The Mini-Fellowship › Clinical tools › The Clinical Companion
The KRI Clinical Companion.
A ketamine practice has to answer a particular set of questions: whether to treat this patient, at what dose given who they are, what to watch for while it runs, what actually happened, and whether any of it worked. General-purpose software answers none of them and records the answers badly. The Companion is built for exactly those, and it comes with the course.

Outcome trends for one patient, plotted against the clinical threshold for each scale. The record shown is a sample used for teaching, not a patient and not results from this program.
What it is
Everything the method asks you to think about, in the room where you think about it.
The Companion began as a way to keep the reasoning behind an infusion in the room where the infusion happens, rather than on a worksheet from the night before. It has grown into the working surface for a ketamine practice: it scores the screen, weighs what the screen found, proposes a dose and the reasons for it, and then holds what actually happened — the labs, the infusion, the adverse events and the outcomes, in the order a patient moves through them.
It is given to participants, used during the clinical sessions, and stays with graduates through the mentorship year. Nothing in it is a substitute for the clinician. It models, it records, and it puts the right question in front of the person making the decision.

The starting point: demographics, the condition being treated, and the body-weight calculations every later number is built on. The target range is a course figure and is not published here.
One patient, start to finish
Step 1
Screen
The KIRT-S runs inside the app: hard-stop flags, a four-domain score, a risk level and a recommendation. Each screen is saved and dated, so a patient returning six months later is compared against their own earlier assessment rather than a memory of it.
Step 2
Correct what can be corrected
Laboratory and pharmacogenomic results are recorded against their reference ranges. Findings that warrant attention before a first infusion surface as action items at that point, rather than being noticed after a disappointing course.
Step 3
Plan
Actual, ideal and adjusted body weight are calculated, the protocol for the indication is applied, and the dose is adjusted for what the screening found. The adjustment and the reason for it are recorded at the time, not reconstructed later.
Step 4
At the chair
A modeled concentration curve shows where the patient is expected to be across the infusion, next to what the monitors are showing. The dose actually given is entered, and the curve is redrawn around what happened rather than what was planned.
Step 5
Record
The session note, the physiological response and any adverse event are captured while they are happening, on templates built around what a controlled-substance record has to contain.
Step 6
Follow the course
Outcome measures, adverse events, labs, notes and sessions accumulate into one timeline, so the next infusion is planned against what the previous ones actually did.
Before the first infusion
Most of the work happens before anyone is in the chair.
Four things decide whether a first infusion is a reasonable idea and what it should look like: whether the patient should be treated at all, what is wrong that can be put right first, what else they are taking, and how they metabolize it. The app asks all four, in that order, and carries the answers forward.
It starts with the reasons not to treat
The screen opens with hard-stop flags — uncontrolled substance use, uncontrolled hypertension, raised intracranial pressure or recent stroke, severe hepatic impairment, active mania or psychosis, unstable cardiovascular disease, pregnancy, known hypersensitivity. Any one of them defers treatment pending specialist review or a documented override.
The order is deliberate. Scoring averages by its nature: a serious concern in one domain is offset by reassuring findings in the others, and the total lands somewhere in the middle. For the findings where proceeding would be difficult to defend whatever the rest of the picture looks like, averaging is the wrong behavior — so those are taken out of the arithmetic altogether and asked first, as yes-or-no questions carrying their own consequence.
The override is there because medicine has exceptions. It is recorded as an override, with a name and a date against it, rather than disappearing into a total that reads as unremarkable afterwards.
Only then does it score. The result is a risk level and a recommendation, saved and dated, so the next screen for the same patient is a comparison rather than a fresh opinion.

The KIRT-S running inside the app: eight hard-stop flags, then the four-domain score, with the most recent screen and its recommendation carried at the top.
It separates what is fixed from what is fixable
MTHFR variants, low testosterone, low magnesium, carnitine deficiency, hypothyroidism and BDNF genotype each carry a threshold value and a recommended correction. Checking one does two things at once: it moves the candidacy score, and it puts a specific action on a list.
The list is short on purpose. It is not a wellness panel: each item is there because there is a plausible mechanism connecting it to ketamine response and something a clinician can actually do about it in the weeks before a first infusion. The correction is named rather than implied — a form and an amount, not “consider supplementation”.
The weighting follows from that. A finding that can be corrected costs less than one that cannot, so the score reflects what the patient will look like at the time of treatment rather than what they look like at the time of screening.
The point is timing. These are findings worth acting on before a first infusion rather than after a course that disappointed everyone — and a course that disappoints is expensive in a way that a three-week delay is not.

