Disposition from Several Source Forms

1library(yamaa)23yamaa_spec("spec.yaml")$run()$output

1import yamaa23yamaa.spec("spec.yaml").run().output

1# Reference solution for the yamaa benchmark sdtm-ds-multi-form-disposition (R track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67library(dplyr, warn.conflicts = FALSE)8library(readr)9library(tidyr)1011odm <- read_csv(12  "/app/input/odm.csv",13  col_types = cols(.default = col_character())14)1516wide <- odm |>17  group_by(StudyOID, SubjectKey, StudyEventOID) |>18  summarise(19    DTC = Value[ItemOID == "IT.DS.DTC"][1],20    COMP = Value[ItemOID == "IT.DS.COMP"][1],21    REASON = Value[ItemOID == "IT.DS.REASON"][1],22    REASONCD = Value[ItemOID == "IT.DS.REASONCD"][1],23    .groups = "drop"24  )2526ds <- wide |>27  rowwise() |>28  mutate(29    DSTERM = case_when(30      StudyEventOID == "CONSENT" ~ "INFORMED CONSENT OBTAINED",31      StudyEventOID == "RAND" ~ "RANDOMIZED",32      StudyEventOID == "EOT" & COMP == "COMPLETED" ~ "COMPLETED",33      StudyEventOID == "EOT" ~ REASON,34      StudyEventOID == "EOS" & COMP == "COMPLETED" ~ "COMPLETED",35      StudyEventOID == "EOS" & COMP == "SCREEN FAILURE" ~ "SCREEN FAILURE",36      StudyEventOID == "EOS" ~ REASON,37      .default = NA_character_38    ),39    DSDECOD = case_when(40      StudyEventOID == "CONSENT" ~ "INFORMED CONSENT OBTAINED",41      StudyEventOID == "RAND" ~ "RANDOMIZED",42      StudyEventOID == "EOT" & COMP == "COMPLETED" ~ "COMPLETED",43      StudyEventOID == "EOT" ~ REASONCD,44      StudyEventOID == "EOS" & COMP == "COMPLETED" ~ "COMPLETED",45      StudyEventOID == "EOS" & COMP == "SCREEN FAILURE" ~ "SCREEN FAILURE",46      StudyEventOID == "EOS" ~ REASONCD,47      .default = NA_character_48    ),49    DSCAT = case_when(50      StudyEventOID %in% c("CONSENT", "RAND") ~ "PROTOCOL MILESTONE",51      StudyEventOID %in% c("EOT", "EOS") ~ "DISPOSITION EVENT",52      .default = NA_character_53    ),54    DSSCAT = case_when(55      StudyEventOID == "CONSENT" ~ "INFORMED CONSENT",56      StudyEventOID == "RAND" ~ "RANDOMIZATION",57      StudyEventOID == "EOT" ~ "END OF TREATMENT",58      StudyEventOID == "EOS" ~ "END OF STUDY",59      .default = NA_character_60    )61  ) |>62  ungroup() |>63  filter(StudyEventOID %in% c("CONSENT", "RAND", "EOT", "EOS")) |>64  mutate(65    DOMAIN = "DS",66    STUDYID = StudyOID,67    USUBJID = paste(StudyOID, SubjectKey, sep = "-"),68    DSSTDTC = suppressWarnings(as.Date(na_if(DTC, ""))),69    DSTERM = na_if(DSTERM, ""),70    DSDECOD = na_if(DSDECOD, "")71  ) |>72  arrange(USUBJID, DSSTDTC, DSTERM) |>73  group_by(USUBJID) |>74  mutate(DSSEQ = row_number()) |>75  ungroup() |>76  select(DOMAIN, STUDYID, USUBJID, DSSEQ, DSTERM, DSDECOD, DSCAT, DSSCAT, DSSTDTC)7778dir.create("/app/output", showWarnings = FALSE)79write_csv(ds, "/app/output/ds.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-ds-multi-form-disposition (Python track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67"""Build the SDTM DS dataset from the collected ODM extract.89One record per subject per collected form (study event). Reads10/app/input/odm.csv (long EAV form: one row per collected item) and11writes /app/output/ds.csv.12"""1314import csv15import os16from collections import OrderedDict1718INPUT_DIR = "/app/input"19OUTPUT_DIR = "/app/output"2021# StudyEventOID -> (DSCAT, DSSCAT, controlled term for a milestone)22EVENT_INFO = {23    "CONSENT": ("PROTOCOL MILESTONE", "INFORMED CONSENT",24                "INFORMED CONSENT OBTAINED"),25    "RAND": ("PROTOCOL MILESTONE", "RANDOMIZATION", "RANDOMIZED"),26    "EOT": ("DISPOSITION EVENT", "END OF TREATMENT", None),27    "EOS": ("DISPOSITION EVENT", "END OF STUDY", None),28}2930COLUMNS = ["DOMAIN", "STUDYID", "USUBJID", "DSSEQ", "DSTERM", "DSDECOD",31           "DSCAT", "DSSCAT", "DSSTDTC"]323334def main():35    with open(os.path.join(INPUT_DIR, "odm.csv"), newline="") as f:36        rows = list(csv.DictReader(f))3738    # Group the EAV rows into forms: one form per subject, study event39    # and repeat keys. A study event with no form gives no record.40    forms = OrderedDict()41    for row in rows:42        key = (row["StudyOID"], row["SubjectKey"], row["StudyEventOID"],43               row["StudyEventRepeatKey"], row["ItemGroupOID"],44               row["ItemGroupRepeatKey"])45        forms.setdefault(key, {})[row["ItemOID"]] = row["Value"] or ""4647    records = []48    for (study_oid, subject_key, event_oid, *_), items in forms.items():49        dscat, dsscat, milestone_term = EVENT_INFO[event_oid]50        completion = items.get("IT.DS.COMP", "")51        if milestone_term is not None:52            dsterm = dsdecod = milestone_term53        elif completion == "COMPLETED":54            dsterm = dsdecod = "COMPLETED"55        elif event_oid == "EOS" and completion == "SCREEN FAILURE":56            dsterm = dsdecod = "SCREEN FAILURE"57        else:58            dsterm = items.get("IT.DS.REASON", "")59            dsdecod = items.get("IT.DS.REASONCD", "")60        records.append({61            "DOMAIN": "DS",62            "STUDYID": study_oid,63            "USUBJID": "{}-{}".format(study_oid, subject_key),64            "DSTERM": dsterm,65            "DSDECOD": dsdecod,66            "DSCAT": dscat,67            "DSSCAT": dsscat,68            "DSSTDTC": items.get("IT.DS.DTC", ""),69        })7071    # DSSEQ numbers each subject's records from 1 in the order they72    # happened, by DSSTDTC and then by DSTERM.73    by_subject = OrderedDict()74    for record in records:75        by_subject.setdefault(record["USUBJID"], []).append(record)7677    out_rows = []78    for subject_records in by_subject.values():79        subject_records.sort(key=lambda r: (r["DSSTDTC"], r["DSTERM"]))80        for seq, record in enumerate(subject_records, start=1):81            record["DSSEQ"] = seq82            out_rows.append(record)8384    os.makedirs(OUTPUT_DIR, exist_ok=True)85    with open(os.path.join(OUTPUT_DIR, "ds.csv"), "w", newline="") as f:86        writer = csv.DictWriter(f, fieldnames=COLUMNS)87        writer.writeheader()88        for record in out_rows:89            writer.writerow({col: record[col] for col in COLUMNS})909192if __name__ == "__main__":93    main()

