Every visit, prepped.
Every suggestion, proven.

The intent is the opposite: the pre-visit summary collects what a clinician would otherwise go hunting for across prior notes, labs and claims, and puts it on one screen. During the visit, suggestions are short, capped per encounter, and dismissible in a single action. We would rather you measured it than took our word for it — chart-prep minutes before and after, on your own panel, is the number to hold us to.

As many as you allow, and no more. The cap per encounter is a setting you control, as is the confidence threshold a suspect has to clear before it surfaces at all. Raising the bar shows fewer, stronger suggestions. This is the single most important dial in the system: a tool that interrupts a clinician too often stops being read, and after that its accuracy is irrelevant.

They dismiss it, and that is the end of it for that encounter. Dismissals are recorded — not to challenge the clinician, but because the pattern of what gets dismissed is the clearest signal of where the suspecting logic is wrong. Those rates come back to you by condition and by signal type, so thresholds can be tightened where the tool is being noisy.

It is built not to. A suspect surfaces as a documentation cue with the evidence behind it attached — the note, the lab, the medication it came from — so the clinician is being shown why something is worth considering, not handed a code to accept. The clinical decision is the clinician's, and a condition that is never addressed at the encounter never becomes a submitted diagnosis.

You do. The signal types are published rather than hidden behind a score, so your clinical and coding leadership can review the logic and decide which sources to trust and at what threshold. Conditions, signal types and confidence tiers can each be turned up, turned down, or switched off entirely for your organization.

Suggestions are delivered through the patterns your EHR already supports — Epic Best Practice Advisories and FHIR CDS Hooks — so clinicians see them where they already look, rather than in another tab. The work is mostly on the data feeds and the tuning, not on building a new interface for your physicians to learn.

It runs inside your own environment — on-premises, in your private cloud, or air-gapped. No PHI leaves your network and there are no external API calls in the data path, so your existing security controls, IAM and monitoring stay in effect. Updates ship as versioned releases your team applies on its own schedule.

Suggestion acceptance rate is the number that matters most, tracked by condition, by signal type and by clinician, because it tells you whether the people using it trust it. Alongside it: alerts per encounter, dismissal patterns, and how much of the panel had its conditions addressed during a visit rather than chased afterwards. If acceptance is falling, the logic needs tightening — and you will see that before it becomes a coding problem.