The co-morbidity assessment: each finding with its critical value, the corrective action, and the weight it carries in the candidacy score.
It reads the medication list for this drug
Benzodiazepines at two thresholds, amphetamines, anticonvulsants, antipsychotics and opioid antagonists at two thresholds — each shown with the threshold that matters and the effect expected on the patient’s response, not a generic interaction warning.
Two thresholds rather than one, because these effects are dose-dependent and not categorical. Below ten milligram-equivalents of diazepam the expectation is a delayed response; above it, that the patient may not respond at all and the benzodiazepine dose is worth revisiting before the infusion is. A yes-or-no interaction checker says “interacts” to both and leaves the clinician no better off than before it was consulted.
What comes back is a specific expectation rather than a warning: a delayed response, a short-lived one, a reduced one, or a recommendation to reconsider treatment altogether. Those are different clinical decisions, and lamotrigine sits in its own row because what it returns is reconsider rather than adjust.
The practical value is in the conversation before the first infusion. A patient told in advance that their medication list makes a late response likely is a patient who is still in the chair at session four. A patient not told that concludes at session three that it does not work.

The medication review, read for effect on ketamine response rather than as a general contraindication list.
It accounts for how the patient clears it
Strong and moderate CYP3A4 inducers and inhibitors are entered by how many the patient is taking, and the app returns a per-medication and a total adjustment. Ultra-rapid and ultra-slow metabolizer status can be entered where a genotype is available.
They are counted rather than named because the effect aggregates. One strong inducer is a manageable adjustment; three is a different patient, and a list that only records which drugs are present does not make that visible.
Genotype is an input here, not a determinant. The human evidence does not support setting a ketamine dose from a CYP2B6 genotype — the one in-vivo study was negative — and the app is not built to do it. Where a result exists it is one factor among several, weighted accordingly.
What comes out is an adjusted dose with the adjustment shown as a percentage and every contributing reason listed beneath it, so the number can be argued with rather than simply accepted. A recommendation whose derivation is hidden is one a clinician has to either trust or ignore.

CYP450 inducers and inhibitors entered by count, with genetic variants, and the adjusted dose that results.
And it ends in something you can act on
The four assessments aggregate into one candidacy score with a named tier — excellent, fair or poor — each carrying its own recommendation and the rationale behind it, alongside a six-step workflow and a list of required actions before treatment.
The required actions are the part that does the work. They are numbered and specific, so they can be cleared one at a time and seen to have been cleared, rather than sitting in a note as advice that everyone agreed with and nobody actioned.
The consolidated risk-factor table beneath shows every contributing finding with the weight it carried. The score is therefore auditable backwards: a clinician who disagrees with the tier can see exactly which entries produced it and change the entry rather than override the conclusion.
It is a structured summary of what was found, not a verdict. The clinician decides whether to treat; the app makes sure the decision is made with the ketamine-specific factors actually in view.

The candidacy summary: score and tier, the consolidated risk factors, and the actions the clinician is expected to complete before treating.
During the infusion
What to watch, and what to do when it moves.
The thresholds are on the screen, not in a binder
Blood pressure, heart rate, oxygen saturation and respiratory rate, each with its frequency, its normal range and the action required when it moves: blood pressure every five minutes with stop-or-reduce above 180 over 110, heart rate and saturation continuous with supplemental oxygen below ninety percent, respiratory rate every five minutes with support below ten or above thirty.
The action column is the reason it is worth having on screen. A normal range tells you a number is abnormal; it does not tell you what to do, and the moment a reading moves is the worst moment to be working that out from first principles.
Alongside it, the neurological assessment: dissociation and sedation scored on the Ramsay or MOAAS scale rather than described, emergence phenomena noted, nystagmus flagged as common and usually benign so it is not mistaken for a complication.
Then the criteria for letting someone leave — thirty to sixty minutes minimum, orientation and coordination assessed, recovery scored, discharge on return to baseline with a responsible escort — and the rating scales with the interval each is repeated at: the depression scale seven to fourteen days after induction, the anxiety and response scales every two weeks, the suicidality scale at every session.

