Evidence
How we grade and verify clinical tools
HumanPulse does not display a single “accuracy percentage”. Evidence labels follow a transparent rubric, and every calculation is versioned and tested.
Evidence levels
Strong evidence
Derived and externally validated in multiple cohorts (including prospective studies) and incorporated into major clinical guidelines.
Moderate evidence
Externally validated, or an established physiological equation whose clinical thresholds rest on consensus; guideline support limited or mixed.
Limited evidence
Derivation and limited validation only, or significant known calibration problems.
Emerging evidence
Recently published; external validation still accumulating.
Formula verification
- Each calculator is implemented from the original publication or an authoritative guideline, never from memory.
- Each calculator ships with automated tests that reproduce published reference cases, boundary values and invalid inputs.
- Tools whose coefficients cannot be verified against an accessible primary source are catalogued as “pending verification” and are not calculable.
What we show for every tool
Validation population, outcome, time horizon, discrimination and calibration where reported, sensitivity/specificity for decision rules, guideline status, geographic and demographic applicability, limitations, references, version and review status.
Versioning
Every result records the calculator version. Any change to a formula, threshold or interpretation is recorded in the tool's version history with the effective date and reviewer.
Clinical review
Tools display whether their content has been independently reviewed by a clinician. Content marked “clinical review pending” has been compiled from the cited sources but awaits that review.