Derive the new anti-cancer therapy start date

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

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

1# Reference solution for the yamaa benchmark adam-adsl-new-therapy (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)910adsl_raw <- read_csv(11  "/app/input/adsl.csv",12  col_types = cols(TRTSDT = col_date(), .default = col_character())13)14cm <- read_csv(15  "/app/input/cm.csv",16  col_types = cols(CMSTDTC_IMP = col_date(), .default = col_character())17)18pr <- read_csv(19  "/app/input/pr.csv",20  col_types = cols(PRSTDTC = col_date(), .default = col_character())21)2223# A qualifying medication starts on or after treatment start; a record24# with no start date is skipped. The completed start fills a year-month25# to the 15th, so the IMP column is the comparison date.26cm_min <- cm |>27  inner_join(select(adsl_raw, STUDYID, USUBJID, TRTSDT), by = c("STUDYID", "USUBJID")) |>28  filter(CMCAT == "ON TREATMENT", !is.na(CMSTDTC_IMP), CMSTDTC_IMP >= TRTSDT) |>29  group_by(STUDYID, USUBJID) |>30  summarise(CMNACTDT = min(CMSTDTC_IMP), .groups = "drop")3132# A qualifying procedure is cancer related and on treatment, also on or33# after treatment start.34pr_min <- pr |>35  inner_join(select(adsl_raw, STUDYID, USUBJID, TRTSDT), by = c("STUDYID", "USUBJID")) |>36  filter(37    PRCAT == "CANCER RELATED", PRSCAT == "ON TREATMENT",38    !is.na(PRSTDTC), PRSTDTC >= TRTSDT39  ) |>40  group_by(STUDYID, USUBJID) |>41  summarise(PRNACTDT = min(PRSTDTC), .groups = "drop")4243adsl <- adsl_raw |>44  left_join(cm_min, by = c("STUDYID", "USUBJID")) |>45  left_join(pr_min, by = c("STUDYID", "USUBJID")) |>46  mutate(47    NACTDT = case_when(48      !is.na(CMNACTDT) & !is.na(PRNACTDT) ~ pmin(CMNACTDT, PRNACTDT),49      !is.na(CMNACTDT) ~ CMNACTDT,50      !is.na(PRNACTDT) ~ PRNACTDT,51      .default = as.Date(NA)52    )53  ) |>54  mutate(55    NACTDY = as.integer(NACTDT - TRTSDT) + 1L,56    NACTFL = if_else(is.na(NACTDT), NA_character_, "Y")57  ) |>58  select(STUDYID, USUBJID, TRTSDT, NACTDT, NACTDY, NACTFL) |>59  arrange(USUBJID)6061dir.create("/app/output", showWarnings = FALSE)62write_csv(adsl, "/app/output/adsl.csv", na = "")

