Lab-Specific Reference Ranges

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

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

1# Reference solution for the yamaa benchmark sdtm-lb-lab-specific-ranges (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(purrr)11library(lubridate)1213odm <- read_csv(14  "/app/input/odm.csv",15  col_types = cols(.default = col_character())16) |>17  mutate(Value = na_if(Value, ""))1819lbrange <- read_csv(20  "/app/input/lbrange.csv",21  col_types = cols(22    AGELO = col_integer(), AGEHI = col_integer(),23    NRLO = col_double(), NRHI = col_double(),24    .default = col_character()25  )26)2728age_years <- function(birth, collected) {29  b <- ymd(birth)30  c <- ymd(collected)31  years <- year(c) - year(b)32  before_birthday <- (month(c) < month(b)) |33    (month(c) == month(b) & day(c) < day(b))34  years - if_else(before_birthday, 1L, 0L)35}3637dm <- odm |>38  filter(ItemGroupOID == "IG.DM") |>39  select(SubjectKey, ItemOID, Value) |>40  pivot_wider(names_from = ItemOID, values_from = Value)4142labs <- odm |>43  filter(ItemGroupOID == "IG.LB") |>44  distinct(StudyOID, SubjectKey, ItemGroupRepeatKey)4546one_lab <- function(study, subj, igrep) {47  sub <- odm |>48    filter(49      SubjectKey == subj, ItemGroupOID == "IG.LB",50      ItemGroupRepeatKey == igrep51    )52  get <- function(oid) {53    hit <- sub |> filter(ItemOID == oid)54    if (nrow(hit) == 0) return(NA_character_)55    hit$Value[[1]]56  }57  dsub <- dm |> filter(SubjectKey == subj)58  sex <- if (nrow(dsub) == 0) NA_character_ else dsub$`IT.DM.SEX`[[1]]59  birth <- if (nrow(dsub) == 0) NA_character_ else dsub$`IT.DM.BRTHDTC`[[1]]60  code <- get("IT.LB.LBTESTCD")61  lbnam <- get("IT.LB.LBNAM")62  lbdtc <- get("IT.LB.LBDTC")63  raw <- get("IT.LB.LBSTRESN")64  stresn <- suppressWarnings(as.numeric(raw))65  unit <- NA_character_66  lo <- NA_real_67  hi <- NA_real_68  if (!is.na(birth) && !is.na(lbdtc) && !is.na(code) && !is.na(lbnam) && !is.na(sex)) {69    age <- age_years(birth, lbdtc)70    hit <- lbrange |>71      filter(72        LBNAM == lbnam, LBTESTCD == code, SEX == sex,73        AGELO <= age, age < AGEHI,74        EFFSTDT <= lbdtc,75        is.na(EFFENDT) | lbdtc <= EFFENDT76      ) |>77      arrange(desc(EFFSTDT), desc(AGELO)) |>78      slice_head(n = 1)79    if (nrow(hit) == 1) {80      unit <- hit$UNIT[[1]]81      lo <- hit$NRLO[[1]]82      hi <- hit$NRHI[[1]]83    }84  }85  nrind <- if (is.na(stresn) || is.na(lo) || is.na(hi)) {86    NA_character_87  } else if (stresn < lo) {88    "LOW"89  } else if (stresn > hi) {90    "HIGH"91  } else {92    "NORMAL"93  }94  tibble(95    DOMAIN = "LB",96    STUDYID = study,97    USUBJID = subj,98    LBSEQ = as.integer(igrep),99    LBTESTCD = code,100    SEX = sex,101    LBNAM = lbnam,102    LBSTRESN = stresn,103    LBORRESU = unit,104    LBSTNRLO = lo,105    LBSTNRHI = hi,106    LBNRIND = nrind107  )108}109110lb <- pmap_dfr(111  list(labs$StudyOID, labs$SubjectKey, labs$ItemGroupRepeatKey),112  one_lab113) |>114  arrange(STUDYID, USUBJID, LBSEQ)115116dir.create("/app/output", showWarnings = FALSE)117write_csv(lb, "/app/output/lb.