Subject elements from DM and an ODM extract

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

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

1# Reference solution for the yamaa benchmark sdtm-se-subject-elements (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)1011dm <- read_csv(12  "/app/input/dm.csv",13  col_types = cols(.default = col_character())14)15odm <- read_csv(16  "/app/input/odm.csv",17  col_types = cols(.default = col_character())18)1920wide <- odm |>21  mutate(Value = na_if(Value, "")) |>22  pivot_wider(23    id_cols = c(SubjectKey, StudyEventOID),24    names_from = ItemOID,25    values_from = Value,26    values_fn = function(x) x[[1]]27  )2829study_day <- function(ref, day) {30  ifelse(31    is.na(ref) | ref == "" | is.na(day) | day == "",32    NA_character_,33    {34      delta <- as.integer(as.Date(day) - as.Date(ref))35      as.character(ifelse(delta >= 0, delta + 1L, delta))36    }37  )38}3940se <- wide |>41  left_join(dm, by = c("SubjectKey" = "USUBJID")) |>42  mutate(43    DOMAIN = "SE",44    STUDYID = STUDYID,45    USUBJID = SubjectKey,46    SESEQ = IT.TA.TAETORD,47    ETCD = StudyEventOID,48    ELEMENT = IT.TE.ELEMENT,49    TAETORD = IT.TA.TAETORD,50    EPOCH = IT.TA.EPOCH,51    first_dtc = IT.EX.FIRSTDTC,52    last_dtc = IT.EX.LASTDTC,53    SESTDTC = case_when(54      EPOCH == "SCREENING" ~ RFICDTC,55      EPOCH == "TREATMENT" ~ first_dtc,56      TRUE ~ last_dtc57    ),58    SEENDTC = case_when(59      EPOCH == "SCREENING" ~ first_dtc,60      EPOCH == "TREATMENT" ~ last_dtc,61      TRUE ~ RFPENDTC62    ),63    SESTDY = study_day(RFSTDTC, SESTDTC),64    SEENDY = study_day(RFSTDTC, SEENDTC),65    SEUPDES = NA_character_66  ) |>67  arrange(USUBJID, suppressWarnings(as.integer(TAETORD))) |>68  select(69    DOMAIN, STUDYID, USUBJID, SESEQ, ETCD, ELEMENT, TAETORD, EPOCH,70    SESTDTC, SEENDTC, SESTDY, SEENDY, SEUPDES71  )7273dir.create("/app/output", showWarnings = FALSE)74write_csv(se, "/app/output/se.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-se-subject-elements (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 SE (Subject Elements) dataset from DM and ODM extracts.89One record per subject per element the subject entered. Dates and study10days follow the epoch-based rules described in the task.11"""1213import csv14from datetime import date1516INPUT_DIR = "/app/input"17OUTPUT_DIR = "/app/output"181920def parse_date(value):21    """Parse an ISO date (or the date part of an ISO datetime).2223    Returns a datetime.date, or None when there is no usable value.24    """25    if value is None:26        return None27    text = str(value).strip()28    if not text:29        return None30    try:31        return date.fromisoformat(text[:10])32    except ValueError:33        return None343536def study_day(target, reference):37    """SDTM study day of target counted from the reference start date.3839    No value when either date has no value. There is no day zero: the40    reference date itself is day 1 and the day before it is day -1.41    """42    if target is None or reference is None:43        return ""44    diff = (target - reference).days45    return diff + 1 if diff >= 0 else diff464748def main():49    # DM: one row per subject with the dates the SE rules draw on.50    dm = {}51    with open(f"{INPUT_DIR}/dm.csv", newline="", encoding="utf-8") as handle:52        for row in csv.DictReader(handle):53            dm[row["USUBJID"]] = {54                "rficdtc": (row.get("RFICDTC") or "").strip(),55                "rfstdtc": parse_date(row.get("RFSTDTC")),56                "rfpendtc": (row.get("RFPENDTC") or "").strip(),57            }5859    # ODM extract: one element record per subject per study event, plus60    # the subject's first and last dosing dates (the same on every event).61    elements = {}62    subject_order = []63    with open(f"{INPUT_DIR}/odm.csv", newline="", encoding="utf-8") as handle:64        for row in csv.DictReader(handle):65            subject = row["SubjectKey"]66            if subject not in elements:67                elements[subject] = {}68                subject_order.append(subject)69            record = elements[subject].setdefault(70                row["StudyEventOID"],71                {72                    "studyid": row["StudyOID"],73                    "usubjid": subject,74                    "etcd": row["StudyEventOID"],75                    "element": "",76                    "taetord": "",77                    "epoch": "",78                    "firstdtc": "",79                    "lastdtc": "",80                },81            )82            item = row["ItemOID"]83            value = (row.get("Value") or "").strip()84            if item == "IT.TE.ELEMENT":85                record["element"] = value86            elif item == "IT.TA.TAETORD":87                record["taetord"] = value88            elif item == "IT.TA.EPOCH":89                record["epoch"] = value90            elif item == "IT.EX.FIRSTDTC":91                record["firstdtc"] = value92            elif item == "IT.EX.LASTDTC":93                record["lastdtc"] = value9495    rows = []96    for subject in subject_order:97        subject_dm = dm.get(subject, {})98        reference = subject_dm.get("rfstdtc")99        records = sorted(100            elements[subject].values(),101            key=lambda rec: (int(rec["taetord"]) if rec["taetord"] else 0),102        )103        for record in records:104            epoch = record["epoch"]105            if epoch == "SCREENING":106                sestdtc = subject_dm.get("rficdtc", "")107                seendtc = record["firstdtc"]108            elif epoch == "TREATMENT":109                sestdtc = record["firstdtc"]110                seendtc = record["lastdtc"]111            else:112                sestdtc = record["lastdtc"]113                seendtc = subject_dm.get("rfpendtc", "")114            rows.append(115                {116                    "DOMAIN": "SE",117                    "STUDYID": record["studyid"],118                    "USUBJID": record["usubjid"],119                    "SESEQ": record["taetord"],120                    "ETCD": record["etcd"],121                    "ELEMENT": record["element"],122                    "TAETORD": record["taetord"],123                    "EPOCH": epoch,124                    "SESTDTC": sestdtc,125                    "SEENDTC": seendtc,126                    "SESTDY": study_day(parse_date(sestdtc), reference),127                    "SEENDY": study_day(parse_date(seendtc), reference),128                    "SEUPDES": "",129                }130            )131132    columns = [133        "DOMAIN",134        "STUDYID",135        "USUBJID",136        "SESEQ",137        "ETCD",138        "ELEMENT",139        "TAETORD",140        "EPOCH",141        "SESTDTC",142        "SEENDTC",143        "SESTDY",144        "SEENDY",145        "SEUPDES",146    ]147    with open(f"{OUTPUT_DIR}/se.csv", "w", newline="", encoding="utf-8") as handle:148        writer = csv.DictWriter(handle, fieldnames=columns)149        writer.writeheader()150        writer.writerows(rows)151152153if __name__ == "__main__":154    main()

