Partial Adverse Event Dates

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

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

1# Reference solution for the yamaa benchmark sdtm-ae-partial-dates (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(stringr)1112odm <- read_csv(13  "/app/input/odm.csv",14  col_types = cols(.default = col_character())15)16dm <- read_csv(17  "/app/input/dm.csv",18  col_types = cols(.default = col_character())19)2021compose_dtc <- function(yr, mo, dy) {22  ifelse(23    is.na(yr) | yr == "", NA_character_,24    ifelse(25      is.na(mo) | mo == "", yr,26      ifelse(27        is.na(dy) | dy == "", paste0(yr, "-", str_pad(mo, 2, pad = "0")),28        paste0(yr, "-", str_pad(mo, 2, pad = "0"), "-", str_pad(dy, 2, pad = "0"))29      )30    )31  )32}3334study_day <- function(adt, ref) {35  diff <- as.integer(as.Date(adt) - as.Date(ref))36  ifelse(is.na(diff), NA_integer_, ifelse(diff >= 0L, diff + 1L, diff))37}3839# One record per form repeat; the repeat number is the sequence number.40# Dates stay at the precision collected: nothing is imputed.41ae <- odm |>42  filter(ItemGroupOID == "IG.AE") |>43  pivot_wider(44    id_cols = c(StudyOID, SubjectKey, FormRepeatKey),45    names_from = ItemOID,46    values_from = Value,47    values_fn = first48  ) |>49  mutate(50    AESEQ = as.integer(FormRepeatKey),51    AETERM = `IT.AE.AETERM`,52    AESTDTC = compose_dtc(`IT.AE.AESTYR`, `IT.AE.AESTMO`, `IT.AE.AESTDY`),53    AEENDTC = compose_dtc(`IT.AE.AEENYR`, `IT.AE.AEENMO`, `IT.AE.AEENDY`),54    START_FULL = !is.na(`IT.AE.AESTYR`) & `IT.AE.AESTYR` != "" &55      !is.na(`IT.AE.AESTMO`) & `IT.AE.AESTMO` != "" &56      !is.na(`IT.AE.AESTDY`) & `IT.AE.AESTDY` != "",57    END_FULL = !is.na(`IT.AE.AEENYR`) & `IT.AE.AEENYR` != "" &58      !is.na(`IT.AE.AEENMO`) & `IT.AE.AEENMO` != "" &59      !is.na(`IT.AE.AEENDY`) & `IT.AE.AEENDY` != ""60  ) |>61  left_join(62    dm |> select(USUBJID, RFSTDTC),63    by = c("SubjectKey" = "USUBJID")64  ) |>65  # Only a complete date gets a study day, and only when the reference66  # start date is known. Day 1 is the reference date; there is no day 0.67  mutate(68    AESTDY = if_else(69      START_FULL & !is.na(RFSTDTC) & RFSTDTC != "",70      study_day(AESTDTC, RFSTDTC), NA_integer_71    ),72    AEENDY = if_else(73      END_FULL & !is.na(RFSTDTC) & RFSTDTC != "",74      study_day(AEENDTC, RFSTDTC), NA_integer_75    ),76    DOMAIN = "AE",77    STUDYID = StudyOID,78    USUBJID = SubjectKey79  ) |>80  arrange(USUBJID, AESEQ) |>81  select(82    DOMAIN, STUDYID, USUBJID, AESEQ, AETERM,83    AESTDTC, AEENDTC, AESTDY, AEENDY84  )8586dir.create("/app/output", showWarnings = FALSE)87write_csv(ae, "/app/output/ae.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-ae-partial-dates (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 DM and ODM long-format inputs.89One record per adverse event (one AE form repeat per subject). Start and10end dates are assembled from their separately collected year, month, and11day items and written as ISO 8601 text at the precision collected; nothing12is imputed. Study days are computed only for complete dates and only when13the subject has a reference start date (RFSTDTC) in DM.14"""1516import csv17from datetime import date1819INPUT_DIR = "/app/input"20OUTPUT_PATH = "/app/output/ae.csv"2122COLUMNS = [23    "DOMAIN",24    "STUDYID",25    "USUBJID",26    "AESEQ",27    "AETERM",28    "AESTDTC",29    "AEENDTC",30    "AESTDY",31    "AEENDY",32]333435def read_csv(path):36    with open(path, newline="", encoding="utf-8") as handle:37        return list(csv.DictReader(handle))383940def assemble_date(parts, prefix):41    """Join collected year/month/day items at the precision collected."""42    year = parts.get(prefix + "YR", "").strip()43    if not year:44        return ""45    month = parts.get(prefix + "MO", "").strip()46    if not month:47        return year48    day = parts.get(prefix + "DY", "").strip()49    if not day:50        return f"{year}-{int(month):02d}"51    return f"{year}-{int(month):02d}-{int(day):02d}"525354def parse_full_date(text):55    """Return a date for a complete ISO date string, else None."""56    if len(text) != 10:57        return None58    try:59        year, month, day = (int(piece) for piece in text.split("-"))60        return date(year, month, day)61    except ValueError:62        return None636465def study_day(date_text, ref_start):66    """SDTM study day of a complete date relative to the reference start.6768    The reference start date itself is day 1; the day before it is day -169    (there is no day 0). Partial dates and missing reference dates have70    no study day.71    """72    value = parse_full_date(date_text)73    if value is None or ref_start is None:74        return ""75    delta = (value - ref_start).days76    return str(delta + 1 if delta >= 0 else delta)777879def main():80    dm_rows = read_csv(f"{INPUT_DIR}/dm.csv")81    ref_start_by_subject = {}82    for row in dm_rows:83        ref = parse_full_date((row.get("RFSTDTC") or "").strip())84        ref_start_by_subject[row["USUBJID"]] = ref8586    odm_rows = read_csv(f"{INPUT_DIR}/odm.csv")8788    # Group the AE form items by subject and form repeat key (AESEQ),89    # keeping first-appearance order of subjects and events.90    events = {}91    order = []92    for row in odm_rows:93        if row["FormOID"] != "FO.AE":94            continue95        key = (row["StudyOID"], row["SubjectKey"], row["FormRepeatKey"])96        if key not in events:97            events[key] = {}98            order.append(key)99        item = row["ItemOID"].rsplit(".", 1)[-1]100        events[key][item] = row["Value"]101102    out_rows = []103    for study_oid, subject_key, repeat_key in order:104        items = events[(study_oid, subject_key, repeat_key)]105        start_dtc = assemble_date(items, "AEST")106        end_dtc = assemble_date(items, "AEEN")107        ref_start = ref_start_by_subject.get(subject_key)108        out_rows.append(109            {110                "DOMAIN": "AE",111                "STUDYID": study_oid,112                "USUBJID": subject_key,113                "AESEQ": repeat_key,114                "AETERM": items.get("AETERM", ""),115                "AESTDTC": start_dtc,116                "AEENDTC": end_dtc,117                "AESTDY": study_day(start_dtc, ref_start),118                "AEENDY": study_day(end_dtc, ref_start),119            }120        )121122    with open(OUTPUT_PATH, "w", newline="", encoding="utf-8") as handle:123        writer = csv.DictWriter(handle, fieldnames=COLUMNS)124        writer.writeheader()125        writer.writerows(out_rows)126127128if __name__ == "__main__":129    main()

