Finding out whether a question is answerable takes as long as answering it.
Each request starts by characterizing the same sources again: which criteria a source can assess, what it does not capture, and where a proxy would be needed.
Real-world evidence & epidemiology
Clear the request queue without adding headcount.
Clinical development, medical affairs, HEOR and safety all need answers from the same small team. StromaOS gives that team specialist AI agents that take a research question through feasibility, cohort creation and analysis under your methods - with every population approved by a named subject matter expert (SME) and reused rather than rebuilt.
Elastic analytical capacity
Governed cohort definitions
Approved population packages
Traceable membership decisions
The real-world evidence operating problem
Every function wants a population. One small team defines all of them.
An RWE and epidemiology team is an internal service provider. Requests arrive from clinical development, medical affairs, HEOR and safety faster than the team can scope them, and much of the effort goes into work that has already been done once.
The same population is defined again for every study.
Entry rules, time windows, code lists and exclusions live in scripts and documents that the next analyst cannot safely reuse, so the work starts over.
Small changes move people in and out without anyone seeing it.
A new code set, a shifted baseline window or a data refresh changes who is in the cohort, and the change surfaces only when two results disagree.
Other functions rebuild what your team already approved.
HEOR, safety and medical writing receive a table, not the definition and approval behind it, so each team reconstructs the population in its own way.
How the solution works
A research question becomes an approved population in one governed chain.
Specialist AI agents perform feasibility, cohort construction and analysis against your sources and your methods. A Sentry Agent checks each step against the configured rules before it proceeds, and every decision that carries scientific weight goes to a named person.
STEP 01
Frame the question
Population, exposure, comparator, outcome, follow-up and intended use, drafted from a protocol excerpt or written directly in plain language.
STEP 02
Assess feasibility
Which criteria each source can assess, counts and attrition at every criterion, and where evidence is missing or needs review.
STEP 03
Define the cohort
Entry event, observation requirements, inclusion and exclusion rules, time windows, exit and censoring, with concept sets pinned to a vocabulary release.
STEP 04
Compare scenarios
Run alternative definitions side by side and see exactly which people moved, and at which rule, before a choice is made.
STEP 05
Approve for use
A named SME reviews the definition, the run and the data it was built on, and approves the cohort for a stated intended use.
STEP 06
Reuse and refresh
The approved cohort package flows to analysis and to other functions by reference. A data refresh produces a new package for review, never a silent change.
Cohort control
Every person in a cohort can be explained.
StromaOS records, for every person, which criteria were met, which were not, and which the data could not settle, together with the source record that decided each one. An earliest recorded event is labeled as exactly that, and missing evidence is never treated as absence.
When a baseline window moves from six months to twelve, your team sees who left the cohort and at which rule, before the change reaches a result.
Model a cohort workflowOne approved population, many functions.
An approved cohort package is published to the teams that need it. Each receives the same versioned population by reference, with its definition and approval attached, so the population a payer, a safety committee or a reviewer asks about is the one your team approved.
HEOR & market access
Utilization, cost and outcomes analyses run on the population the evidence team approved.
Safety & pharmacovigilance
Exposure cohorts and observation windows for safety studies, defined once and traceable to the person.
Medical writing
Methods text and population descriptions that match the approved definition word for word.
Clinical development & operations
Feasibility counts and attrition for protocol planning before money is committed to a design.
Governance throughout real-world evidence
AI accelerates the work without inheriting scientific authority.
StromaOS separates drafting, feasibility, cohort construction, analysis, review, approval and publication. Your epidemiologists and SMEs retain every scientific decision.
Explore the Governance Spine01
Method governance
Definitions, concept sets, vocabulary releases, time windows, diagnostics and reviewer acceptance remain part of the record.
02
Approval authority
Only a designated SME can approve a cohort, and never the person who authored it. Approval names the data source and the intended use, and does not extend beyond them.
03
Exploratory and confirmatory, kept apart
Working runs are labeled exploratory wherever they appear. Every version, scenario and comparison is kept, so a population chosen after seeing outcomes cannot pass as pre-specified.
04
Privacy by design
Aggregate users see counts with small numbers hidden, and person-level evidence requires patient-level access. In Your VPC, Embedded and Air-gapped deployments, patient-level data and execution remain inside your boundary.
05
Evidence Data Trace
Definition, run, data identity, approval, refresh and every downstream use remain reconstructable.
Start with one research question, feasibility request, or cohort your team keeps rebuilding.
Connect the sources you already license. Define the methods, review and approval authority. Then run the queue as one governed real-world evidence operation.