Last Observation Before Exposure

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

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

1# Reference solution for the yamaa benchmark sdtm-lb-last-observation-before-exposure (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(purrr)1011odm <- read_csv(12  "/app/input/odm.csv",13  col_types = cols(.default = col_character())14) |>15  mutate(Value = na_if(Value, ""))1617dm <- read_csv(18  "/app/input/dm.csv",19  col_types = cols(.default = col_character())20) |>21  mutate(RFSTDTC = na_if(RFSTDTC, ""))2223mapping <- read_csv(24  "/app/input/lb_mapping.csv",25  col_types = cols(.default = col_character())26)2728groups <- odm |>29  distinct(StudyOID, SubjectKey, ItemGroupRepeatKey)3031one_group <- function(study, subj, igrep) {32  sub <- odm |>33    filter(SubjectKey == subj, ItemGroupRepeatKey == igrep)34  get <- function(oid) {35    hit <- sub |> filter(ItemOID == oid)36    if (nrow(hit) == 0) return(NA_character_)37    hit$Value[[1]]38  }39  lbdtc <- get("IT.LB.LBDTC")40  found <- mapping |>41    filter(ItemOID %in% sub$ItemOID) |>42    slice_head(n = 1)43  if (nrow(found) == 0) return(tibble())44  raw <- get(found$ItemOID[[1]])45  has_result <- !is.na(raw)46  tibble(47    DOMAIN = "LB",48    STUDYID = study,49    USUBJID = subj,50    LBSEQ = as.integer(igrep),51    LBTESTCD = found$LBTESTCD[[1]],52    LBSPEC = found$LBSPEC[[1]],53    LBORRES = raw,54    LBDTC = lbdtc,55    LBSTAT = if_else(has_result, NA_character_, "NOT DONE"),56    HAS = has_result57  )58}5960lb <- pmap_dfr(61  list(groups$StudyOID, groups$SubjectKey, groups$ItemGroupRepeatKey),62  one_group63)6465flags <- lb |>66  filter(HAS, !is.na(LBDTC)) |>67  left_join(dm |> select(USUBJID, RFSTDTC), by = "USUBJID") |>68  mutate(DAYDATE = substr(LBDTC, 1, 10), REFDAY = substr(RFSTDTC, 1, 10)) |>69  filter(is.na(REFDAY) | DAYDATE <= REFDAY) |>70  arrange(USUBJID, LBTESTCD, LBSPEC, DAYDATE, LBSEQ) |>71  group_by(USUBJID, LBTESTCD, LBSPEC) |>72  slice_tail(n = 1) |>73  ungroup() |>74  transmute(USUBJID, LBTESTCD, LBSPEC, FLAGSEQ = LBSEQ)7576lb <- lb |>77  left_join(flags, by = c("USUBJID", "LBTESTCD", "LBSPEC")) |>78  mutate(79    LBLOBXFL = if_else(HAS & !is.na(FLAGSEQ) & LBSEQ == FLAGSEQ, "Y", NA_character_)80  ) |>81  arrange(STUDYID, USUBJID, LBSEQ) |>82  select(83    DOMAIN, STUDYID, USUBJID, LBSEQ, LBTESTCD, LBSPEC, LBORRES,84    LBDTC, LBSTAT, LBLOBXFL85  )8687dir.create("/app/output", showWarnings = FALSE)88write_csv(lb, "/app/output/lb.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-lb-last-observation-before-exposure (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 LB dataset from DM, LB_MAPPING and ODM extracts.89One output record per laboratory result item group in the ODM extract.10LBLOBXFL flags the last observation with a result, within each11(LBTESTCD, LBSPEC) series for a subject, whose collection date falls on12or before the subject's reference start date (RFSTDTC in DM). A subject13without a reference start date has the flag on the latest record with a14result in the series.15"""1617import csv18from datetime import date1920INPUT_DIR = "/app/input"21OUTPUT_FILE = "/app/output/lb.csv"2223COLUMNS = [24    "DOMAIN",25    "STUDYID",26    "USUBJID",27    "LBSEQ",28    "LBTESTCD",29    "LBSPEC",30    "LBORRES",31    "LBDTC",32    "LBSTAT",33    "LBLOBXFL",34]353637def read_csv(path):38    with open(path, newline="", encoding="utf-8") as handle:39        return list(csv.DictReader(handle))404142def day(text):43    """Parse an ISO date/datetime at day precision; None when blank."""44    text = (text or "").strip()45    if not text:46        return None47    return date.fromisoformat(text[:10])484950def main():51    dm = read_csv(f"{INPUT_DIR}/dm.csv")52    mapping = read_csv(f"{INPUT_DIR}/lb_mapping.csv")53    odm = read_csv(f"{INPUT_DIR}/odm.csv")5455    ref_start = {}56    for row in dm:57        ref_start[row["USUBJID"]] = day(row["RFSTDTC"])5859    test_map = {row["ItemOID"]: (row["LBTESTCD"], row["LBSPEC"]) for row in mapping}6061    # Group the ODM items into one record per item group occurrence,62    # keeping the collected order.63    groups = {}64    order = []65    for row in odm:66        key = (row["StudyOID"], row["SubjectKey"], row["ItemGroupRepeatKey"])67        if key not in groups:68            groups[key] = {69                "STUDYID": row["StudyOID"],70                "USUBJID": row["SubjectKey"],71                "LBSEQ": row["ItemGroupRepeatKey"],72                "items": {},73            }74            order.append(key)75        groups[key]["items"][row["ItemOID"]] = row["Value"]7677    records = []78    for key in order:79        group = groups[key]80        items = group["items"]81        result_items = [oid for oid in items if oid in test_map]82        if not result_items:83            continue84        item_oid = result_items[0]85        testcd, spec = test_map[item_oid]86        result = (items.get(item_oid) or "").strip()87        stat = (items.get("IT.LB.LBSTAT") or "").strip()88        if not result:89            stat = stat or "NOT DONE"90        records.append(91            {92                "DOMAIN": "LB",93                "STUDYID": group["STUDYID"],94                "USUBJID": group["USUBJID"],95                "LBSEQ": group["LBSEQ"],96                "LBTESTCD": testcd,97                "LBSPEC": spec,98                "LBORRES": result,99                "LBDTC": (items.get("IT.LB.LBDTC") or "").strip(),100                "LBSTAT": stat,101                "LBLOBXFL": "",102            }103        )104105    # Assign the flag within each subject and (test, specimen) series.106    series = {}107    for idx, record in enumerate(records):108        key = (record["USUBJID"], record["LBTESTCD"], record["LBSPEC"])109        series.setdefault(key, []).append(idx)110111    for (usubjid, _testcd, _spec), idxs in series.items():112        ref = ref_start.get(usubjid)113        candidates = []114        for idx in idxs:115            record = records[idx]116            if not record["LBORRES"]:117                continue  # a record without a result never carries the flag118            collected = day(record["LBDTC"])119            if ref is not None and (collected is None or collected > ref):120                continue  # collected after the reference start date121            candidates.append(idx)122        if not candidates:123            continue124        best = max(125            candidates,126            key=lambda i: (day(records[i]["LBDTC"]) or date.min, int(records[i]["LBSEQ"])),127        )128        records[best]["LBLOBXFL"] = "Y"129130    with open(OUTPUT_FILE, "w", newline="", encoding="utf-8") as handle:131        writer = csv.DictWriter(handle, fieldnames=COLUMNS)132        writer.writeheader()133        writer.writerows(records)134135136if __name__ == "__main__":137    main()

