Overview
Budget Impact takes a research question and produces a fully sourced, editable budget-impact model — without the weeks of manual literature review and spreadsheet building that usually sit behind one. Everything the model uses is visible: the decision tree, every parameter, and the citation behind each number.
The workflow
The analysis moves through seven screens. You can review and edit at every step.
1.Research Setup
Describe the question in plain language (or upload a paper). EviValue parses it into a PICO structure — Population, Intervention, Comparator, Outcomes — and infers the perspective and time horizon, then matches it to a supported model.
2.Study Design & Protocol
Review and edit the AI-extracted study protocol before anything runs. Every field is editable, so you stay in control of the inputs that drive the model.
3.Model Builder
Pick the evidence source and review the model structure EviValue builds — a transparent decision tree of the clinical pathway. If the question doesn't match a supported model, EviValue says so rather than guessing. Approve the tree to move on to the parameters.
4.Model Parameters
Review the editable table of every input in the model. Each value shows its source citation, and where the evidence offers alternatives you can switch between candidate values. Any edit recomputes the model live. (Model Builder and Model Parameters are two screens for one model-building step — structure first, then the numbers — so you can focus on one at a time.)
5.Results
See the total cost of care, per-member-per-month impact, budget impact, and a year-by-year breakdown, with an auto-written interpretation of the numbers.
6.Sensitivity Analysis
Run one-way sensitivity analysis with literature-anchored ranges and a tornado diagram that ranks the parameters driving the result.
7.Report
Generate a payer-ready report following budget-impact good-practice structure, ready to export and share.
Evidence engine
The built-in EviValue Literature Extract searches across medical research databases in parallel, then reads, deduplicates, and ranks the studies before pulling out the parameters your model needs.
- check_circleSearches PubMed and Semantic Scholar in parallel
- check_circleDeduplicates and ranks studies by relevance
- check_circleExtracts parameters with the source citation attached to each value
- check_circleFlags single-source values so you know what to double-check
Validation
Before results are shown, EviValue runs a panel of deterministic checks on the model inputs — no AI, just math. The checks fall into four groups:
Data integrity
Required inputs present, probabilities in [0,1], costs non-negative and finite.
Clinical plausibility
Every clinical driver checked against its literature-anchored range.
Logical consistency
Directional invariants between arms hold (e.g. the comparator's sensitivity isn't lower than expected).
Structural
Tree branch probabilities sum to one per arm; population and time horizon are sane.
FAQ
Where do the numbers come from?
Every parameter is shown with the source it came from, and you can edit any value — the model recomputes immediately.
Can I change the inputs?
Yes. Every protocol field and every model parameter is editable. The decision tree and results update live as you edit.
Is the model a black box?
No. The decision tree shows every node and probability, each cost traces to its source, and a validation panel runs deterministic checks on the inputs before results are shown.
What evidence sources are supported?
The built-in EviValue Literature Extract searches PubMed and Semantic Scholar. Connectors for external evidence platforms are on the roadmap.
Contact
Questions or feedback?
Reach the team and we'll get back to you.