1# Reference solution for the yamaa benchmark adam-adsl-new-therapy (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"""Create an ADSL dataset with new anticancer therapy variables.89Reads the ADSL, CM and PR datasets from /app/input and writes10/app/output/adsl.csv with one record per subject and the columns11STUDYID, USUBJID, TRTSDT, NACTDT, NACTDY, NACTFL.1213A medication record qualifies when its CMCAT is "ON TREATMENT". A14procedure record qualifies when its PRCAT is "CANCER RELATED" and its15PRSCAT is "ON TREATMENT". A qualifying record counts only when its16start date is on or after TRTSDT; a qualifying record with no start17date does not count. A start recorded as a year and month only is18imputed to the 15th of that month; a full recorded start date stands19as recorded.2021NACTDT is the earlier of the subject's earliest qualifying medication22start date and earliest qualifying procedure start date, and has no23value when no qualifying record counts. NACTDY is the study day of24NACTDT counted from TRTSDT (TRTSDT itself is day 1). NACTFL is "Y"25when NACTDT has a value and has no value otherwise.26"""2728import csv29from datetime import date30from pathlib import Path3132INPUT_DIR = Path("/app/input")33OUTPUT_DIR = Path("/app/output")343536def parse_start_date(value):37    """Return a date from an ISO 8601 date string, or None.3839    A full date (YYYY-MM-DD) stands as recorded. A year-and-month40    value (YYYY-MM) is imputed to the 15th of that month. A missing41    value returns None.42    """43    value = (value or "").strip()44    if not value:45        return None46    parts = value.split("-")47    if len(parts) == 3:48        year, month, day = (int(part) for part in parts)49        return date(year, month, day)50    if len(parts) == 2:51        year, month = (int(part) for part in parts)52        return date(year, month, 15)53    return None545556def read_csv(path):57    with open(path, newline="", encoding="utf-8") as handle:58        return list(csv.DictReader(handle))596061def main():62    adsl = read_csv(INPUT_DIR / "adsl.csv")63    cm = read_csv(INPUT_DIR / "cm.csv")64    pr = read_csv(INPUT_DIR / "pr.csv")6566    trtsdt_by_subject = {67        row["USUBJID"]: parse_start_date(row["TRTSDT"]) for row in adsl68    }6970    # Earliest counting start date per subject across both sources.71    earliest = {}7273    def consider(usubjid, start):74        if start is None:75            return76        trtsdt = trtsdt_by_subject.get(usubjid)77        if trtsdt is None or start < trtsdt:78            return79        if usubjid not in earliest or start < earliest[usubjid]:80            earliest[usubjid] = start8182    for row in cm:83        if (row.get("CMCAT") or "").strip() != "ON TREATMENT":84            continue85        consider(row["USUBJID"], parse_start_date(row.get("CMSTDTC", "")))8687    for row in pr:88        if (row.get("PRCAT") or "").strip() != "CANCER RELATED":89            continue90        if (row.get("PRSCAT") or "").strip() != "ON TREATMENT":91            continue92        consider(row["USUBJID"], parse_start_date(row.get("PRSTDTC", "")))9394    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)95    fieldnames = ["STUDYID", "USUBJID", "TRTSDT", "NACTDT", "NACTDY", "NACTFL"]96    with open(OUTPUT_DIR / "adsl.csv", "w", newline="", encoding="utf-8") as handle:97        writer = csv.DictWriter(handle, fieldnames=fieldnames)98        writer.writeheader()99        for row in adsl:100            usubjid = row["USUBJID"]101            trtsdt = trtsdt_by_subject[usubjid]102            nactdt = earliest.get(usubjid)103            if nactdt is None:104                nactdt_text, nactdy_text, nactfl = "", "", ""105            else:106                nactdt_text = nactdt.isoformat()107                nactdy_text = str((nactdt - trtsdt).days + 1)108                nactfl = "Y"109            writer.writerow(110                {111                    "STUDYID": row["STUDYID"],112                    "USUBJID": usubjid,113                    "TRTSDT": row["TRTSDT"],114                    "NACTDT": nactdt_text,115                    "NACTDY": nactdy_text,116                    "NACTFL": nactfl,117                }118            )119120121if __name__ == "__main__":122    main()

1/* opensas reference solution for the yamaa benchmark adam-adsl-new-therapy.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data adsl;6  length STUDYID USUBJID TRTSDT $1024;7  infile "/app/input/adsl.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024. TRTSDT : $char1024.;9run;1011data cm;12  length STUDYID USUBJID CMCAT CMSCAT CMTRT CMSTDTC CMSTDTC_IMP $1024;13  infile "/app/input/cm.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;14  input STUDYID : $char1024. USUBJID : $char1024. CMSEQ CMCAT : $char1024. CMSCAT : $char1024. CMTRT : $char1024. CMSTDTC : $char1024. CMSTDTC_IMP : $char1024.;15run;1617data pr;18  length STUDYID USUBJID PRCAT PRSCAT PRTRT PRSTDTC $1024;19  infile "/app/input/pr.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;20  input STUDYID : $char1024. USUBJID : $char1024. PRSEQ PRCAT : $char1024. PRSCAT : $char1024. PRTRT : $char1024. PRSTDTC : $char1024.;21run;2223data medication;24  length therapy $10;25  set cm;26  if CMCAT='ON TREATMENT';27  /* Use the supplied completed date, including its collected corrections. */28  therapy='';29  if lengthn(CMSTDTC_IMP)>=10 then therapy=substr(CMSTDTC_IMP,1,10);30run;31data procedure;32  length therapy $10;33  set pr;34  if PRCAT='CANCER RELATED' and PRSCAT='ON TREATMENT';35  therapy=substr(PRSTDTC,1,10);36run;37data therapies; set medication procedure; keep STUDYID USUBJID therapy; run;38proc sql;39  create table firsttherapy as select a.STUDYID,a.USUBJID,min(b.therapy) as NACTDT from adsl as a inner join therapies as b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID where a.TRTSDT ne '' and b.therapy ne '' and b.therapy>=a.TRTSDT group by a.STUDYID,a.USUBJID;40  create table joined as select a.*,b.NACTDT from adsl as a left join firsttherapy as b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID;41quit;42data result;43  length NACTFL $1; set joined;44  NACTFL=''; NACTDY=.;45  if not missing(NACTDT) then do; NACTFL='Y'; NACTDY=input(NACTDT,yymmdd10.)-input(TRTSDT,yymmdd10.)+1; end;46run;4748proc sql;49  create table final as select STUDYID, USUBJID, TRTSDT, NACTDT, NACTDY, NACTFL from result;50quit;51proc export data=final outfile="/app/output/adsl.csv" dbms=csv replace; run;