RWE Architect™From the research question to a submission-ready manuscript
A conversational platform that conducts real-world studies using Brazil's Unified Health System data — protocol, cohort, statistics, figures, and manuscript — with scientific rigor automatically verified at every stage.
4B+
outpatient care records
224M
hospital admission authorizations
246M
patients in the cross-database linkage catalog
STROBE + RECORD
reporting aligned with journal standards
Brazil's public health data exists. The evidence rarely does.
Brazil's Unified Health System generates one of the world's largest health data collections. Turning it into publishable evidence still requires data engineering, biostatistics, and months of work — and an error in cohort definition becomes a plausible but incorrect number inside a paper.
Data that is difficult to access
Huge databases, different coding across systems, and patient linkage between databases.
Expensive, sequential teams
Data engineer, statistician, and writer working in sequence.
Silent errors
Ordinary AI writes convincing text even when the number is wrong.
From protocol to manuscript, with traceability
Seven connected stages, with every decision recorded and every result linked to its source.
01
Protocol first
Question, design, inclusion and exclusion criteria, index date, observation period, and statistical plan are recorded before any query, with the source of each decision — researcher, AI, or pending.
02
Database queries with transparent cost
Queries are generated and run automatically. Each is stored in full, with the volume read and cost.
03
Cohort validation
A step-by-step attrition table following STROBE, with every number produced by its source query. Patient counts are checked through two independent paths.
04
Statistical analysis
Kaplan-Meier, Cox, competing risks, propensity scores with covariate balance, and rates with confidence intervals.
05
Publication figures and tables
11 figure types at 300 DPI. Table 1 and the flowchart are assembled from computed data, never retyped.
06
Integrated literature
Searches in PubMed and ClinicalTrials.gov, with every cited reference supported by the recorded search.
07
Manuscript
AI drafts the sections. The researcher can edit any section on screen, and those edits are respected.
National data connected to the study question
Outpatient care (SIA)
More than 4 billion individualized procedures.
Specialized medicines
More than 420 million dispensations for treatment patterns and longitudinal follow-up.
Chemotherapy and radiotherapy
Approximately 62 million and 4 million records.
Hospital admissions (SIH)
224 million authorizations, with diagnoses, length of stay, ICU, and outcome.
Hospital Cancer Registries
The national INCA database, with 5.4 million cases, and the São Paulo Oncocentro Foundation database, with hormone receptors, HER2, recurrence, and vital status.
IBGE population
By municipality and year, for rates and incidence.
Your own data
Attach Excel or CSV files and integrate them into the study.
A catalog of 246 million patients makes it possible to follow the same patient across outpatient and hospital care. For cancer registries, the platform performs probabilistic linkage with chemotherapy and hospitalization data.
Rigor verified, not promised
A plausible number, a polished figure, or a convincing sentence does not leave the platform if the method performed differs from the method declared.
The method described in the text is the method executed on the data — age range, period, and exclusion window.
Written hazard ratios, confidence intervals, medians, and p-values match the recorded analyses.
The strength of the conclusion matches the strength of the evidence: an observational study does not claim causality.
Every number in the selection flowchart comes from the query that produced it.
Rates have a declared and computed denominator.
Every cited figure and table exists, and every existing figure is cited.
Missing data is treated as missing, never as a clinical category.
Ethics, funding, and conflicts of interest are resolved.
The seal is granted, not requested
The “ready for publication” status is assigned by automatic evaluation, never by the AI, and is withdrawn if the study changes after approval.
Four dimensions, no average
Scientific integrity, reporting completeness (STROBE and RECORD), reproducibility, and submission completeness appear separately, because a completed checklist is not the same as correct science.
AI and data governance
Governed AI, from the first turn to submission
Data classification
Each study states whether data is public and anonymized, pseudonymized, or identifiable; an ethical justification that contradicts that classification blocks the study.
AI model tracking
Every response records which model produced it. The model is selected and tested by Techtrials before use.
Protected identifiers
Columns that identify people, such as CPF and SUS Card, are withheld before reaching AI, including in client spreadsheets, and the researcher is warned before sending them.
Mandatory human review
No study reaches “ready for publication” without researcher confirmation. If the study changes, confirmation must be repeated.
AI use disclosure (ICMJE)
Generated from what actually happened. Sections written by the researcher are not attributed to AI, and human review is declared only when it occurred.
Robust methods, publication-ready outputs
Analyses
Figures
300 DPI, opened full screen.
A study prepared for review and submission
Manuscript in Word and PDF
A clean journal version and a blinded version without authors for peer review.
Submission package
Protocol, STROBE/RECORD checklist, declarations, supplementary material with the code list, attrition table, source data for every figure, and all executed queries.
Reproducibility stated honestly
Traceable, auditable, or computationally reproducible, according to what the study supports, with a guide for reproducing the analyses.
Governance record
AI and data governance record included in the package.
Evidence for those who research, decide, and transform care
Pharmaceutical industry
Medical Affairs, HEOR, and Market Access
Disease burden and epidemiology in Brazil's public health system, treatment patterns and therapy lines, resource use — admissions, length of stay, and ICU —, real-world survival, and local evidence for access and adoption discussions.
Researchers and academic institutions
National observational studies
Studies using national data without a data engineering team, with manuscripts structured from the outset for journal requirements.
Healthcare managers and medical societies
Care and population
Views of care by state and over time, with population rates.
The experience
Work in natural language
Study panel with protocol, evidence, and manuscript side by side
Your edits are respected by AI
Dark and light themes
What researchers want to know