1/* opensas reference solution for the yamaa benchmark sdtm-lb-last-observation-before-exposure.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 lb_mapping;12  length ItemOID LBTESTCD LBSPEC $1024;13  infile "/app/input/lb_mapping.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;14  input ItemOID : $char1024. LBTESTCD : $char1024. LBSPEC : $char1024.;15run;1617data odm;18  length StudyOID MetaDataVersionOID SubjectKey StudyEventOID StudyEventRepeatKey FormOID FormRepeatKey ItemGroupOID ItemGroupRepeatKey ItemOID Value $1024;19  infile "/app/input/odm.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;20  input StudyOID : $char1024. MetaDataVersionOID : $char1024. SubjectKey : $char1024. StudyEventOID : $char1024. StudyEventRepeatKey : $char1024. FormOID : $char1024. FormRepeatKey : $char1024. ItemGroupOID : $char1024. ItemGroupRepeatKey : $char1024. ItemOID : $char1024. Value : $char1024.;21run;2223proc sort data=odm out=ordered; by StudyOID SubjectKey StudyEventOID StudyEventRepeatKey FormOID FormRepeatKey ItemGroupRepeatKey; run;24data dates;25  length LBDTC LBSTAT $1024;26  retain LBDTC LBSTAT;27  set ordered;2829  by StudyOID SubjectKey StudyEventOID StudyEventRepeatKey FormOID FormRepeatKey ItemGroupRepeatKey;30  if first.ItemGroupRepeatKey then do;31    LBDTC=''; LBSTAT='';32  end;33    if ItemOID='IT.LB.LBDTC' then LBDTC=Value;34  if ItemOID='IT.LB.LBSTAT' then LBSTAT=Value;35  if last.ItemGroupRepeatKey then output;36run;37proc sql;38create table joined as select a.*,b.LBDTC,b.LBSTAT,c.LBTESTCD,c.LBSPEC,d.RFSTDTC 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 and a.ItemGroupRepeatKey=b.ItemGroupRepeatKey inner join lb_mapping c on a.ItemOID=c.ItemOID left join dm d on a.StudyOID=d.STUDYID and a.SubjectKey=d.USUBJID; quit;39data collected; set joined; length DOMAIN $2 STUDYID USUBJID LBORRES DAYDATE REFDAY $1024; DOMAIN='LB'; STUDYID=StudyOID; USUBJID=SubjectKey; LBSEQ=input(ItemGroupRepeatKey,best32.); LBORRES=Value;40if missing(Value) then LBSTAT='NOT DONE'; if LBSTAT='NOT DONE' then LBORRES=''; DAYDATE=''; REFDAY=''; if lengthn(LBDTC)>=10 then DAYDATE=substr(LBDTC,1,10); if lengthn(RFSTDTC)>=10 then REFDAY=substr(RFSTDTC,1,10); run;41data eligible; set collected; if not missing(LBORRES) and not missing(DAYDATE) and (missing(REFDAY) or DAYDATE<=REFDAY); run;42proc sort data=eligible; by STUDYID USUBJID LBTESTCD LBSPEC descending DAYDATE descending LBSEQ; run;43data flagged; set eligible; by STUDYID USUBJID LBTESTCD LBSPEC; if first.LBSPEC; length LBLOBXFL $1; LBLOBXFL='Y'; run;44proc sql; create table result as select a.*,b.LBLOBXFL from collected a left join flagged b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID and a.LBSEQ=b.LBSEQ and a.LBTESTCD=b.LBTESTCD and a.LBSPEC=b.LBSPEC; quit;45proc sql;46  create table final as select DOMAIN, STUDYID, USUBJID, LBSEQ, LBTESTCD, LBSPEC, LBORRES, LBDTC, LBSTAT, LBLOBXFL from result;47quit;48proc export data=final outfile="/app/output/lb.csv" dbms=csv replace; run;