Standardize Non-Numeric Lab Results

1library(yamaa)23lb <- yamaa_domain("spec.yaml")$output4lb

1import yamaa23lb = yamaa.yamaa_domain("spec.yaml").output4lb

1# Reference solution for the yamaa benchmark sdtm-lb-character-results (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)10library(stringr)1112odm <- read_csv(13  "/app/input/odm.csv",14  col_types = cols(.default = col_character())15) |>16  mutate(Value = na_if(Value, ""))1718form_order <- c(19  "FO.LB_PROT" = 0L, "FO.LB_CK" = 1L, "FO.LB_GLUC" = 2L,20  "FO.LB_KETON" = 3L, "FO.LB_CREAT" = 4L21)22form_test <- c(23  "FO.LB_PROT" = "PROT", "FO.LB_CK" = "CK", "FO.LB_GLUC" = "GLUC",24  "FO.LB_KETON" = "KETON", "FO.LB_CREAT" = "CREAT"25)26test_name <- c(27  PROT = "Protein", CK = "Creatine Kinase", GLUC = "Glucose",28  KETON = "Ketones", CREAT = "Creatinine"29)3031standardize <- function(raw) {32  if (is.na(raw)) return(NA_character_)33  low <- str_to_lower(str_replace_all(str_trim(raw), " ", ""))34  if (low %in% c("negative", "neg")) return("NEGATIVE")35  if (low %in% c("trace", "tr")) return("TRACE")36  if (low %in% c("1+", "+1", "1plus")) return("1+")37  if (low %in% c("2+", "+2", "2plus")) return("2+")38  raw39}4041forms <- odm |>42  distinct(StudyOID, SubjectKey, FormOID, FormRepeatKey) |>43  filter(FormOID %in% names(form_order))4445one_form <- function(study, subj, formoid, formrep) {46  sub <- odm |>47    filter(48      SubjectKey == subj, FormOID == formoid, FormRepeatKey == formrep49    )50  get <- function(oid) {51    hit <- sub |> filter(ItemOID == oid)52    if (nrow(hit) == 0) return(NA_character_)53    hit$Value[[1]]54  }55  raw <- get("IT.LB.RESULT")56  if (is.na(raw)) return(tibble())57  code <- form_test[[formoid]]58  stresc <- standardize(raw)59  stresn <- suppressWarnings(as.numeric(raw))60  if (!is.na(stresc) && stresc %in% c("NEGATIVE", "TRACE", "1+", "2+")) {61    stresn <- NA_real_62  }63  if (str_starts(str_trim(raw), "<") || str_starts(str_trim(raw), ">")) {64    stresn <- NA_real_65  }66  nrind <- if (str_starts(str_trim(raw), "<")) {67    "LOW"68  } else if (str_starts(str_trim(raw), ">")) {69    "HIGH"70  } else {71    NA_character_72  }73  unit <- get("IT.LB.LBORRESU")74  tibble(75    DOMAIN = "LB",76    STUDYID = study,77    USUBJID = subj,78    ORDER = form_order[[formoid]],79    REP = as.integer(formrep),80    LBTESTCD = code,81    LBTEST = test_name[[code]],82    LBORRES = raw,83    LBORRESU = unit,84    LBSTRESC = stresc,85    LBSTRESN = stresn,86    LBSTRESU = unit,87    LBNRIND = nrind,88    LBDTC = get("IT.LB.LBDTC")89  )90}9192lb <- pmap_dfr(93  list(forms$StudyOID, forms$SubjectKey, forms$FormOID, forms$FormRepeatKey),94  one_form95) |>96  arrange(STUDYID, USUBJID, ORDER, REP) |>97  group_by(STUDYID, USUBJID) |>98  mutate(LBSEQ = row_number()) |>99  ungroup() |>100  select(101    DOMAIN, STUDYID, USUBJID, LBSEQ, LBTESTCD, LBTEST, LBORRES,102    LBORRESU, LBSTRESC, LBSTRESN, LBSTRESU, LBNRIND, LBDTC103  )104105dir.create("/app/output", showWarnings = FALSE)106write_csv(lb, "/app/output/lb.