Medication, dose, negation and laterality checker
Catches wrong doses, flipped negations and left-versus-right slips.
Checked against clinicians
Recall on seeded and real error sets.
Where a person decides
Rules authored and signed off by customer pharmacy or clinicians.
Works on
Scribes, voice, chat, EHR agents.
Delivered inside a Quality Audit today. Reviewed .
The problem it solves
A few error types do outsized harm: wrong dose, flipped negation, left versus right. They're small edits that an average score barely registers.
How it works
Extract the sensitive items
Doses, negations, sides and drug names are pulled from source and output.
Compare them exactly
Rules check each pair, not just similar wording.
Block or flag
Mismatches are flagged for review, or blocked if a gate is configured.
An example, illustrative
Source: 'no chest pain'. Output: 'chest pain'.
Mismatch: negation flipped. Blocked until reviewed.
Source and output pairs, plus any local formulary or dose limits.
Rules catch known patterns. New kinds of errors still need clinician review.
What it helps measure
Quick answers
What does the dose and negation checker do?
Deterministic and targeted checks for high-harm error classes.
How is the dose and negation checker checked against clinicians?
Recall on seeded and real error sets.
Why use rules instead of a model?
For a few high-harm classes, exact rules give predictable results and are easy to explain to a safety committee.
Does this run in production?
It can, as a gate. In audits it measures how often these errors occur.
What can the dose and negation checker not do?
Rules authored and signed off by customer pharmacy or clinicians. No agent makes a clinical decision.