Exposure from Collected Records

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

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

1# Reference solution for the yamaa benchmark sdtm-ex-from-ec (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)15kit <- read_csv(16  "/app/input/kit_list.csv",17  col_types = cols(.default = col_character())18)1920kit_trt <- setNames(kit$TREATMENT, kit$KIT)21kit_dose <- setNames(suppressWarnings(as.numeric(kit$DOSE_MG)), kit$KIT)2223ec_forms <- odm |> filter(FormOID == "FO.EC")24vs_weight <- odm |>25  filter(FormOID == "FO.VS", ItemOID == "IT.VS.WEIGHT") |>26  group_by(StudyOID, SubjectKey, StudyEventOID, StudyEventRepeatKey) |>27  summarise(WEIGHT = suppressWarnings(as.numeric(first(Value))), .groups = "drop")2829ec_wide <- ec_forms |>30  group_by(StudyOID, SubjectKey, StudyEventOID, StudyEventRepeatKey, FormRepeatKey) |>31  summarise(32    ECOCCUR = Value[ItemOID == "IT.EC.ECOCCUR"][1],33    ECTRT = Value[ItemOID == "IT.EC.ECTRT"][1],34    FORM = Value[ItemOID == "IT.EC.FORM"][1],35    TABLETS = suppressWarnings(as.numeric(Value[ItemOID == "IT.EC.TABLETS"][1])),36    STRENGTHMG = suppressWarnings(as.numeric(Value[ItemOID == "IT.EC.STRENGTHMG"][1])),37    DOSEMKG = suppressWarnings(as.numeric(Value[ItemOID == "IT.EC.DOSEMKG"][1])),38    KIT = Value[ItemOID == "IT.EC.KIT"][1],39    AUCTARGET = suppressWarnings(as.numeric(Value[ItemOID == "IT.EC.AUCTARGET"][1])),40    ECSTDTC = Value[ItemOID == "IT.EC.ECSTDTC"][1],41    ECENDTC = Value[ItemOID == "IT.EC.ECENDTC"][1],42    .groups = "drop"43  ) |>44  left_join(vs_weight, by = c("StudyOID", "SubjectKey", "StudyEventOID", "StudyEventRepeatKey")) |>45  filter(ECOCCUR == "Y")4647ex <- ec_wide |>48  rowwise() |>49  mutate(50    DOMAIN = "EX",51    STUDYID = StudyOID,52    USUBJID = SubjectKey,53    EXTRT = if (FORM == "KITDOSE") unname(kit_trt[KIT]) else ECTRT,54    EXDOSE = case_when(55      FORM == "TABLET" ~ TABLETS * STRENGTHMG,56      FORM == "INFUSION" ~ DOSEMKG * WEIGHT,57      FORM == "KITDOSE" ~ unname(kit_dose[KIT]),58      FORM == "AUCDOSE" ~ AUCTARGET,59      .default = NA_real_60    ),61    EXDOSU = ifelse(FORM == "AUCDOSE", "AUC", "mg"),62    FORMREPEAT = suppressWarnings(as.integer(FormRepeatKey)),63    EXSTDTC_DATE = suppressWarnings(as.Date(na_if(ECSTDTC, "")))64  ) |>65  ungroup() |>66  arrange(STUDYID, USUBJID, EXSTDTC_DATE, FORMREPEAT) |>67  group_by(STUDYID, USUBJID) |>68  mutate(EXSEQ = row_number()) |>69  ungroup() |>70  select(DOMAIN, STUDYID, USUBJID, EXSEQ, EXTRT, EXDOSE, EXDOSU, EXSTDTC = ECSTDTC, EXENDTC = ECENDTC)7172dir.create("/app/output", showWarnings = FALSE)73write_csv(ex, "/app/output/ex.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-ex-from-ec (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 1.67from pathlib import Path89import polars as pl1011odm = pl.read_csv("/app/input/odm.csv", infer_schema=False)12kit = pl.read_csv("/app/input/kit_list.csv", infer_schema=False)1314kit_map = {r["KIT"]: (r["TREATMENT"], r["DOSE_MG"]) for r in kit.to_dicts()}1516ec_forms = odm.filter(pl.col("FormOID") == "FO.EC")17vs_forms = odm.filter(pl.col("FormOID") == "FO.VS")1819# Body weight per visit: SubjectKey + StudyEventOID + StudyEventRepeatKey.20vs_wide = (21    vs_forms.filter(pl.col("ItemOID") == "IT.VS.WEIGHT")22    .group_by("StudyOID", "SubjectKey", "StudyEventOID", "StudyEventRepeatKey")23    .agg(pl.col("Value").first().alias("WEIGHT"))24    .with_columns(pl.col("WEIGHT").cast(pl.Float64, strict=False))25)2627ec_wide = (28    ec_forms.group_by(29        "StudyOID",30        "SubjectKey",31        "StudyEventOID",32        "StudyEventRepeatKey",33        "FormRepeatKey",34    )35    .agg(36        pl.col("Value").filter(pl.col("ItemOID") == "IT.EC.ECOCCUR").first().alias("ECOCCUR"),37        pl.col("Value").filter(pl.col("ItemOID") == "IT.EC.ECTRT").first().alias("ECTRT"),38        pl.col("Value").filter(pl.col("ItemOID") == "IT.EC.FORM").first().alias("FORM"),39        pl.col("Value").filter(pl.col("ItemOID") == "IT.EC.TABLETS").first().alias("TABLETS"),40        pl.col("Value").filter(pl.col("ItemOID") == "IT.EC.STRENGTHMG").first().alias("STRENGTHMG"),41        pl.col("Value").filter(pl.col("ItemOID") == "IT.EC.DOSEMKG").first().alias("DOSEMKG"),42        pl.col("Value").filter(pl.col("ItemOID") == "IT.EC.KIT").first().alias("KIT"),43        pl.col("Value").filter(pl.col("ItemOID") == "IT.EC.AUCTARGET").first().alias("AUCTARGET"),44        pl.col("Value").filter(pl.col("ItemOID") == "IT.EC.ECSTDTC").first().alias("ECSTDTC"),45        pl.col("Value").filter(pl.col("ItemOID") == "IT.EC.ECENDTC").first().alias("ECENDTC"),46    )47    .join(48        vs_wide,49        on=["StudyOID", "SubjectKey", "StudyEventOID", "StudyEventRepeatKey"],50        how="left",51    )52    .filter(pl.col("ECOCCUR") == "Y")53)545556def num(v):57    if v is None or v == "":58        return None59    try:60        return float(v)61    except ValueError:62        return None636465records: list[dict] = []66for row in ec_wide.to_dicts():67    study = row["StudyOID"]68    subj = row["SubjectKey"]69    form = row["FORM"] or ""70    usubjid = subj71    if form == "KITDOSE":72        treatment, dose_mg = kit_map.get(row["KIT"] or "", (None, None))73        trt = treatment74        dose = num(dose_mg)75        dosu = "mg"76    elif form == "TABLET":77        trt = row["ECTRT"] or None78        tablets = num(row["TABLETS"])79        strength = num(row["STRENGTHMG"])80        dose = tablets * strength if tablets is not None and strength is not None else None81        dosu = "mg"82    elif form == "INFUSION":83        trt = row["ECTRT"] or None84        perkg = num(row["DOSEMKG"])85        weight = row["WEIGHT"]86        dose = perkg * weight if perkg is not None and weight is not None else None87        dosu = "mg"88    elif form == "AUCDOSE":89        trt = row["ECTRT"] or None90        dose = num(row["AUCTARGET"])91        dosu = "AUC"92    else:93        trt = row["ECTRT"] or None94        dose = None95        dosu = "mg"96    records.append(97        {98            "STUDYID": study,99            "USUBJID": usubjid,100            "EXTRT": trt,101            "EXDOSE": dose,102            "EXDOSU": dosu,103            "EXSTDTC": row["ECSTDTC"] or None,104            "EXENDTC": row["ECENDTC"] or None,105            "FORMREPEAT": int(row["FormRepeatKey"]) if row["FormRepeatKey"] else 0,106        }107    )108109ex = (110    pl.DataFrame(111        records,112        schema={113            "STUDYID": pl.String,114            "USUBJID": pl.String,115            "EXTRT": pl.String,116            "EXDOSE": pl.Float64,117            "EXDOSU": pl.String,118            "EXSTDTC": pl.String,119            "EXENDTC": pl.String,120            "FORMREPEAT": pl.Int64,121        },122    )123    .with_columns(124        DOMAIN=pl.lit("EX"),125        EXSTDTC_DT=pl.col("EXSTDTC").str.to_date(strict=False),126    )127    .sort(["STUDYID", "USUBJID", "EXSTDTC_DT", "FORMREPEAT"], nulls_last=True)128    .with_columns(EXSEQ=pl.col("USUBJID").cum_count().over(["STUDYID", "USUBJID"]))129    .with_columns(EXSEQ=pl.col("EXSEQ").cast(pl.Int64))130    .select(131        "DOMAIN",132        "STUDYID",133        "USUBJID",134        "EXSEQ",135        "EXTRT",136        "EXDOSE",137        "EXDOSU",138        "EXSTDTC",139        "EXENDTC",140    )141)142143Path("/app/output").mkdir(exist_ok=True)144ex.write_csv("/app/output/ex.csv")