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-lb-lab-specific-ranges (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 datetime import date8from pathlib import Path910import polars as pl1112odm = pl.read_csv("/app/input/odm.csv", infer_schema=False)13lbrange = pl.read_csv("/app/input/lbrange.csv", infer_schema=False).with_columns(14    pl.col("AGELO").cast(pl.Int64),15    pl.col("AGEHI").cast(pl.Int64),16    pl.col("NRLO").cast(pl.Float64, strict=False),17    pl.col("NRHI").cast(pl.Float64, strict=False),18)1920subjects = {}21labs = {}22study_of = {}23for r in odm.to_dicts():24    study_of[r["SubjectKey"]] = r["StudyOID"]25    if r["ItemGroupOID"] == "IG.DM":26        subjects.setdefault(r["SubjectKey"], {})[r["ItemOID"]] = (27            r["Value"] if r["Value"] not in (None, "") else None28        )29    elif r["ItemGroupOID"] == "IG.LB":30        key = (r["SubjectKey"], r["ItemGroupRepeatKey"])31        labs.setdefault(key, {})[r["ItemOID"]] = (32            r["Value"] if r["Value"] not in (None, "") else None33        )343536def age_years(birth, collected):37    b = date.fromisoformat(birth)38    c = date.fromisoformat(collected)39    years = c.year - b.year40    if (c.month, c.day) < (b.month, b.day):41        years -= 142    return years434445ranges = sorted(lbrange.to_dicts(), key=lambda r: (r["EFFSTDT"], r["AGELO"]), reverse=True)4647rows = []48for (subj, repkey), lab in labs.items():49    dm = subjects.get(subj, {})50    sex = dm.get("IT.DM.SEX")51    birth = dm.get("IT.DM.BRTHDTC")52    code = lab.get("IT.LB.LBTESTCD")53    lbnam = lab.get("IT.LB.LBNAM")54    lbdtc = lab.get("IT.LB.LBDTC")55    raw = lab.get("IT.LB.LBSTRESN")56    try:57        stresn = float(raw) if raw not in (None, "") else None58    except ValueError:59        stresn = None60    unit = lo = hi = None61    if birth and lbdtc and code and lbnam and sex:62        age = age_years(birth, lbdtc)63        best = None64        for rg in ranges:65            if rg["LBNAM"] != lbnam or rg["LBTESTCD"] != code or rg["SEX"] != sex:66                continue67            if not (rg["AGELO"] <= age < rg["AGEHI"]):68                continue69            eff_start = rg["EFFSTDT"]70            eff_end = rg["EFFENDT"] if rg["EFFENDT"] not in (None, "") else None71            if lbdtc < eff_start:72                continue73            if eff_end is not None and lbdtc > eff_end:74                continue75            best = rg76            break77        if best is not None:78            unit, lo, hi = best["UNIT"], best["NRLO"], best["NRHI"]79    if stresn is None or lo is None or hi is None:80        nrind = None81    elif stresn < lo:82        nrind = "LOW"83    elif stresn > hi:84        nrind = "HIGH"85    else:86        nrind = "NORMAL"87    rows.append(88        {89            "DOMAIN": "LB",90            "STUDYID": study_of[subj],91            "USUBJID": subj,92            "LBSEQ": int(repkey),93            "LBTESTCD": code,94            "SEX": sex,95            "LBNAM": lbnam,96            "LBSTRESN": stresn,97            "LBORRESU": unit,98            "LBSTNRLO": lo,99            "LBSTNRHI": hi,100            "LBNRIND": nrind,101        }102    )103104lb = pl.DataFrame(105    rows,106    schema={107        "DOMAIN": pl.String,108        "STUDYID": pl.String,109        "USUBJID": pl.String,110        "LBSEQ": pl.Int64,111        "LBTESTCD": pl.String,112        "SEX": pl.String,113        "LBNAM": pl.String,114        "LBSTRESN": pl.Float64,115        "LBORRESU": pl.String,116        "LBSTNRLO": pl.Float64,117        "LBSTNRHI": pl.Float64,118        "LBNRIND": pl.String,119    },120).sort(["STUDYID", "USUBJID", "LBSEQ"])121122Path("/app/output").mkdir(parents=True, exist_ok=True)123lb.write_csv("/app/output/lb.csv")

