Planned Time Points Around Dosing

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

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

1# Reference solution for the yamaa benchmark sdtm-vs-planned-time-points (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(lubridate)9library(readr)10library(tidyr)1112study_day <- function(ref, coll) {13  if (is.na(ref) || is.na(coll)) {14    return(NA_integer_)15  }16  delta <- as.integer(coll - ref)17  if (delta >= 0L) delta + 1L else delta18}1920time_point <- function(form) {21  name <- toupper(form)22  if (grepl("PREDOSE", name, fixed = TRUE)) {23    list(vstpt = "PRE-DOSE", vstptnum = 1L, vseltm = "-PT15M")24  } else if (grepl("30MIN", name, fixed = TRUE)) {25    list(vstpt = "30 MIN POST-DOSE", vstptnum = 2L, vseltm = "PT30M")26  } else if (grepl("1H", name, fixed = TRUE)) {27    list(vstpt = "1 H POST-DOSE", vstptnum = 3L, vseltm = "PT1H")28  } else if (grepl("4H", name, fixed = TRUE)) {29    list(vstpt = "4 H POST-DOSE", vstptnum = 4L, vseltm = "PT4H")30  } else {31    NULL32  }33}3435dm <- read_csv(36  "/app/input/dm.csv",37  col_types = cols(.default = col_character())38)39odm <- read_csv(40  "/app/input/odm.csv",41  col_types = cols(.default = col_character())42)43mapping <- read_csv(44  "/app/input/vs_mapping.csv",45  col_types = cols(.default = col_character())46)4748ref_map <- setNames(suppressWarnings(ymd(dm$RFSTDTC)), dm$USUBJID)4950# The first-dose datetime on each subject's dosing form is the reference51# time point every planned elapsed time counts from.52doses <- odm |>53  filter(FormOID == "FO.EX", ItemOID == "IT.EX.EXSTDTC") |>54  select(SubjectKey, VSRFTDTC = Value)55test_map <- setNames(mapping$VSTESTCD, mapping$ItemOID)56unit_map <- setNames(mapping$UNIT, mapping$ItemOID)5758forms <- odm |>59  filter(StudyEventOID == "SE.D1", grepl("^FO\\.VS_", FormOID)) |>60  group_by(SubjectKey, FormOID, StudyOID) |>61  summarise(62    vsdtc = Value[ItemOID == "IT.VS.VSDTC"][1],63    .groups = "drop"64  )6566results <- odm |>67  filter(68    StudyEventOID == "SE.D1",69    grepl("^FO\\.VS_", FormOID),70    ItemOID %in% names(test_map),71    !is.na(Value) & Value != ""72  ) |>73  left_join(forms, by = c("SubjectKey", "FormOID", "StudyOID")) |>74  left_join(doses, by = "SubjectKey") |>75  rowwise() |>76  mutate(77    point = list(time_point(FormOID)),78    VSTPT = point$vstpt,79    VSTPTNUM = point$vstptnum,80    VSELTM = point$vseltm,81    VSTPTREF = "FIRST DOSE",82    VSTESTCD = test_map[[ItemOID]],83    VSORRES = Value,84    VSORRESU = unit_map[[ItemOID]],85    VSSTRESN = suppressWarnings(as.numeric(Value)),86    VSSTRESU = VSORRESU,87    VSDTC = vsdtc,88    ref = ref_map[[SubjectKey]],89    VSDY = study_day(ref, suppressWarnings(as_date(ymd_hms(VSDTC)))),90    test_ord = case_when(91      VSTESTCD == "PULSE" ~ 1L,92      VSTESTCD == "SYSBP" ~ 2L,93      .default = 3L94    )95  ) |>96  ungroup() |>97  filter(!is.na(VSTPT), !is.na(VSDTC) & VSDTC != "") |>98  arrange(SubjectKey, VSTPTNUM, test_ord) |>99  group_by(SubjectKey) |>100  mutate(VSSEQ = row_number()) |>101  ungroup() |>102  arrange(SubjectKey, VSSEQ) |>103  transmute(104    DOMAIN = "VS",105    STUDYID = StudyOID,106    USUBJID = SubjectKey,107    VSSEQ,108    VISIT = "DAY 1",109    VSTPT, VSTPTNUM, VSELTM, VSTPTREF, VSRFTDTC, VSTESTCD, VSORRES, VSORRESU,110    VSSTRESN, VSSTRESU, VSDTC, VSDY111  )112113dir.create("/app/output", showWarnings = FALSE)114write_csv(results, "/app/output/vs.