1/* opensas reference solution for the yamaa benchmark sdtm-ex-from-ec.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data kit_list;6  length KIT TREATMENT DOSE_MG $1024;7  infile "/app/input/kit_list.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input KIT : $char1024. TREATMENT : $char1024. DOSE_MG : $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 StudyEventOID StudyEventRepeatKey FormRepeatKey; run;18data doses;19  length ECTRT ECSTDTC ECENDTC ECOCCUR FORM TABLETS STRENGTHMG DOSEMKG KIT AUCTARGET $1024;20  retain ECTRT ECSTDTC ECENDTC ECOCCUR FORM TABLETS STRENGTHMG DOSEMKG KIT AUCTARGET;21  set ordered;22  where FormOID='FO.EC';23  by StudyOID SubjectKey StudyEventOID StudyEventRepeatKey FormRepeatKey;24  if first.FormRepeatKey then do;25    ECTRT=''; ECSTDTC=''; ECENDTC=''; ECOCCUR=''; FORM=''; TABLETS=''; STRENGTHMG=''; DOSEMKG=''; KIT=''; AUCTARGET='';26  end;27    if ItemOID='IT.EC.ECTRT' then ECTRT=Value;28  if ItemOID='IT.EC.ECSTDTC' then ECSTDTC=Value;29  if ItemOID='IT.EC.ECENDTC' then ECENDTC=Value;30  if ItemOID='IT.EC.ECOCCUR' then ECOCCUR=Value;31  if ItemOID='IT.EC.FORM' then FORM=Value;32  if ItemOID='IT.EC.TABLETS' then TABLETS=Value;33  if ItemOID='IT.EC.STRENGTHMG' then STRENGTHMG=Value;34  if ItemOID='IT.EC.DOSEMKG' then DOSEMKG=Value;35  if ItemOID='IT.EC.KIT' then KIT=Value;36  if ItemOID='IT.EC.AUCTARGET' then AUCTARGET=Value;37  if last.FormRepeatKey then output;38run;39proc sort data=odm out=ordered; by StudyOID SubjectKey StudyEventOID StudyEventRepeatKey; run;40data weights;41  length VISITWEIGHT $1024;42  retain VISITWEIGHT;43  set ordered;44  where FormOID='FO.VS';45  by StudyOID SubjectKey StudyEventOID StudyEventRepeatKey;46  if first.StudyEventRepeatKey then do;47    VISITWEIGHT='';48  end;49    if ItemOID='IT.VS.WEIGHT' then VISITWEIGHT=Value;50  if last.StudyEventRepeatKey then output;51run;52proc sql;53create table joined as select a.*,b.VISITWEIGHT,c.TREATMENT,c.DOSE_MG from doses a left join weights b on a.StudyOID=b.StudyOID and a.SubjectKey=b.SubjectKey and a.StudyEventOID=b.StudyEventOID and a.StudyEventRepeatKey=b.StudyEventRepeatKey left join kit_list c on a.KIT=c.KIT; quit;54data prepared; set joined; if ECOCCUR='Y'; length DOMAIN $2 STUDYID USUBJID EXTRT EXDOSU EXSTDTC EXENDTC $1024;55DOMAIN='EX'; STUDYID=StudyOID; USUBJID=SubjectKey; EXTRT=ECTRT; EXSTDTC=ECSTDTC; EXENDTC=ECENDTC; EXDOSU='mg'; EXDOSE=.;56if FORM='TABLET' then EXDOSE=input(TABLETS,best32.)*input(STRENGTHMG,best32.);57if FORM='INFUSION' then EXDOSE=input(DOSEMKG,best32.)*input(VISITWEIGHT,best32.);58if FORM='KITDOSE' then do; EXTRT=TREATMENT; EXDOSE=input(DOSE_MG,best32.); end;59if FORM='AUCDOSE' then do; EXDOSE=input(AUCTARGET,best32.); EXDOSU='AUC'; end;60FORMREP=input(FormRepeatKey,best32.); NODATE=missing(EXSTDTC); run;61proc sort data=prepared; by STUDYID USUBJID NODATE EXSTDTC FORMREP; run;62data result; set prepared; by STUDYID USUBJID; retain EXSEQ;63  if first.USUBJID then EXSEQ=0;64  EXSEQ+1;65run;6667proc sql;68  create table final as select DOMAIN, STUDYID, USUBJID, EXSEQ, EXTRT, EXDOSE, EXDOSU, EXSTDTC, EXENDTC from result;69quit;70proc export data=final outfile="/app/output/ex.csv" dbms=csv replace; run;