Repeated Adverse Events from ODM

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

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

1# Reference solution for the yamaa benchmark sdtm-ae-odm-repeated (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)1516# One event per AE item group occurrence: subject, visit, visit repeat, item17# group, and item group repeat key. An occurrence without a reported term18# gives no record; a repeat key reused at a later visit starts a new event.19events <- odm |>20  filter(ItemGroupOID == "IG.AE") |>21  pivot_wider(22    id_cols = c(23      StudyOID, SubjectKey, StudyEventOID, StudyEventRepeatKey,24      ItemGroupRepeatKey25    ),26    names_from = ItemOID,27    values_from = Value,28    values_fn = first29  ) |>30  mutate(31    AETERM = `IT.AE.AETERM`,32    AESTDTC = `IT.AE.AESTDTC`,33    AEENDTC = `IT.AE.AEENDTC`,34    AESEV = `IT.AE.AESEV`,35    AESER = `IT.AE.AESER`36  ) |>37  filter(!is.na(AETERM) & AETERM != "") |>38  # Visit order is screening, then baseline; then visit repeat and item39  # group repeat key. The same term in two occurrences stays two records.40  mutate(41    VISITORD = case_when(42      StudyEventOID == "SCREENING" ~ 0L,43      StudyEventOID == "BASELINE" ~ 1L,44      .default = 2L45    ),46    VISITREP = as.integer(StudyEventRepeatKey),47    IGREP = as.integer(ItemGroupRepeatKey)48  ) |>49  arrange(SubjectKey, VISITORD, VISITREP, IGREP) |>50  group_by(SubjectKey) |>51  mutate(AESEQ = row_number()) |>52  ungroup() |>53  mutate(DOMAIN = "AE", STUDYID = StudyOID, USUBJID = SubjectKey) |>54  select(55    DOMAIN, STUDYID, USUBJID, AESEQ, AETERM,56    AESTDTC, AEENDTC, AESEV, AESER57  )5859dir.create("/app/output", showWarnings = FALSE)60write_csv(events, "/app/output/ae.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-ae-odm-repeated (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 AE dataset from the ODM long-format extract.89One record per reported event: an event is one item-group occurrence on an10AE form occurrence (subject, visit, visit repeat, form, form repeat).11Occurrences without a reported term contribute no record.12"""1314import csv1516INPUT_PATH = "/app/input/odm.csv"17OUTPUT_PATH = "/app/output/ae.csv"1819ITEM_NAMES = {20    "IT.AE.AETERM": "AETERM",21    "IT.AE.AESTDTC": "AESTDTC",22    "IT.AE.AEENDTC": "AEENDTC",23    "IT.AE.AESEV": "AESEV",24    "IT.AE.AESER": "AESER",25}2627VISIT_ORDER = {"SCREENING": 0, "BASELINE": 1}282930def repeat_number(value):31    try:32        return int(value)33    except (TypeError, ValueError):34        return 0353637def main():38    with open(INPUT_PATH, newline="", encoding="utf-8") as handle:39        rows = list(csv.DictReader(handle))4041    # Group the AE form items by event occurrence, keeping first-appearance42    # order of visits for any visit not in the known visit order.43    events = {}44    visit_rank_seen = {}45    for row in rows:46        if row["FormOID"] != "FO.AE":47            continue48        visit = row["StudyEventOID"]49        if visit not in visit_rank_seen:50            visit_rank_seen[visit] = len(visit_rank_seen)51        key = (52            row["StudyOID"],53            row["SubjectKey"],54            visit,55            row["StudyEventRepeatKey"],56            row["FormOID"],57            row["FormRepeatKey"],58            row["ItemGroupOID"],59            row["ItemGroupRepeatKey"],60        )61        event = events.setdefault(key, {})62        field = ITEM_NAMES.get(row["ItemOID"])63        if field is not None:64            event[field] = row["Value"]6566    def sort_key(item):67        key, _event = item68        (_study, _subject, visit, visit_repeat, _form, form_repeat,69         _group, group_repeat) = key70        visit_rank = VISIT_ORDER.get(visit, 100 + visit_rank_seen[visit])71        return (72            visit_rank,73            repeat_number(visit_repeat),74            repeat_number(form_repeat),75            repeat_number(group_repeat),76        )7778    records_by_subject = {}79    for key, event in events.items():80        term = event.get("AETERM", "")81        if term == "":82            continue83        studyid, usubjid = key[0], key[1]84        records_by_subject.setdefault((studyid, usubjid), []).append(85            (key, event))8687    output_rows = []88    for (studyid, usubjid), subject_events in records_by_subject.items():89        subject_events.sort(key=sort_key)90        for seq, (_key, event) in enumerate(subject_events, start=1):91            output_rows.append({92                "DOMAIN": "AE",93                "STUDYID": studyid,94                "USUBJID": usubjid,95                "AESEQ": seq,96                "AETERM": event.get("AETERM", ""),97                "AESTDTC": event.get("AESTDTC", ""),98                "AEENDTC": event.get("AEENDTC", ""),99                "AESEV": event.get("AESEV", ""),100                "AESER": event.get("AESER", ""),101            })102103    fieldnames = ["DOMAIN", "STUDYID", "USUBJID", "AESEQ", "AETERM",104                  "AESTDTC", "AEENDTC", "AESEV", "AESER"]105    with open(OUTPUT_PATH, "w", newline="", encoding="utf-8") as handle:106        writer = csv.DictWriter(handle, fieldnames=fieldnames)107        writer.writeheader()108        writer.writerows(output_rows)109110111if __name__ == "__main__":112    main()