1/* opensas reference solution for the yamaa benchmark sdtm-se-subject-elements.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data dm;6  length STUDYID USUBJID ARMCD RFICDTC RFSTDTC RFPENDTC $1024;7  infile "/app/input/dm.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024. ARMCD : $char1024. RFICDTC : $char1024. RFSTDTC : $char1024. RFPENDTC : $char1024.;9run;1011data odm;12  length StudyOID MetaDataVersionOID SubjectKey StudyEventOID StudyEventRepeatKey ItemGroupOID ItemGroupRepeatKey ItemOID Value $1024;13  infile "/app/input/odm.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;14  input StudyOID : $char1024. MetaDataVersionOID : $char1024. SubjectKey : $char1024. StudyEventOID : $char1024. StudyEventRepeatKey : $char1024. ItemGroupOID : $char1024. ItemGroupRepeatKey : $char1024. ItemOID : $char1024. Value : $char1024.;15run;1617proc sort data=odm out=ordered; by StudyOID SubjectKey StudyEventOID; run;18data elements;19  length ELEMENT ORD EPOCH $1024;20  retain ELEMENT ORD EPOCH;21  set ordered;2223  by StudyOID SubjectKey StudyEventOID;24  if first.StudyEventOID then do;25    ELEMENT=''; ORD=''; EPOCH='';26  end;27    if ItemOID='IT.TE.ELEMENT' then ELEMENT=Value;28  if ItemOID='IT.TA.TAETORD' then ORD=Value;29  if ItemOID='IT.TA.EPOCH' then EPOCH=Value;30  if last.StudyEventOID then output;31run;32proc sort data=odm out=ordered; by StudyOID SubjectKey; run;33data dosing;34  length FIRSTDTC LASTDTC $1024;35  retain FIRSTDTC LASTDTC;36  set ordered;3738  by StudyOID SubjectKey;39  if first.SubjectKey then do;40    FIRSTDTC=''; LASTDTC='';41  end;42    if ItemOID='IT.EX.FIRSTDTC' then FIRSTDTC=Value;43  if ItemOID='IT.EX.LASTDTC' then LASTDTC=Value;44  if last.SubjectKey then output;45run;46proc sql; create table joined as select a.*,b.FIRSTDTC,b.LASTDTC,c.RFICDTC,c.RFSTDTC,c.RFPENDTC from elements a left join dosing b on a.StudyOID=b.StudyOID and a.SubjectKey=b.SubjectKey inner join dm c on a.StudyOID=c.STUDYID and a.SubjectKey=c.USUBJID where not missing(a.ELEMENT); quit;47data result; set joined; length DOMAIN $2 STUDYID USUBJID ETCD SESTDTC SEENDTC SEUPDES $1024; DOMAIN='SE'; STUDYID=StudyOID; USUBJID=SubjectKey; ETCD=StudyEventOID; TAETORD=input(ORD,best32.); SESEQ=TAETORD; SEUPDES=''; SESTDY=.; SEENDY=.;48if EPOCH='SCREENING' then do; SESTDTC=RFICDTC; SEENDTC=FIRSTDTC; end; else if EPOCH='TREATMENT' then do; SESTDTC=FIRSTDTC; SEENDTC=LASTDTC; end; else do; SESTDTC=LASTDTC; SEENDTC=RFPENDTC; end;49if not missing(RFSTDTC) then do; if not missing(SESTDTC) then do; SESTDY=input(SESTDTC,yymmdd10.)-input(RFSTDTC,yymmdd10.); if SESTDY>=0 then SESTDY+1; end; if not missing(SEENDTC) then do; SEENDY=input(SEENDTC,yymmdd10.)-input(RFSTDTC,yymmdd10.); if SEENDY>=0 then SEENDY+1; end; end; run;50proc sql;51  create table final as select DOMAIN, STUDYID, USUBJID, SESEQ, ETCD, ELEMENT, TAETORD, EPOCH, SESTDTC, SEENDTC, SESTDY, SEENDY, SEUPDES from result;52quit;53proc export data=final outfile="/app/output/se.csv" dbms=csv replace; run;