1/* opensas reference solution for the yamaa benchmark sdtm-lb-lab-specific-ranges.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data lbrange;6  length LBNAM LBTESTCD SEX EFFSTDT EFFENDT UNIT $1024;7  infile "/app/input/lbrange.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input LBNAM : $char1024. LBTESTCD : $char1024. SEX : $char1024. AGELO AGEHI EFFSTDT : $char1024. EFFENDT : $char1024. UNIT : $char1024. NRLO NRHI;9  RANGE_ORDER=_n_;10run;1112data odm;13  length StudyOID MetaDataVersionOID SubjectKey StudyEventOID StudyEventRepeatKey ItemGroupOID ItemGroupRepeatKey ItemOID Value $1024;14  infile "/app/input/odm.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;15  input StudyOID : $char1024. MetaDataVersionOID : $char1024. SubjectKey : $char1024. StudyEventOID : $char1024. StudyEventRepeatKey : $char1024. ItemGroupOID : $char1024. ItemGroupRepeatKey : $char1024. ItemOID : $char1024. Value : $char1024.;16run;1718proc sort data=odm out=ordered; by StudyOID SubjectKey; run;19data subjects;20  length SEX BRTHDTC $1024;21  retain SEX BRTHDTC;22  set ordered;23  where ItemGroupOID='IG.DM';24  by StudyOID SubjectKey;25  if first.SubjectKey then do;26    SEX=''; BRTHDTC='';27  end;28    if ItemOID='IT.DM.SEX' then SEX=Value;29  if ItemOID='IT.DM.BRTHDTC' then BRTHDTC=Value;30  if last.SubjectKey then output;31run;32proc sort data=odm out=ordered; by StudyOID SubjectKey ItemGroupRepeatKey; run;33data panels;34  length LBDTC LBNAM LBTESTCD RESULTTEXT $1024;35  retain LBDTC LBNAM LBTESTCD RESULTTEXT;36  set ordered;37  where ItemGroupOID='IG.LB';38  by StudyOID SubjectKey ItemGroupRepeatKey;39  if first.ItemGroupRepeatKey then do;40    LBDTC=''; LBNAM=''; LBTESTCD=''; RESULTTEXT='';41  end;42    if ItemOID='IT.LB.LBDTC' then LBDTC=Value;43  if ItemOID='IT.LB.LBNAM' then LBNAM=Value;44  if ItemOID='IT.LB.LBTESTCD' then LBTESTCD=Value;45  if ItemOID='IT.LB.LBSTRESN' then RESULTTEXT=Value;46  if last.ItemGroupRepeatKey then output;47run;48proc sql; create table withbirth as select a.*,b.SEX,b.BRTHDTC from panels a left join subjects b on a.StudyOID=b.StudyOID and a.SubjectKey=b.SubjectKey; quit;49data collected; set withbirth; AGE=.; if lengthn(BRTHDTC)=10 and lengthn(LBDTC)=10 then AGE=intck('year',input(BRTHDTC,yymmdd10.),input(LBDTC,yymmdd10.),'continuous'); LBSTRESN=input(RESULTTEXT,?? best32.); LBSEQ=input(ItemGroupRepeatKey,best32.); run;50proc sql; create table joined as select a.*,b.UNIT as LBORRESU,b.NRLO as LBSTNRLO,b.NRHI as LBSTNRHI,b.EFFSTDT as RANGE_START,b.AGELO as RANGE_AGE,b.RANGE_ORDER from collected a left join lbrange b on a.LBNAM=b.LBNAM and a.LBTESTCD=b.LBTESTCD and a.SEX=b.SEX and a.AGE>=b.AGELO and a.AGE<b.AGEHI and a.LBDTC>=b.EFFSTDT and (missing(b.EFFENDT) or a.LBDTC<=b.EFFENDT); quit;51proc sort data=joined; by StudyOID SubjectKey LBSEQ descending RANGE_START descending RANGE_AGE RANGE_ORDER; run;52data selected; set joined; by StudyOID SubjectKey LBSEQ; if first.LBSEQ; run;53data result; set selected; length DOMAIN $2 STUDYID USUBJID $1024 LBNRIND $8; DOMAIN='LB'; STUDYID=StudyOID; USUBJID=SubjectKey; LBNRIND=''; if not missing(LBSTRESN) and not missing(LBSTNRLO) and not missing(LBSTNRHI) then do; if LBSTRESN<LBSTNRLO then LBNRIND='LOW'; else if LBSTRESN>LBSTNRHI then LBNRIND='HIGH'; else LBNRIND='NORMAL'; end; run;54proc sql;55  create table final as select DOMAIN, STUDYID, USUBJID, LBSEQ, LBTESTCD, SEX, LBNAM, LBSTRESN, LBORRESU, LBSTNRLO, LBSTNRHI, LBNRIND from result;56quit;57proc export data=final outfile="/app/output/lb.csv" dbms=csv replace; run;