Review ask

As a clinical trial statistician or programmer, challenge anything in this benchmark: describe a problem in Comments, or propose a fix with the Edit button on any source file.

  1. Task: Would a real study need this derivation, specified this way? Does it follow CDISC SDTM: variable names, controlled terms, record structure? Flag any rule that is ambiguous or that a sponsor would handle differently.
  2. Inputs: Do they look like data collected in a real trial, and do they include the situations you would expect to meet in one?
  3. Expected output: Would independent QC programming reproduce every value from these inputs and rules?

The other tabs beside Comments are background for assessment, not under review.

Summary

Lifecycle: reviewedEditUnder review
What this benchmark means
README.md

Goal: create one supplemental record per collected qualifier, carrying IDVARVAL, QLABEL, QVAL, QORIG, and QEVAL.

Input: two tables sharing study and subject identifiers: a parent medical history (MH) table with one record per condition, and a qualifier table with one record per condition carrying the two collected qualifiers.

Variables:

  • IDVARVAL is the parent sequence number as text, matched on subject and condition term.
  • QLABEL is the qualifier label: Family History for MHFAMHX, Confirmed by Medical Records for MHCONF.
  • QVAL is the collected answer, Y or N.
  • QORIG is the origin, always Collected (Define-XML 2.1) for collected values.
  • QEVAL is blank, since a collected value is not an assessment.

Note: a qualifier that finds no parent record is an error rather than a record with an empty link, so every supplemental record points at a real parent. A parent record with no collected qualifiers simply contributes nothing.

Inputs

Under review
2 source files

input/mh.csv

Edit
5 rows
input/mh.csv
STUDYIDUSUBJIDMHSEQMHTERM
CATH1ASTHMA
CATH2ECZEMA
CATH1HYPERTENSION
CATH2ASTHMA
CATH1DIABETES

input/mh_supp_raw.csv

Edit
4 rows
input/mh_supp_raw.csv
STUDYIDUSUBJIDMHTERMMHFAMHXMHCONF
CATHASTHMAYY
CATHECZEMAN—
CATHASTHMAYN
CATHDIABETESNY

Expected output

Under review
7 expected rows

expected/suppmh.csv

Edit
7 rows
expected/suppmh.csv
STUDYIDRDOMAINUSUBJIDIDVARIDVARVALQNAMQLABELQVALQORIGQEVAL
CATHMHMHSEQ1MHFAMHXFamily HistoryYCollected—
CATHMHMHSEQ2MHFAMHXFamily HistoryNCollected—
CATHMHMHSEQ2MHFAMHXFamily HistoryYCollected—
CATHMHMHSEQ1MHFAMHXFamily HistoryNCollected—
CATHMHMHSEQ1MHCONFConfirmed by Medical RecordsYCollected—
CATHMHMHSEQ2MHCONFConfirmed by Medical RecordsNCollected—
CATHMHMHSEQ1MHCONFConfirmed by Medical RecordsYCollected—
Shaded columns are derived— = missing value

Comments and assessment materials

Comments

Signed in with GitHub; saved in repository Discussions

Mapping spec

Generated from the benchmark YAML; do not edit by hand

Variable NameVariable LabelTypeLengthControlled Terms or FormatOriginCoreConversion DefinitionVariable TypeVariable OrderComments for Define
STUDYIDStudy IdentifierCharCollectedCopy value from MH_SUPP_RAW.STUDYID.SUPP1
RDOMAINRelated Domain AbbreviationCharAssignedSet constant value "MH".SUPP2
USUBJIDUnique Subject IdentifierCharCollectedCopy value from MH_SUPP_RAW.USUBJID.SUPP3
IDVARIdentifying VariableCharAssignedSet constant value "MHSEQ".SUPP4
IDVARVALIdentifying Variable ValueCharDerivedTake MHSEQ from the MH record whose USUBJID equals MH_SUPP_RAW.USUBJID and MHTERM equals MH_SUPP_RAW.MHTERM, and a row with no match is an error.SUPP5
QNAMQualifier Variable NameCharMHFAMHX, MHCONFDerivedSet per constructed row; see the Row construction sheet.SUPP6
QLABELQualifier LabelCharDerivedSet per constructed row; see the Row construction sheet.SUPP7
QVALData ValueCharY, NDerivedSet per constructed row; see the Row construction sheet.SUPP8
QORIGOriginCharAssignedSet constant value "Collected".SUPP9
QEVALEvaluatorCharAssignedSet constant value "None".SUPP10
TemplateInclude whenQNAMQLABELQVAL
family_historyMH_SUPP_RAW.MHFAMHX IS NOT NULLConstant "MHFAMHX"Constant "Family History"Copy MH_SUPP_RAW.MHFAMHX
medical_record_confirmationMH_SUPP_RAW.MHCONF IS NOT NULLConstant "MHCONF"Constant "Confirmed by Medical Records"Copy MH_SUPP_RAW.MHCONF
VariablePermitted values
QNAMMHFAMHX, MHCONF
QVALY, N
VersionDateAuthorDescriptionReviewerSign-off
1.32026-09-29yamaa docs buildShow study codelist names, external dictionaries, and ISO 8601 formats in variable rows; list the document's codelist values and metadata
1.22026-09-29yamaa docs buildClassify ADaM copies from a declared input as Predecessor and recodes as Derived, and SDTM recodes of locally derived columns as Derived
1.12026-09-29yamaa docs buildResolve lookups into the donor record they select, and give every derivation shape a plain-language rule instead of its raw YAML
1.0yamaa docs buildInitial generation from spec.yaml