Monitoring parameters with the action threshold beside each one, the neurological assessment, the rating scales and their schedule.
The equipment list is part of the record
Continual blood-pressure monitoring, pulse oximetry, ECG, oxygen and its delivery, suction and IV access; then the emergency drug box — benzodiazepines, antihypertensives, atropine, epinephrine, naloxone, flumazenil and saline — which has to be in the room and in date rather than somewhere in the building.
Then the four adverse events worth rehearsing, each with a prevalence and a management strategy: hypertensive crisis and respiratory depression, both reported at under one percent; transition reactions and dissociative symptoms, both an order of magnitude more common than that. Knowing which of the four you are likely to meet is what stops a self-limiting reaction being managed as an emergency.
Emergency numbers sit below them — emergency services, poison control, and the current psychiatric crisis line.
The facility and staffing requirements close it out: private rooms with monitoring capability, a recovery area, accessible emergency equipment, and one-to-one nursing during treatment. It is the same standard taught on the course, held in the same place as the patient it applies to, which is where it is actually useful.

Required equipment and emergency medications, adverse-event management with prevalence, emergency contacts and facility requirements.
Every session leaves a record behind
Each infusion is logged with its date, indication, dose, duration and any adverse events, numbered in sequence so the course is legible at a glance rather than reassembled from notes.
The dose actually given is what gets recorded — not the dose that was recommended — because the next session is planned against what happened.

The session list for one patient: date, indication, dose given, duration, and whether an adverse event was recorded. Modeled peak concentrations are course figures and have been removed.
Afterwards, and for the next twelve months
Whether it worked is a measurement, not an impression.
Ketamine practice is unusually vulnerable to a pleasant conversation at the end of a session standing in for data. The app scores the patient on validated instruments, classifies the result against published thresholds, and draws the course.
Scored, classified and dated
Four instruments — depression, anxiety, post-traumatic stress and a pain scale — so a chronic-pain patient and a PTSD patient are each measured on the instrument built for them rather than on whichever one the practice happens to use. Questionnaires print, so the patient can complete them on paper.
Each assessment is kept with its date and its severity band, which is what makes the history auditable rather than anecdotal. A score of nine means nothing on its own; nine against a baseline of twenty-one, on a stated date, means something.
The response classification then compares the most recent score against the earliest and says which threshold it met — and names the number it used to decide.
That last part matters more than it sounds. Response and remission are different things — by the usual convention a response is a substantial fall from baseline, while remission means the score has reached or passed below a stated threshold — and a patient can meet one without the other. Software that reports the label without the arithmetic invites the two to be used interchangeably, which is how a practice comes to believe it is doing better than it is.

PHQ-9, GAD-7, PCL-5 and VAS with the full assessment history, and a response classification that states the threshold it used. The record shown is a sample used for teaching, not a patient and not results from this program.
And drawn, so a relapse is visible
The scores plot over time against the clinical threshold for each scale, with the severity bands named in the legend so a point on the line is read against what that score means rather than against the other points.
A course that improved and then drifted back looks different from one that held, and it looks different months before anyone would have said so out loud. Drift is the failure mode a conversation at the end of a session is worst at detecting, because each session is compared with the last one rather than with the best one.
The summary cards carry the change from baseline in both points and percent, and the number of assessments behind it — which is the figure that stops two data points being mistaken for a trend.
This is also what turns maintenance into a decision rather than a habit. A responder who has lost a quarter of their peak improvement appears on the practice dashboard without anyone going looking, and the graph is where that shows up as something you can see rather than something you were told.

The same data plotted over time, with the clinical threshold marked and the severity bands in the legend. The record shown is a sample used for teaching, not a patient and not results from this program.
Adverse events are coded, not described
Every event is coded for severity, relationship to treatment, action taken and outcome, per patient and cumulatively across the practice, exportable as CSV or PDF.
The reporting rule is printed above the log rather than left to memory: an event ending in transport to a higher level of care, or an emergency-department visit within twenty-four hours of an infusion, is reportable to the state within twenty-four hours.

The coded adverse-event log, with the reporting rule stated on screen and each event coded for severity, relationship, action taken and outcome.
Across a practice
The things that go wrong quietly are the ones nobody is looking at.
The responder who slipped back four months ago. The adverse event that should have been reported. The patient who never came back. None of these announce themselves, and none of them are visible from inside a single chart.