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-lb-character-results (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)1213forms = {}14study_of = {}15for r in odm.to_dicts():16    study_of[r["SubjectKey"]] = r["StudyOID"]17    key = (r["SubjectKey"], r["FormOID"], r["FormRepeatKey"])18    forms.setdefault(key, {})[r["ItemOID"]] = r["Value"]1920form_info = {21    "FO.LB_PROT": ("PROT", "Protein", 0),22    "FO.LB_CK": ("CK", "Creatine Kinase", 1),23    "FO.LB_GLUC": ("GLUC", "Glucose", 2),24    "FO.LB_KETON": ("KETON", "Ketones", 3),25    "FO.LB_CREAT": ("CREAT", "Creatinine", 4),26}272829def standardize(raw):30    if raw is None:31        return None32    t = raw.strip()33    low = t.lower().replace(" ", "")34    if low in ("negative", "neg"):35        return "NEGATIVE"36    if low in ("trace", "tr"):37        return "TRACE"38    if low in ("1+", "+1", "1plus"):39        return "1+"40    if low in ("2+", "+2", "2plus"):41        return "2+"42    return raw434445def as_number(text):46    if text is None or text == "":47        return None48    try:49        return float(text)50    except ValueError:51        return None525354rows = []55for (subj, formoid, formrep), items in forms.items():56    if formoid not in form_info:57        continue58    code, name, order = form_info[formoid]59    raw = items.get("IT.LB.RESULT")60    if raw is None or raw == "":61        continue62    lbdtc = items.get("IT.LB.LBDTC")63    unit = items.get("IT.LB.LBORRESU")64    unit = unit if unit not in (None, "") else None65    stresc = standardize(raw)66    # A true numeric result only: grades and censored values have no number.67    stresn = as_number(raw)68    if stresc in ("NEGATIVE", "TRACE", "1+", "2+"):69        stresn = None70    if raw.strip().startswith("<") or raw.strip().startswith(">"):71        stresn = None72    nrind = None73    if raw.strip().startswith("<"):74        nrind = "LOW"75    elif raw.strip().startswith(">"):76        nrind = "HIGH"77    rows.append(78        {79            "DOMAIN": "LB",80            "STUDYID": study_of[subj],81            "USUBJID": subj,82            "ORDER": order,83            "REP": int(formrep),84            "LBTESTCD": code,85            "LBTEST": name,86            "LBORRES": raw,87            "LBORRESU": unit,88            "LBSTRESC": stresc,89            "LBSTRESN": stresn,90            "LBSTRESU": unit,91            "LBNRIND": nrind,92            "LBDTC": lbdtc,93        }94    )9596lb = (97    pl.DataFrame(98        rows,99        schema={100            "DOMAIN": pl.String,101            "STUDYID": pl.String,102            "USUBJID": pl.String,103            "ORDER": pl.Int64,104            "REP": pl.Int64,105            "LBTESTCD": pl.String,106            "LBTEST": pl.String,107            "LBORRES": pl.String,108            "LBORRESU": pl.String,109            "LBSTRESC": pl.String,110            "LBSTRESN": pl.Float64,111            "LBSTRESU": pl.String,112            "LBNRIND": pl.String,113            "LBDTC": pl.String,114        },115    )116    .sort(["STUDYID", "USUBJID", "ORDER", "REP"])117    .with_columns(118        pl.int_range(1, pl.len() + 1).over(["STUDYID", "USUBJID"]).alias("LBSEQ")119    )120    .select(121        "DOMAIN",122        "STUDYID",123        "USUBJID",124        "LBSEQ",125        "LBTESTCD",126        "LBTEST",127        "LBORRES",128        "LBORRESU",129        "LBSTRESC",130        "LBSTRESN",131        "LBSTRESU",132        "LBNRIND",133        "LBDTC",134    )135)136137Path("/app/output").mkdir(parents=True, exist_ok=True)138lb.write_csv("/app/output/lb.csv")