1/* opensas reference solution for the yamaa benchmark sdtm-ae-partial-dates.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 RFSTDTC $1024;7  infile "/app/input/dm.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024. RFSTDTC : $char1024.;9run;1011data odm;12  length StudyOID MetaDataVersionOID SubjectKey StudyEventOID StudyEventRepeatKey FormOID FormRepeatKey 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. FormOID : $char1024. FormRepeatKey : $char1024. ItemGroupOID : $char1024. ItemGroupRepeatKey : $char1024. ItemOID : $char1024. Value : $char1024.;15run;1617proc sort data=odm out=ordered; by StudyOID SubjectKey FormRepeatKey; run;18data wide;19  length AETERM AESTYR AESTMO AEENYR AEENMO STARTDAY ENDDAY $1024;20  retain AETERM AESTYR AESTMO AEENYR AEENMO STARTDAY ENDDAY;21  set ordered;22  where ItemGroupOID='IG.AE';23  by StudyOID SubjectKey FormRepeatKey;24  if first.FormRepeatKey then do;25    AETERM=''; AESTYR=''; AESTMO=''; AEENYR=''; AEENMO=''; STARTDAY=''; ENDDAY='';26  end;27    if ItemOID='IT.AE.AETERM' then AETERM=Value;28  if ItemOID='IT.AE.AESTYR' then AESTYR=Value;29  if ItemOID='IT.AE.AESTMO' then AESTMO=Value;30  if ItemOID='IT.AE.AEENYR' then AEENYR=Value;31  if ItemOID='IT.AE.AEENMO' then AEENMO=Value;32  if ItemOID='IT.AE.AESTDY' then STARTDAY=Value;33  if ItemOID='IT.AE.AEENDY' then ENDDAY=Value;34  if last.FormRepeatKey then output;35run;36data dates; set wide; length DOMAIN $2 STUDYID USUBJID AESTDTC AEENDTC $1024;37DOMAIN='AE'; STUDYID=StudyOID; USUBJID=SubjectKey; AESEQ=input(FormRepeatKey,best32.);38AESTDTC=''; AEENDTC='';39if not missing(AESTYR) then do; AESTDTC=AESTYR; if not missing(AESTMO) then do; AESTDTC=catx('-',AESTDTC,AESTMO); if not missing(STARTDAY) then AESTDTC=catx('-',AESTDTC,STARTDAY); end; end;40if not missing(AEENYR) then do; AEENDTC=AEENYR; if not missing(AEENMO) then do; AEENDTC=catx('-',AEENDTC,AEENMO); if not missing(ENDDAY) then AEENDTC=catx('-',AEENDTC,ENDDAY); end; end; run;41proc sql; create table joined as select a.*,b.RFSTDTC from dates a left join dm b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID; quit;42data result; set joined; AESTDY=.; AEENDY=.; if lengthn(RFSTDTC)=10 then do;43if lengthn(AESTDTC)=10 then do; AESTDY=input(AESTDTC,yymmdd10.)-input(RFSTDTC,yymmdd10.); if AESTDY>=0 then AESTDY+1; end;44if lengthn(AEENDTC)=10 then do; AEENDY=input(AEENDTC,yymmdd10.)-input(RFSTDTC,yymmdd10.); if AEENDY>=0 then AEENDY+1; end; end; run;45proc sql;46  create table final as select DOMAIN, STUDYID, USUBJID, AESEQ, AETERM, AESTDTC, AEENDTC, AESTDY, AEENDY from result;47quit;48proc export data=final outfile="/app/output/ae.csv" dbms=csv replace; run;