1/* opensas reference solution for the yamaa benchmark sdtm-ae-odm-repeated.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 FormOID FormRepeatKey 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. FormOID : $char1024. FormRepeatKey : $char1024. ItemGroupOID : $char1024. ItemGroupRepeatKey : $char1024. ItemOID : $char1024. Value : $char1024.;9run;1011proc sort data=odm out=ordered; by StudyOID SubjectKey StudyEventOID StudyEventRepeatKey ItemGroupOID ItemGroupRepeatKey; run;12data wide;13  length AETERM AESTDTC AEENDTC AESEV AESER $1024;14  retain AETERM AESTDTC AEENDTC AESEV AESER;15  set ordered;16  where ItemGroupOID='IG.AE';17  by StudyOID SubjectKey StudyEventOID StudyEventRepeatKey ItemGroupOID ItemGroupRepeatKey;18  if first.ItemGroupRepeatKey then do;19    AETERM=''; AESTDTC=''; AEENDTC=''; AESEV=''; AESER='';20  end;21    if ItemOID='IT.AE.AETERM' then AETERM=Value;22  if ItemOID='IT.AE.AESTDTC' then AESTDTC=Value;23  if ItemOID='IT.AE.AEENDTC' then AEENDTC=Value;24  if ItemOID='IT.AE.AESEV' then AESEV=Value;25  if ItemOID='IT.AE.AESER' then AESER=Value;26  if last.ItemGroupRepeatKey then output;27run;28data prepared; set wide; if not missing(AETERM);29length DOMAIN $2 STUDYID USUBJID $1024; DOMAIN='AE'; STUDYID=StudyOID; USUBJID=SubjectKey;30VISITORD=2; if StudyEventOID='SCREENING' then VISITORD=0; if StudyEventOID='BASELINE' then VISITORD=1;31VISITREP=input(StudyEventRepeatKey,best32.); IGREP=input(ItemGroupRepeatKey,best32.); run;32proc sort data=prepared; by STUDYID USUBJID VISITORD VISITREP IGREP; run;33data result; set prepared; by STUDYID USUBJID; retain AESEQ;34  if first.USUBJID then AESEQ=0;35  AESEQ+1;36run;3738proc sql;39  create table final as select DOMAIN, STUDYID, USUBJID, AESEQ, AETERM, AESTDTC, AEENDTC, AESEV, AESER from result;40quit;41proc export data=final outfile="/app/output/ae.csv" dbms=csv replace; run;