1/* opensas reference solution for the yamaa benchmark sdtm-lb-character-results.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data odm;6  length StudyOID MetaDataVersionOID SubjectKey StudyEventOID StudyEventRepeatKey FormOID FormRepeatKey ItemGroupOID ItemGroupRepeatKey ItemOID Value $1024;7  infile "/app/input/odm.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input StudyOID : $char1024. MetaDataVersionOID : $char1024. SubjectKey : $char1024. StudyEventOID : $char1024. StudyEventRepeatKey : $char1024. FormOID : $char1024. FormRepeatKey : $char1024. ItemGroupOID : $char1024. ItemGroupRepeatKey : $char1024. ItemOID : $char1024. Value : $char1024.;9run;1011proc sort data=odm out=ordered; by StudyOID SubjectKey FormOID FormRepeatKey; run;12data wide;13  length LBORRES LBORRESU LBDTC $1024;14  retain LBORRES LBORRESU LBDTC;15  set ordered;1617  by StudyOID SubjectKey FormOID FormRepeatKey;18  if first.FormRepeatKey then do;19    LBORRES=''; LBORRESU=''; LBDTC='';20  end;21    if ItemOID='IT.LB.RESULT' then LBORRES=Value;22  if ItemOID='IT.LB.LBORRESU' then LBORRESU=Value;23  if ItemOID='IT.LB.LBDTC' then LBDTC=Value;24  if last.FormRepeatKey then output;25run;26data prepared; set wide; if not missing(LBORRES); length DOMAIN $2 STUDYID USUBJID LBTESTCD LBTEST LBSTRESC LBSTRESU LBNRIND TOKEN $1024;27DOMAIN='LB'; STUDYID=StudyOID; USUBJID=SubjectKey; LBSTRESU=LBORRESU; LBSTRESC=LBORRES; LBNRIND=''; LBSTRESN=.; TOKEN=lowcase(compress(LBORRES,' '));28if TOKEN in ('negative','neg') then LBSTRESC='NEGATIVE'; if TOKEN in ('trace','tr') then LBSTRESC='TRACE'; if TOKEN in ('1+','+1','1plus') then LBSTRESC='1+'; if TOKEN in ('2+','+2','2plus') then LBSTRESC='2+';29if prxmatch('/^[+-]?([0-9]+(\.[0-9]*)?|\.[0-9]+)([eE][+-]?[0-9]+)?$/',strip(LBORRES)) and not (LBSTRESC in ('NEGATIVE','TRACE','1+','2+')) then LBSTRESN=input(LBORRES,best32.);30if substr(strip(LBORRES),1,1)='<' then LBNRIND='LOW'; if substr(strip(LBORRES),1,1)='>' then LBNRIND='HIGH';31if FormOID='FO.LB_PROT' then do; LBTESTCD='PROT'; LBTEST='Protein'; TESTORD=1; end;32else if FormOID='FO.LB_CK' then do; LBTESTCD='CK'; LBTEST='Creatine Kinase'; TESTORD=2; end;33else if FormOID='FO.LB_GLUC' then do; LBTESTCD='GLUC'; LBTEST='Glucose'; TESTORD=3; end;34else if FormOID='FO.LB_KETON' then do; LBTESTCD='KETON'; LBTEST='Ketones'; TESTORD=4; end;35else if FormOID='FO.LB_CREAT' then do; LBTESTCD='CREAT'; LBTEST='Creatinine'; TESTORD=5; end; else delete;36REP=input(FormRepeatKey,best32.); run;37proc sort data=prepared; by STUDYID USUBJID TESTORD REP; run;38data result; set prepared; by STUDYID USUBJID; retain LBSEQ;39  if first.USUBJID then LBSEQ=0;40  LBSEQ+1;41run;4243proc sql;44  create table final as select DOMAIN, STUDYID, USUBJID, LBSEQ, LBTESTCD, LBTEST, LBORRES, LBORRESU, LBSTRESC, LBSTRESN, LBSTRESU, LBNRIND, LBDTC from result;45quit;46proc export data=final outfile="/app/output/lb.csv" dbms=csv replace; run;