The practice view: active patients, recent infusions, open serious adverse events, who is overdue, and who is losing benefit — without opening a chart. The record shown is a sample used for teaching, not a patient and not results from this program.
A view across every patient
Active patients, infusions in the last thirty days, open serious adverse events and who is overdue for follow-up, on one screen.
Maintenance that surfaces itself
A responder who has lost a quarter of their peak improvement appears on that screen. Nobody has to remember to look.
Reportable events, flagged as such
The cumulative log carries the reporting rule with it, so the twenty-four-hour obligation is in front of the clinician rather than in a regulation.
It checks its own arithmetic
The optimization logic ships with a validation suite of fifteen patient scenarios that can be run on demand. Software that carries clinical reasoning should be testable, and this is testable.
What it covers
The screening instrument
KIRT-S assessments, saved, dated and reloadable, with the risk level and the recommendation that followed.
Labs and pharmacogenomics
Numeric and genetic results with reference ranges, flagged when abnormal, and carried into the clinical picture rather than filed away from it.
Interaction analysis
A ketamine-specific read of the medication list, reported as the effect expected on this patient’s response — not a generic contraindication list.
Protocols by indication
Separate induction, frequency, escalation and ceiling parameters for depression and bipolar, anxiety and PTSD, chronic pain, and maintenance.
A coded adverse-event log
Each event coded for severity, relationship, action taken and outcome, per patient and across the practice, exportable as CSV or PDF.
Outcome measures
PHQ-9, GAD-7, PCL-5 and VAS, scored on entry, classified against response and remission thresholds, and charted over time.
Notes and templates
Initial evaluation, infusion note and infusion record templates, editable for your own practice, with documents and images attached to the patient.
The clinical timeline
Notes, sessions, labs, outcomes, adverse events and screens in one chronological stream, filterable and printable.
Screenshots are from the working application. Target concentrations have been removed, because the specific target ranges are course material and are not published here. Records shown are teaching samples.
What the Companion is not
Six limits, stated up front.
It is not a dosing autopilot
The app models and records. The clinician decides. Every number it produces is a proposal a trained person is expected to evaluate, and the training is what makes that possible.
It is not sold separately
The Clinical Companion comes with the Mini-Fellowship and is not currently available outside it. That is deliberate: the modeling is of use to someone who understands the pharmacology underneath it, and the training is where that comes from.
It does not measure plasma concentration
No bedside assay exists. The curve is a pharmacokinetic model of the expected concentration, not a measurement of the patient’s blood. The course is explicit about this distinction and so is the app.
It does not replace monitoring
The model tells you where the patient should be. The monitors and the patient tell you where they are. When those disagree, the patient wins.
It is not a diagnostic device
It is not FDA-cleared, and it is not offered as a device. It is a decision-support and record-keeping tool for clinicians who have been trained in the method it supports.
It is not your EHR
Not as it stands. It is built for the clinician giving the infusions, and it exports to CSV and PDF so what it holds can go into the chart you already keep.
What the evidence supports on individualizing a dose →
How the concentration is targeted →
The screening tool it starts with →
Questions
About the Companion.
Does the Companion measure plasma levels?
No. There is no point-of-care assay for ketamine. The app models the expected concentration from a pharmacokinetic model, using the patient’s adjusted body weight, the protocol for the indication and the metabolic factors entered. It is a plan, checked against the patient, not a reading.
Are the target concentrations shown on this page?
No. The specific target ranges are course material and are not published here, so they have been removed from the screenshots. The labels remain so you can see where they sit.
Does the app require genetic testing?
No. Pharmacogenomic results can be entered where a patient has them, and the course teaches how to read one. The human evidence does not support setting a dose from genotype, and the app is not used that way.
Can I buy the Companion without taking the course?
Not at present. It comes with the Mini-Fellowship and stays with graduates afterward.
Does it run on a phone or a tablet?
Yes. It installs on iOS, Android and desktop as a progressive web app, so it can be open on a tablet at the chair and on a desktop in the office.
Can I get my data out?
Yes. Patient records, the adverse-event log and outcome measures export to CSV and PDF at any time.
Is it a replacement for my EHR?
Not as it stands. It covers screening, planning, the infusion record, adverse events and outcome measures — the parts of a ketamine practice a general EHR handles poorly — and exports so the rest can live where it already lives.
What happens if I stop practicing the method?
The app and the materials are tied to the training and to the twelve-month re-assessment. Graduates who stop practicing the method, or who do not complete the re-assessment, are asked to retire the credential.
Where does the patient data live?
In your account, under your control, and you can export it at any time. Graduates remain responsible for their own privacy and security obligations, which the course covers.
Ketamine is FDA-approved as an anesthetic. Its use for depression, chronic pain and other psychiatric indications is off-label, which means the FDA has not evaluated it for those uses. The KRI Clinical Companion is not an FDA-cleared medical device and is not a diagnostic instrument. It is a decision-support and record-keeping tool for clinicians trained in the method it supports, and clinical decisions remain the responsibility of the treating clinician.
It comes home with you
Along with the KIRT-S, the Clinical Reference Manual, evaluation and consent templates, infusion records, controlled-substance logs and standard operating procedures.