Agent prompt

The request an agent receives with the input files; each tier states no more than the one above

prompts/sdtm-suppmh-linkage/full.md

Edit
20 lines

Following CDISC SDTM standards, use the provided MH_SUPP_RAW and MH datasets to create a SUPPMH dataset with one supplemental record per collected qualifier.

The output dataset should contain the following columns in this order: STUDYID, RDOMAIN, USUBJID, IDVAR, IDVARVAL, QNAM, QLABEL, QVAL, QORIG, QEVAL

RDOMAIN is MH and IDVAR is MHSEQ. IDVARVAL is the parent sequence number as text, matched on subject and condition term. QNAM is the qualifier name, MHFAMHX or MHCONF. QLABEL is Family History for MHFAMHX and Confirmed by Medical Records for MHCONF. QVAL is the collected answer, Y or N. QORIG is Collected for collected values. QEVAL has no value, since a collected value is not an assessment.

A qualifier left blank gives no record, and a parent record with no collected qualifiers contributes nothing.

Read the source datasets from /app/input and save the completed dataset as /app/output/suppmh.csv.

prompts/sdtm-suppmh-linkage/conventions.md

Edit
15 lines

Following CDISC SDTM standards, use the provided MH_SUPP_RAW and MH datasets to create a SUPPMH dataset with one supplemental record per collected qualifier.

The output dataset should contain the following columns in this order: STUDYID, RDOMAIN, USUBJID, IDVAR, IDVARVAL, QNAM, QLABEL, QVAL, QORIG, QEVAL

RDOMAIN is MH and IDVAR is MHSEQ. IDVARVAL is the parent sequence number as text. QNAM is the qualifier name, MHFAMHX or MHCONF. QLABEL is Family History for MHFAMHX and Confirmed by Medical Records for MHCONF. QVAL is Y or N. QORIG is Collected for collected values. QEVAL has no value.

Read the source datasets from /app/input and save the completed dataset as /app/output/suppmh.csv.

prompts/sdtm-suppmh-linkage/brief.md

Edit
10 lines

Following CDISC SDTM standards, use the provided MH_SUPP_RAW and MH datasets to create a SUPPMH dataset with one supplemental record per collected qualifier.

The output dataset should contain the following columns in this order: STUDYID, RDOMAIN, USUBJID, IDVAR, IDVARVAL, QNAM, QLABEL, QVAL, QORIG, QEVAL

Read the source datasets from /app/input and save the completed dataset as /app/output/suppmh.csv.

yamaa code

Runs the yamaa specification; its result must match the expected output

run.R

Edit
4 lines
library(yamaa)suppmh <- yamaa_domain("spec.yaml")$outputsuppmh

run.py

Edit
4 lines
import yamaasuppmh = yamaa.yamaa_domain("spec.yaml").outputsuppmh

Solution without yamaa

Written from the full prompt and the inputs alone; CI checks it reproduces the expected output

solutions/sdtm-suppmh-linkage/result.R

Edit
48 lines
# Reference solution for the yamaa benchmark sdtm-suppmh-linkage (R track).## Written from the benchmark's full prompt and its inputs alone, as an agent# would write it. Harbor's oracle agent runs it to check that the prompt can# be solved and that the grader scores a correct answer 1.library(dplyr, warn.conflicts = FALSE)library(readr)library(tidyr)mh <- read_csv(  "/app/input/mh.csv",  col_types = cols(.default = col_character()))raw <- read_csv(  "/app/input/mh_supp_raw.csv",  col_types = cols(.default = col_character()))# The parent sequence number, matched on subject and condition term.parents <- mh |> select(STUDYID, USUBJID, MHSEQ, MHTERM)suppmh <- raw |>  pivot_longer(    cols = c(MHFAMHX, MHCONF),    names_to = "QNAM",    values_to = "QVAL"  ) |>  filter(!is.na(QVAL) & QVAL != "") |>  inner_join(parents, by = c("STUDYID", "USUBJID", "MHTERM")) |>  mutate(    RDOMAIN = "MH",    IDVAR = "MHSEQ",    IDVARVAL = MHSEQ,    QLABEL = case_when(      QNAM == "MHFAMHX" ~ "Family History",      QNAM == "MHCONF" ~ "Confirmed by Medical Records"    ),    QORIG = "Collected",    QEVAL = NA_character_  ) |>  arrange(USUBJID, suppressWarnings(as.integer(MHSEQ)), QNAM) |>  select(    STUDYID, RDOMAIN, USUBJID, IDVAR, IDVARVAL, QNAM, QLABEL, QVAL, QORIG, QEVAL  )dir.create("/app/output", showWarnings = FALSE)write_csv(suppmh, "/app/output/suppmh.csv", na = "")

