Loading scenario…

AKI

μg/mL vs. hr

Population

Individual

Selected

Observations

AKI Periods

Creatinine (SCr)

Waiting for Data

Complete demographics and events to generate a simulation.

Simulation Paused

Update the highlighted field to resume.

Events

Doses, measured levels, and serum creatinine. Add or edit rows and the simulation reruns as you go.

Steady-state mode: the first dose's interval sets the assumed starting concentration, so it's required.

Date	Time		Value	Dur Duration	Int Interval
				hr BLQ —	hr —

Regimens

Candidate regimens scored against your targets. Click a row to plot it on the chart, or enter your own in the bottom row.

	Amount	Dur Duration	Int Interval	mg/kg	AUC_SS	Trough_SS	Target
No candidate regimens yet — they appear once the simulation has run.
	mg	hr	hr	mg/kg	AUC/MIC (24h) Predicted ± On target Time to target Model disagreement	Trough Predicted ± mg/L Toxicity Efficacy Model disagreement	✓ —
	mg	hr	hr	mg/kg	AUC/MIC (24h) Predicted ± On target Time to target Model disagreement	Trough Predicted ± mg/L Toxicity Efficacy Model disagreement	✓ —
	mg	hr	hr	mg/kg	AUC/MIC (24h) Predicted ± On target Time to target Model disagreement	Trough Predicted ± mg/L Toxicity Efficacy Model disagreement	✓ —

Regimens start

Definitions

On target: — the probability that the regimen's steady-state AUC/MIC lands inside your target window, given the prediction's uncertainty.
Time to target: — how long after the regimen starts before exposure first reaches the target. "N/A" means it never gets there over the simulated horizon.
Model disagreement: — the share of a prediction's uncertainty that comes from the models disagreeing, rather than from each model's residual (assay) error.
Toxicity: — the probability that the steady-state trough exceeds the concentration associated with harm.
Efficacy: — the probability that the steady-state trough sits at or above the concentration needed to be effective.

No simulation has run yet. Add doses and demographics to trigger one.

Ensemble Fit

Population (Base)

RMSD μg/mL

Bias μg/mL

Individual (MAP)

RMSD μg/mL

Bias μg/mL

Eff. Models of

Weight Entropy 0–1

#	Time	Measured	Pop Pred	Ind Pred	Weight
	hr	mg/L	mg/L	mg/L	0–1

	Population		Individual (MAP)

%	mg/L	mg/L	mg/L	mg/L

Definitions

Population (Base): — the model's prediction from its published average parameters, before fitting to the measured levels.
Individual (MAP): — the same model after Bayesian fitting to the measured levels.
RMSD (Fit): — typical gap between the predicted and measured levels (µg/mL). Lower is a tighter fit; big misses count more.
Bias: — which way the misses lean: positive predicts high, negative predicts low, near 0 means no consistent lean.
Averaging weight: — how much each model counts toward the blended (ensemble) prediction.
Eff. models: — roughly how many models meaningfully contribute to the blend (1 ÷ Σ weight²). Near 1 means one model dominates; higher means weight is spread.
Weight entropy: — spread of the averaging weights, scaled 0–1. 0 means one model carries everything; 1 means all contributing models weigh equally.
Model disagreement: — the share of a prediction's uncertainty that comes from the models disagreeing, rather than from each model's residual (assay) error.

Full definitions & equations →

No simulation has run yet. Add doses and demographics to trigger one.

Targets

	In Target	Below	Above	Time to target

Selected

Not enough history yet — needs at least one completed dosing interval.

Per-model kinetics

Pop Kinetics					Ind Kinetics					CrCl
CL	V1	V2	Q2	V3	CL	V1	V2	Q2	V3	CrCl
L/h	L	L	L/h	L	L/h	L	L	L/h	L	mL/min

Bayesian Model Averaging

Outside its training range.

Use all models, or manually select a subset. Predictions are averaged, weighted by fit — best once levels are available.

Your model set

Filter

/ sel

	Model	N	Info
	running —		Notes
No models match.

Optimization Candidates

Select the doses and intervals to simulate when finding a target regimen.

Dose (mg)

Interval (h)

Chorus

Chorus is a vancomycin pharmacokinetic simulator that uses Bayesian model averaging to fit 46 published models to a scenario at once.

Not a validated clinical tool, and not a source of dosing advice.

vanc.app

Chorus analysis view: the model-averaged fit curve through measured vancomycin levels, beside a per-model table listing each published model's averaging weight and fit error

Features

Client-Side & Zero-Footprint:: The simulation runs entirely in your browser, with nothing transmitted.
Longitudinal Renal Tracking:: Incorporates dynamic clearance changes across serial creatinine values, automatically flagging KDIGO-defined acute kidney injury.
Configurable BLQ Handling:: Enter below-quantification values directly, with customizable imputation rules and fitting weights.
Transparent Model Diagnostics:: Review each model’s averaging weight, residual error, and bias in an open diagnostic table.
Customizable Ensembles:: Filter the library by clinical cohort, select individual models, and choose between equal, error-based, or penalty-adjusted weighting.
Cohort Analysis & Model Selection:: Compare model performance across many scenarios, and train a selector that narrows the ensemble before the first level.
Regimen Optimization:: Simulates candidate dose-and-interval combinations, scoring each by its ensemble probability of achieving target steady-state AUC₂₄.

Events

Regimens

Observations

Per-model

Per-model kinetics

Optimization Candidates

Dose (mg)

Interval (h)

Chorus

Features

Simulation settings

Models

Dose candidates (mg)

Interval candidates (h)

Demographics

Copy scenario

	Wt (kg)	CrCl
Actual
Ideal
Adjusted
Lean