1/* opensas reference solution for the yamaa benchmark sdtm-ds-multi-form-disposition.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data odm;6  length StudyOID MetaDataVersionOID SubjectKey StudyEventOID StudyEventRepeatKey ItemGroupOID ItemGroupRepeatKey ItemOID Value $1024;7  infile "/app/input/odm.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input StudyOID : $char1024. MetaDataVersionOID : $char1024. SubjectKey : $char1024. StudyEventOID : $char1024. StudyEventRepeatKey : $char1024. ItemGroupOID : $char1024. ItemGroupRepeatKey : $char1024. ItemOID : $char1024. Value : $char1024.;9run;1011proc sort data=odm out=ordered; by StudyOID SubjectKey StudyEventOID; run;12data wide;13  length COMP DTC REASON REASONCD $1024;14  retain COMP DTC REASON REASONCD;15  set ordered;1617  by StudyOID SubjectKey StudyEventOID;18  if first.StudyEventOID then do;19    COMP=''; DTC=''; REASON=''; REASONCD='';20  end;21    if ItemOID='IT.DS.COMP' then COMP=Value;22  if ItemOID='IT.DS.DTC' then DTC=Value;23  if ItemOID='IT.DS.REASON' then REASON=Value;24  if ItemOID='IT.DS.REASONCD' then REASONCD=Value;25  if last.StudyEventOID then output;26run;27data prepared; set wide; length DOMAIN $2 STUDYID USUBJID DSTERM DSDECOD DSCAT DSSCAT DSSTDTC $1024;28DOMAIN='DS'; STUDYID=StudyOID; USUBJID=SubjectKey; DSCAT='DISPOSITION EVENT'; DSSTDTC=DTC; DSTERM=REASON; DSDECOD=REASONCD;29if COMP='COMPLETED' then do; DSTERM='COMPLETED'; DSDECOD='COMPLETED'; end;30USUBJID=catx('-',StudyOID,SubjectKey);31if StudyEventOID='CONSENT' then do; DSTERM='INFORMED CONSENT OBTAINED'; DSDECOD=DSTERM; DSCAT='PROTOCOL MILESTONE'; DSSCAT='INFORMED CONSENT'; end;32if StudyEventOID='RAND' then do; DSTERM='RANDOMIZED'; DSDECOD=DSTERM; DSCAT='PROTOCOL MILESTONE'; DSSCAT='RANDOMIZATION'; end;33if StudyEventOID='EOT' then DSSCAT='END OF TREATMENT';34if StudyEventOID='EOS' then do; DSSCAT='END OF STUDY'; if COMP='SCREEN FAILURE' then do; DSTERM='SCREEN FAILURE'; DSDECOD=DSTERM; end; end; run;35proc sort data=prepared; by STUDYID USUBJID DSSTDTC DSTERM; run;36data result; set prepared; by STUDYID USUBJID; retain DSSEQ;37  if first.USUBJID then DSSEQ=0;38  DSSEQ+1;39run;4041proc sql;42  create table final as select DOMAIN, STUDYID, USUBJID, DSSEQ, DSTERM, DSDECOD, DSCAT, DSSCAT, DSSTDTC from result;43quit;44proc export data=final outfile="/app/output/ds.csv" dbms=csv replace; run;