Incoming Studies for a Biostat Core
Investigators send a core their phase 1 and 2 trials and chart reviews as Excel files of different shapes. Each one gets a visual cell check, SAS or R consistency checks and a new custom script, because the data is different every time. Cleaning alone takes a quarter of the study's time; programming, analysis and the report take the rest, and turnaround runs from half an hour to a week.
Teams need the checks and the cleaning to run the same way on every file, the statistical choices confirmed with the investigator instead of guessed, and the script written from an approved plan rather than from scratch.
Done by hand, the core writes a new script for each study, the empty cells are chased by email, and the follow-up requests arrive after the report has gone out.
The quarter of each study spent cleaning, cut to a review
A fixed set of cleaning operations is proposed with counts and approved at a gate, so the statistician reviews instead of rebuilding the checks.
Questions confirmed with the investigator, 2 to 4 options each
Where methods differ in ways that matter, Sutrix asks one question with two to four options, a rationale and a recommendation, and the answer is recorded.
Every number derived twice
The analysis runs from the approved plan in Python or R, and an independent validation re-derives each number from the plan alone.
How it works
Receive the file as sent: Excel, CSV or an EDC export, hashed and kept as received, screened for identifier classes.
Audit, then approve the cleaning: the same checks on every file; the proposal lists each operation with its count and waits for the statistician's decision.
Confirm the plan: the plan is extracted from the question and protocol, the objection ledger is written, and the investigator's open questions are asked once.
Run, validate and report: the analysis executes, the independent validation reconciles it, and the report carries Methods, Results tables, Missing data and a plain-words section.