agents (EMPEROR-Reduced, DAPA-HF…), HFpEF (EMPEROR-Preserved, DELIVER…), and device therapy (MADIT-II, SCD-HeFT, RAFT…).

  • AF

    Atrial fibrillation: 13 trials, 3 groups

    Anticoagulation (RE-LY, ROCKET-AF, ARISTOTLE, ENGAGE AF…), rhythm control (EAST-AFNET 4, CABANA, AFFIRM…), and rate control trials.

  • VHD

    Valvular disease: 9 trials, 2 groups

    TAVR landmark trials (PARTNER 3, Evolut Low Risk, NOTION, SURTAVI…) and mitral intervention trials. Coronary artery disease topic scaffolded for future expansion.

Visualizations

Match-strength bar chart at the top of results: all trials sorted by best fit, color-coded by status. Click a bar to jump to that trial’s card.

Per-trial radar chart: each axis is one inclusion or exclusion criterion. Fills the outer ring when met, drops to center when not met, sits at the midpoint for unknowns.

Race composition bar: minimalist stacked bar for each trial’s enrolled population. Trials where the original publication did not report race show a striped placeholder with explanatory text.

Status filter chips: the four status counts at the top of the results panel act as toggles.

Privacy & HIPAA

Designed outside HIPAA scope

The tool is designed to operate outside the scope of HIPAA by never accepting, storing, or transmitting protected health information.

No PHI fields: the form accepts only de-identified clinical variables (NYHA class, LVEF, NT-proBNP, eGFR, medications, comorbidities). No name, MRN, date of birth, address, or any of the 18 HIPAA identifiers.

No backend: the site is a static page hosted on GitHub Pages. There is no server that receives, processes, or logs patient inputs. All evaluation runs in the browser.

No persistence: patient inputs live only in browser memory for the session. Closing the tab discards everything.

For institutional use

If your hospital, residency program, or health system is considering embedding this tool in a clinical-education workflow, the above design means there is no PHI handling that requires a Business Associate Agreement.

However, your institution’s privacy or compliance office may still want to review the tool independently. The complete source code, deployment configuration, and analytics provider (GoatCounter: no cookies, no IP storage) are all open and inspectable.

Important: This tool is for education, not for clinical decision support in named patients. When in doubt, use hypothetical or de-identified examples.