Event Findings

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

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

1# Reference solution for the yamaa benchmark sdtm-fa-event-findings (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)10library(purrr)1112odm <- read_csv(13  "/app/input/odm.csv",14  col_types = cols(.default = col_character())15)1617long <- odm |>18  mutate(Value = na_if(Value, ""))1920ae <- long |>21  filter(ItemGroupOID == "IG.AE") |>22  select(SubjectKey, StudyOID, StudyEventOID, StudyEventRepeatKey, ItemOID, Value) |>23  pivot_wider(names_from = ItemOID, values_from = Value) |>24  rename(25    AETERM = `IT.AE.AETERM`,26    AELNKID_AE = `IT.AE.AELNKID`27  )2829fa_forms <- long |>30  filter(ItemGroupOID == "IG.FA") |>31  select(32    SubjectKey, StudyOID, StudyEventOID, StudyEventRepeatKey,33    ItemGroupRepeatKey, ItemOID, Value34  ) |>35  pivot_wider(names_from = ItemOID, values_from = Value) |>36  rename(37    LINK_FA = `IT.FA.AELNKID`,38    LOCATION = `IT.FA.LOCATION`,39    SIZE = `IT.FA.SIZE`,40    SIZEU = `IT.FA.SIZEU`,41    BIOPSY = `IT.FA.BIOPSY`,42    FADTC = `IT.FA.FADTC`43  )4445fa <- fa_forms |>46  left_join(47    ae,48    by = c("SubjectKey", "StudyEventOID", "StudyEventRepeatKey")49  ) |>50  filter(!is.na(AETERM) & AETERM != "") |>51  mutate(52    LINK = coalesce(LINK_FA, AELNKID_AE),53    STUDYID = StudyOID.x,54    USUBJID = SubjectKey55  ) |>56  select(STUDYID, USUBJID, LINK, FADTC, SIZEU, AETERM, LOCATION, SIZE, BIOPSY) |>57  pivot_longer(58    cols = c(LOCATION, SIZE, BIOPSY),59    names_to = "WHICH",60    values_to = "FAORRES"61  ) |>62  filter(!is.na(FAORRES)) |>63  mutate(64    ORDER = case_when(65      WHICH == "LOCATION" ~ 0L,66      WHICH == "SIZE" ~ 1L,67      .default = 2L68    ),69    FATESTCD = case_when(70      WHICH == "LOCATION" ~ "LOC",71      WHICH == "SIZE" ~ "SIZE",72      .default = "BIOPSY"73    ),74    FATEST = case_when(75      WHICH == "LOCATION" ~ "Location",76      WHICH == "SIZE" ~ "Size",77      .default = "Biopsied"78    ),79    FAOBJ = AETERM,80    FACAT = "AE",81    IS_SIZE = WHICH == "SIZE",82    FAORRESU = if_else(IS_SIZE, SIZEU, NA_character_),83    FASTRESC = FAORRES,84    FASTRESN = if_else(85      IS_SIZE, suppressWarnings(as.numeric(FAORRES)), NA_real_86    ),87    FASTRESU = if_else(IS_SIZE, SIZEU, NA_character_),88    FALNKID = LINK,89    DOMAIN = "FA",90    TIE = FAORRES91  ) |>92  arrange(STUDYID, USUBJID, LINK, ORDER, TIE) |>93  group_by(STUDYID, USUBJID) |>94  mutate(FASEQ = row_number()) |>95  ungroup() |>96  select(97    DOMAIN, STUDYID, USUBJID, FASEQ, FATESTCD, FATEST, FAOBJ, FACAT,98    FAORRES, FAORRESU, FASTRESC, FASTRESN, FASTRESU, FALNKID, FADTC99  )100101dir.create("/app/output", showWarnings = FALSE)102write_csv(fa, "/app/output/fa.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-fa-event-findings (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 FA dataset (findings about adverse events) from ODM data.89Reads the long-format ODM extract from /app/input and writes one FA10record per answered test (LOC, SIZE, BIOPSY) on each findings form11linked to an adverse event occurrence, to /app/output/fa.csv.12"""1314import csv15import os1617INPUT_DIR = "/app/input"18OUTPUT_DIR = "/app/output"1920COLUMNS = [21    "DOMAIN", "STUDYID", "USUBJID", "FASEQ", "FATESTCD", "FATEST",22    "FAOBJ", "FACAT", "FAORRES", "FAORRESU", "FASTRESC", "FASTRESN",23    "FASTRESU", "FALNKID", "FADTC",24]2526TEST_ORDER = {"LOC": 0, "SIZE": 1, "BIOPSY": 2}27TEST_NAMES = {"LOC": "Location", "SIZE": "Size", "BIOPSY": "Biopsied"}282930def read_odm(input_dir):31    """Return the rows of the ODM long-format extract in /app/input."""32    for name in sorted(os.listdir(input_dir)):33        if not name.lower().endswith(".csv"):34            continue35        path = os.path.join(input_dir, name)36        with open(path, newline="", encoding="utf-8") as handle:37            reader = csv.DictReader(handle)38            fields = reader.fieldnames or []39            if {"StudyOID", "SubjectKey", "ItemOID", "Value"} <= set(fields):40                return list(reader)41    raise FileNotFoundError("No ODM extract found in " + input_dir)424344def group_forms(rows):45    """Group ODM rows into form occurrences keyed by their identity.4647    Returns (ae_forms, fa_forms): lists of dicts with the occurrence48    keys and an ``items`` mapping of ItemOID to collected value.49    """50    forms = {}51    order = []52    for row in rows:53        key = (54            row["StudyOID"], row["SubjectKey"], row["StudyEventOID"],55            row["StudyEventRepeatKey"], row["ItemGroupOID"],56            row["ItemGroupRepeatKey"],57        )58        if key not in forms:59            forms[key] = {60                "study": row["StudyOID"],61                "subject": row["SubjectKey"],62                "event": row["StudyEventOID"],63                "event_repeat": row["StudyEventRepeatKey"],64                "group": row["ItemGroupOID"],65                "group_repeat": row["ItemGroupRepeatKey"],66                "items": {},67                "position": len(order),68            }69            order.append(key)70        forms[key]["items"][row["ItemOID"]] = row["Value"]71    ae_forms = [forms[k] for k in order if forms[k]["group"] == "IG.AE"]72    fa_forms = [forms[k] for k in order if forms[k]["group"] == "IG.FA"]73    return ae_forms, fa_forms747576def as_number(text):77    """Format a collected measurement as a plain number string."""78    value = float(text)79    if value == int(value):80        return str(int(value))81    return str(value)828384def main():85    rows = read_odm(INPUT_DIR)86    ae_forms, fa_forms = group_forms(rows)8788    # Index AE occurrences by subject and event occurrence so each89    # findings form can read the term of its linked event.90    ae_by_event = {}91    for form in ae_forms:92        key = (form["study"], form["subject"], form["event"],93               form["event_repeat"])94        ae_by_event[key] = form9596    records = []97    for form in fa_forms:98        items = form["items"]99        lnkid = items.get("IT.FA.AELNKID", "")100        ae_key = (form["study"], form["subject"], form["event"],101                  form["event_repeat"])102        ae_form = ae_by_event.get(ae_key)103        faobj = ae_form["items"].get("IT.AE.AETERM", "") if ae_form else ""104        event_order = ae_form["position"] if ae_form else form["position"]105        fadtc = items.get("IT.FA.FADTC", "")106        size_unit = items.get("IT.FA.SIZEU", "")107108        answers = []109        location = items.get("IT.FA.LOCATION", "")110        if location != "":111            answers.append(("LOC", location, "", "", ""))112        size = items.get("IT.FA.SIZE", "")113        if size != "":114            answers.append(("SIZE", size, size_unit, as_number(size),115                            size_unit))116        biopsy = items.get("IT.FA.BIOPSY", "")117        if biopsy != "":118            answers.append(("BIOPSY", biopsy, "", "", ""))119120        for testcd, orres, orresu, stresn, stresu in answers:121            records.append({122                "subject": form["subject"],123                "event_order": event_order,124                "test_order": TEST_ORDER[testcd],125                "tiebreak": orres,126                "row": {127                    "DOMAIN": "FA",128                    "STUDYID": form["study"],129                    "USUBJID": form["subject"],130                    "FASEQ": "",131                    "FATESTCD": testcd,132                    "FATEST": TEST_NAMES[testcd],133                    "FAOBJ": faobj,134                    "FACAT": "AE",135                    "FAORRES": orres,136                    "FAORRESU": orresu,137                    "FASTRESC": orres,138                    "FASTRESN": stresn,139                    "FASTRESU": stresu,140                    "FALNKID": lnkid,141                    "FADTC": fadtc,142                },143            })144145    # Subject order follows first appearance in the collected data.146    subject_order = {}147    for row in rows:148        if row["SubjectKey"] not in subject_order:149            subject_order[row["SubjectKey"]] = len(subject_order)150151    records.sort(key=lambda r: (subject_order[r["subject"]],152                                r["event_order"], r["test_order"],153                                r["tiebreak"]))154    counters = {}155    for record in records:156        subject = record["subject"]157        counters[subject] = counters.get(subject, 0) + 1158        record["row"]["FASEQ"] = str(counters[subject])159160    os.makedirs(OUTPUT_DIR, exist_ok=True)161    out_path = os.path.join(OUTPUT_DIR, "fa.csv")162    with open(out_path, "w", newline="", encoding="utf-8") as handle:163        writer = csv.DictWriter(handle, fieldnames=COLUMNS)164        writer.writeheader()165        for record in records:166            writer.writerow(record["row"])167168169if __name__ == "__main__":170    main()

1/* opensas reference solution for the yamaa benchmark sdtm-fa-event-findings.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 StudyEventRepeatKey ItemGroupRepeatKey; run;12data events;13  length FAOBJ AELNKID $1024;14  retain FAOBJ AELNKID;15  set ordered;16  where ItemGroupOID='IG.AE';17  by StudyOID SubjectKey StudyEventOID StudyEventRepeatKey ItemGroupRepeatKey;18  if first.ItemGroupRepeatKey then do;19    FAOBJ=''; AELNKID='';20  end;21    if ItemOID='IT.AE.AETERM' then FAOBJ=Value;22  if ItemOID='IT.AE.AELNKID' then AELNKID=Value;23  if last.ItemGroupRepeatKey then output;24run;25proc sort data=odm out=ordered; by StudyOID SubjectKey StudyEventOID StudyEventRepeatKey ItemGroupRepeatKey; run;26data findings;27  length FALNKID LOC SIZE SIZEU BIOPSY FADTC $1024;28  retain FALNKID LOC SIZE SIZEU BIOPSY FADTC;29  set ordered;30  where ItemGroupOID='IG.FA';31  by StudyOID SubjectKey StudyEventOID StudyEventRepeatKey ItemGroupRepeatKey;32  if first.ItemGroupRepeatKey then do;33    FALNKID=''; LOC=''; SIZE=''; SIZEU=''; BIOPSY=''; FADTC='';34  end;35    if ItemOID='IT.FA.AELNKID' then FALNKID=Value;36  if ItemOID='IT.FA.LOCATION' then LOC=Value;37  if ItemOID='IT.FA.SIZE' then SIZE=Value;38  if ItemOID='IT.FA.SIZEU' then SIZEU=Value;39  if ItemOID='IT.FA.BIOPSY' then BIOPSY=Value;40  if ItemOID='IT.FA.FADTC' then FADTC=Value;41  if last.ItemGroupRepeatKey then output;42run;43proc sql;44create table joined as select a.*,b.FAOBJ from findings a inner join events b on a.StudyOID=b.StudyOID and a.SubjectKey=b.SubjectKey and a.FALNKID=b.AELNKID; quit;45data prepared; set joined; length DOMAIN $2 STUDYID USUBJID FATESTCD FATEST FACAT FAORRES FAORRESU FASTRESC FASTRESU $1024;46DOMAIN='FA'; STUDYID=StudyOID; USUBJID=SubjectKey; FACAT='AE'; FAORRESU=''; FASTRESU=''; FASTRESN=.; EVENTREP=input(ItemGroupRepeatKey,best32.);47if not missing(LOC) then do; FATESTCD='LOC'; FATEST='Location'; FAORRES=LOC; FASTRESC=FAORRES; TESTORD=1; output; end;48if not missing(SIZE) then do; FATESTCD='SIZE'; FATEST='Size'; FAORRES=SIZE; FASTRESC=FAORRES; FAORRESU=SIZEU; FASTRESU=SIZEU; FASTRESN=input(SIZE,best32.); TESTORD=2; output; end;49if not missing(BIOPSY) then do; FATESTCD='BIOPSY'; FATEST='Biopsied'; FAORRES=BIOPSY; FASTRESC=FAORRES; FAORRESU=''; FASTRESU=''; FASTRESN=.; TESTORD=3; output; end; run;50proc sort data=prepared; by STUDYID USUBJID FALNKID TESTORD FAORRES; run;51data result; set prepared; by STUDYID USUBJID; retain FASEQ;52  if first.USUBJID then FASEQ=0;53  FASEQ+1;54run;5556proc sql;57  create table final as select DOMAIN, STUDYID, USUBJID, FASEQ, FATESTCD, FATEST, FAOBJ, FACAT, FAORRES, FAORRESU, FASTRESC, FASTRESN, FASTRESU, FALNKID, FADTC from result;58quit;59proc export data=final outfile="/app/output/fa.csv" dbms=csv replace; run;