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-vs-planned-time-points (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 VS dataset from DM, ODM, and VS_MAPPING.89One record per collected vital-signs result. Each vital-signs form fixes10the planned time point of every result on it; the form's datetime item11gives the actual collection datetime for all results on that form.12"""1314import csv15from datetime import date, datetime16from pathlib import Path1718INPUT = Path("/app/input")19OUTPUT = Path("/app/output")2021# Form OID -> (VSTPT, VSTPTNUM, VSELTM)22FORM_TIMEPOINTS = {23    "FO.VS_PREDOSE": ("PRE-DOSE", 1, "-PT15M"),24    "FO.VS_30MIN": ("30 MIN POST-DOSE", 2, "PT30M"),25    "FO.VS_1H": ("1 H POST-DOSE", 3, "PT1H"),26    "FO.VS_4H": ("4 H POST-DOSE", 4, "PT4H"),27}2829# Test order within a time point for VSSEQ.30TEST_ORDER = {"PULSE": 0, "SYSBP": 1, "DIABP": 2}3132COLUMNS = [33    "DOMAIN", "STUDYID", "USUBJID", "VSSEQ", "VISIT", "VSTPT",34    "VSTPTNUM", "VSELTM", "VSTPTREF", "VSRFTDTC", "VSTESTCD", "VSORRES",35    "VSORRESU", "VSSTRESN", "VSSTRESU", "VSDTC", "VSDY",36]373839def read_csv(path):40    with open(path, newline="", encoding="utf-8") as handle:41        return list(csv.DictReader(handle))424344def study_day(collection, reference):45    """SDTM study day: reference day is day 1, no day zero."""46    if not collection or not reference:47        return ""48    delta = (collection - reference).days49    return delta + 1 if delta >= 0 else delta505152def numeric_text(value):53    number = float(value)54    return str(int(number)) if number == int(number) else str(number)555657def main():58    dm_rows = read_csv(INPUT / "dm.csv")59    odm_rows = read_csv(INPUT / "odm.csv")60    mapping_rows = read_csv(INPUT / "vs_mapping.csv")6162    reference_start = {}63    for row in dm_rows:64        value = (row.get("RFSTDTC") or "").strip()65        reference_start[row["USUBJID"]] = (66            datetime.strptime(value[:10], "%Y-%m-%d").date() if value else None67        )6869    dictionary = {70        row["ItemOID"]: (row["VSTESTCD"], row["UNIT"]) for row in mapping_rows71    }7273    # The first-dose datetime on each subject's dosing form is the reference74    # time point every planned elapsed time counts from.75    first_dose = {76        row["SubjectKey"]: row["Value"].strip()77        for row in odm_rows78        if row["FormOID"] == "FO.EX" and row["ItemOID"] == "IT.EX.EXSTDTC"79    }8081    # Group ODM item rows by subject and form occurrence.82    forms = {}83    for row in odm_rows:84        form_oid = row["FormOID"]85        if form_oid not in FORM_TIMEPOINTS:86            continue87        key = (88            row["StudyOID"],89            row["SubjectKey"],90            row["StudyEventOID"],91            row["StudyEventRepeatKey"],92            form_oid,93            row["FormRepeatKey"],94            row["ItemGroupOID"],95            row["ItemGroupRepeatKey"],96        )97        forms.setdefault(key, {})[row["ItemOID"]] = row["Value"]9899    records = []100    for key, items in forms.items():101        study_id, subject = key[0], key[1]102        vstpt, vstptnum, vseltm = FORM_TIMEPOINTS[key[4]]103        vsdtc = (items.get("IT.VS.VSDTC") or "").strip()104        collection_date = (105            datetime.strptime(vsdtc[:10], "%Y-%m-%d").date() if vsdtc else None106        )107        vsdy = study_day(collection_date, reference_start.get(subject))108        for item_oid, (testcd, unit) in dictionary.items():109            if item_oid not in items:110                continue111            value = items[item_oid]112            records.append(113                {114                    "DOMAIN": "VS",115                    "STUDYID": study_id,116                    "USUBJID": subject,117                    "VISIT": "DAY 1",118                    "VSTPT": vstpt,119                    "VSTPTNUM": vstptnum,120                    "VSELTM": vseltm,121                    "VSTPTREF": "FIRST DOSE",122                    "VSRFTDTC": first_dose.get(subject, ""),123                    "VSTESTCD": testcd,124                    "VSORRES": value,125                    "VSORRESU": unit,126                    "VSSTRESN": numeric_text(value),127                    "VSSTRESU": unit,128                    "VSDTC": vsdtc,129                    "VSDY": vsdy,130                    "_test_order": TEST_ORDER[testcd],131                }132            )133134    records.sort(135        key=lambda r: (r["USUBJID"], r["VSTPTNUM"], r["_test_order"])136    )137    counters = {}138    for record in records:139        subject = record["USUBJID"]140        counters[subject] = counters.get(subject, 0) + 1141        record["VSSEQ"] = counters[subject]142        del record["_test_order"]143144    OUTPUT.mkdir(parents=True, exist_ok=True)145    with open(OUTPUT / "vs.csv", "w", newline="", encoding="utf-8") as handle:146        writer = csv.DictWriter(handle, fieldnames=COLUMNS)147        writer.writeheader()148        writer.writerows(records)149150151if __name__ == "__main__":152    main()

1/* opensas reference solution for the yamaa benchmark sdtm-vs-planned-time-points.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;1617data vs_mapping;18  length ItemOID VSTESTCD VSTEST UNIT $1024;19  infile "/app/input/vs_mapping.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;20  input ItemOID : $char1024. VSTESTCD : $char1024. VSTEST : $char1024. UNIT : $char1024.;21run;2223proc sort data=odm out=ordered; by StudyOID SubjectKey StudyEventOID StudyEventRepeatKey FormOID FormRepeatKey; run;24data dates;25  length VSDTC $1024;26  retain VSDTC;27  set ordered;28  where ItemGroupOID='IG.VS';29  by StudyOID SubjectKey StudyEventOID StudyEventRepeatKey FormOID FormRepeatKey;30  if first.FormRepeatKey then do;31    VSDTC='';32  end;33    if ItemOID='IT.VS.VSDTC' then VSDTC=Value;34  if last.FormRepeatKey then output;35run;36/* The first-dose datetime on the dosing form is the reference time point. */37data doses; set odm; where FormOID='FO.EX' and ItemOID='IT.EX.EXSTDTC'; length VSRFTDTC $1024; VSRFTDTC=Value; keep StudyOID SubjectKey VSRFTDTC; run;38proc sql; create table joined as select a.*,b.VSDTC,c.VSTESTCD,c.UNIT,d.RFSTDTC,e.VSRFTDTC from odm a inner join dates b on a.StudyOID=b.StudyOID and a.SubjectKey=b.SubjectKey and a.StudyEventOID=b.StudyEventOID and a.StudyEventRepeatKey=b.StudyEventRepeatKey and a.FormOID=b.FormOID and a.FormRepeatKey=b.FormRepeatKey inner join vs_mapping c on a.ItemOID=c.ItemOID left join dm d on a.StudyOID=d.STUDYID and a.SubjectKey=d.USUBJID left join doses e on a.StudyOID=e.StudyOID and a.SubjectKey=e.SubjectKey; quit;39data prepared; set joined; length DOMAIN $2 STUDYID USUBJID VISIT VSTPT VSELTM VSTPTREF VSORRES VSORRESU VSSTRESU $1024; DOMAIN='VS'; VSTPTREF='FIRST DOSE'; STUDYID=StudyOID; USUBJID=SubjectKey; VISIT='DAY 1'; VSTPT=''; VSELTM=''; VSTPTNUM=.;40if FormOID='FO.VS_PREDOSE' then do; VSTPT='PRE-DOSE'; VSTPTNUM=1; VSELTM='-PT15M'; end;41if FormOID='FO.VS_30MIN' then do; VSTPT='30 MIN POST-DOSE'; VSTPTNUM=2; VSELTM='PT30M'; end;42if FormOID='FO.VS_1H' then do; VSTPT='1 H POST-DOSE'; VSTPTNUM=3; VSELTM='PT1H'; end;43if FormOID='FO.VS_4H' then do; VSTPT='4 H POST-DOSE'; VSTPTNUM=4; VSELTM='PT4H'; end;44VSORRES=Value; VSORRESU=UNIT; VSSTRESU=UNIT; VSSTRESN=input(Value,best32.); VSDY=.; if lengthn(VSDTC)>=10 and lengthn(RFSTDTC)>=10 then do; VSDY=input(substr(VSDTC,1,10),yymmdd10.)-input(substr(RFSTDTC,1,10),yymmdd10.); if VSDY>=0 then VSDY+1; end;45TESTORD=3; if VSTESTCD='PULSE' then TESTORD=1; if VSTESTCD='SYSBP' then TESTORD=2; run;46proc sort data=prepared; by STUDYID USUBJID VSTPTNUM TESTORD; run;47data result; set prepared; by STUDYID USUBJID; retain VSSEQ;48  if first.USUBJID then VSSEQ=0;49  VSSEQ+1;50run;5152proc sql;53  create table final as select DOMAIN, STUDYID, USUBJID, VSSEQ, VISIT, VSTPT, VSTPTNUM, VSELTM, VSTPTREF, VSRFTDTC, VSTESTCD, VSORRES, VSORRESU, VSSTRESN, VSSTRESU, VSDTC, VSDY from result;54quit;55proc export data=final outfile="/app/output/vs.csv" dbms=csv replace; run;