The comparison is explicit
Output is framed as departure from this patient’s established state, not as a percentile. The clinician can see what the comparison set is, which is not true of most scores in routine use.
Clinicians & health systems
The reference range is the patient.
Built by a bedside neuroscience nurse, for people who have watched a patient deteriorate inside a normal range. Here is what changes, what the regulatory status actually is, and what we will not claim.
Regulatory status, before anything else
Pathos Vigil is investigational. It has not been cleared or approved by the U.S. Food and Drug Administration or any other regulatory authority. It is not available for sale and not for clinical use. Nothing on this page is a claim of clinical performance, safety or efficacy.
We put this first rather than in a footnote because clinicians are entitled to know the stage of a thing before they read a word about what it might do.
The clinical problem
Deterioration inside a normal range.
You have seen this. A patient whose observations sit inside every threshold on the chart, whose trend is unmistakably wrong, and whose deterioration is obvious to the nurse who has been with them for six hours and invisible to the scoring system that has been with them the whole time.
The scoring system is not broken. It is answering the question it was built to answer — is this observation abnormal for people in general — when the question that matters is whether it is abnormal for this patient, now, compared to this patient, earlier.
The gap between those two questions is time. It is the interval in which a change is real, is measurable, and is not yet visible to any instrument comparing the patient to a population.
The asymmetry
The direction that gets missed.
Level of consciousness is the clearest example. Agitation is documented, escalated and responded to. Its opposite — the patient becoming quieter, slower, less engaged, less present — generates no disruption, and in a system whose attention is allocated by disruption, that is a hazard.
The populations who experience the quiet failure mode are not random. They are the sedated, the delirious-hypoactive, the non-verbal, the cognitively impaired, the patients whose first language is not the ward’s, and the patients who have learned that speaking up does not improve their care.
Symmetric treatment is not a philosophical position. It is the correction for a directional gap that is well described in the literature on hypoactive delirium and unrecognized deterioration — see the hypoactive-delirium citation, a gap you can go and check rather than take from us.
At the bedside
What a per-subject baseline changes in practice.
Trend is the primary object
Regime change — is this system moving? — rather than threshold crossing. Continuous, rather than reassessed when someone next has a hand free.
Direction is preserved, not ranked
Elevated and suppressed are reported distinctly, because the clinical response differs — but neither is weighted as more severe by the model.
Absence of baseline is stated
A newly admitted patient has no baseline yet. The system says so, rather than silently substituting a population range and presenting the result with the same confidence.
What it is not
Not a replacement for clinical assessment, and not designed to be. Not an autonomous alarm system. Not a diagnosis. The intended role is to make a change visible earlier to the person who is going to make the decision — who remains the clinician.
The patient who stops making noise is the one to look at hardest.
Attention in a busy unit is allocated by disruption. Deterioration is not obliged to be disruptive.
Evidence
Where the evidence stands.
The underlying mathematics is not ours and is not new. Critical slowing down as an early-warning signature of state transition in complex dynamical systems is an established body of work spanning ecology, climate science and physiology — we set out the published lineage on the evidence page. What we are doing is applying it under a per-subject baseline and a symmetry constraint, in monitoring contexts.
The application is what requires evidence, and that evidence is what a regulatory pathway exists to establish. We are in development. We have not published clinical performance data, and we will not describe performance we have not demonstrated.
If you would find this useful, the most valuable thing you can offer us is not enthusiasm — it is a hard question about where it would fail in your unit.
What we will claim
- The framework and how it is constructed.
- The mathematical basis and its published lineage.
- The architectural guarantees, which are inspectable.
What we will not claim
- Clinical performance, sensitivity or specificity.
- Improved outcomes, length of stay, or mortality.
- Regulatory status we do not have.
- Comparative superiority over any existing tool.
Safety, quality and data
How the regulated part is kept separate.
Collaboration
What we are looking for from clinical partners.
-
Clinical input on failure modes
Specifically: where would a per-subject baseline mislead you? Which patients arrive without a usable baseline? Where is the quiet deterioration in your unit?
-
Research collaboration
Retrospective analysis, prospective observational study design, and academic partnership — the work that has to happen before any performance claim is legitimate.
-
Design partnership for workflow
A monitoring output nobody has time to look at is worthless. We want to be told, early and bluntly, what would not survive contact with a real shift.