solutions/sdtm-suppmh-linkage/result.py

Edit
61 lines
# Reference solution for the yamaa benchmark sdtm-suppmh-linkage (Python track).## Written from the benchmark's full prompt and its inputs alone, as an agent# would write it. Harbor's oracle agent runs it to check that the prompt can# be solved and that the grader scores a correct answer 1.from pathlib import Pathimport polars as plmh = pl.read_csv("/app/input/mh.csv", infer_schema=False)raw = pl.read_csv("/app/input/mh_supp_raw.csv", infer_schema=False)labels = {"MHFAMHX": "Family History", "MHCONF": "Confirmed by Medical Records"}# The parent sequence number, matched on subject and condition term.parents = mh.select("STUDYID", "USUBJID", "MHSEQ", "MHTERM")records = []for qnam in ("MHFAMHX", "MHCONF"):    collected = (        raw.filter(pl.col(qnam).is_not_null() & (pl.col(qnam) != ""))        .join(parents, on=["STUDYID", "USUBJID", "MHTERM"], how="inner")        .select(            "STUDYID",            "USUBJID",            "MHSEQ",            QNAM=pl.lit(qnam),            QLABEL=pl.lit(labels[qnam]),            QVAL=pl.col(qnam),        )    )    records.append(collected)suppmh = (    pl.concat(records)    .with_columns(        RDOMAIN=pl.lit("MH"),        IDVAR=pl.lit("MHSEQ"),        IDVARVAL=pl.col("MHSEQ"),        QORIG=pl.lit("Collected"),        QEVAL=pl.lit(None, dtype=pl.String),    )    .with_columns(pl.col("MHSEQ").cast(pl.Int64).alias("_seq"))    .sort(["USUBJID", "_seq", "QNAM"])    .select(        "STUDYID",        "RDOMAIN",        "USUBJID",        "IDVAR",        "IDVARVAL",        "QNAM",        "QLABEL",        "QVAL",        "QORIG",        "QEVAL",    ))Path("/app/output").mkdir(exist_ok=True)suppmh.write_csv("/app/output/suppmh.csv")

solutions/sdtm-suppmh-linkage/result.sas

Edit
24 lines
/* opensas reference solution for the yamaa benchmark sdtm-suppmh-linkage.   Reads the benchmark inputs and derives the requested output.   Execute with the existing opensas runtime: opensas /app/output/result.sas. */data mh;  length STUDYID USUBJID MHTERM $1024;  infile "/app/input/mh.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;  input STUDYID : $char1024. USUBJID : $char1024. MHSEQ MHTERM : $char1024.;run;data mh_supp_raw;  length STUDYID USUBJID MHTERM MHFAMHX MHCONF $1024;  infile "/app/input/mh_supp_raw.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;  input STUDYID : $char1024. USUBJID : $char1024. MHTERM : $char1024. MHFAMHX : $char1024. MHCONF : $char1024.;run;proc sql; create table qualified as select a.*,b.MHSEQ from mh_supp_raw a inner join mh b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID and a.MHTERM=b.MHTERM; quit;data result; set qualified; length RDOMAIN IDVAR IDVARVAL QNAM QLABEL QVAL QORIG QEVAL $1024; RDOMAIN='MH'; IDVAR='MHSEQ'; IDVARVAL=strip(put(MHSEQ,best32.)); QORIG='Collected'; QEVAL='';if not missing(MHFAMHX) then do; QNAM='MHFAMHX'; QLABEL='Family History'; QVAL=MHFAMHX; output; end;if not missing(MHCONF) then do; QNAM='MHCONF'; QLABEL='Confirmed by Medical Records'; QVAL=MHCONF; output; end; run;proc sql;  create table final as select STUDYID, RDOMAIN, USUBJID, IDVAR, IDVARVAL, QNAM, QLABEL, QVAL, QORIG, QEVAL from result;quit;proc export data=final outfile="/app/output/suppmh.csv" dbms=csv replace; run;