INSPIRE-MEDS

The INSPIRE dataset is a publicly available research dataset in perioperative medicine, which includes approximately 130,000 cases (50% of all surgical cases) who underwent anesthesia for surgery at an academic institution in South Korea between 2011 and 2020. This comprehensive dataset includes patient characteristics such as age, sex, American Society of Anesthesiologists physical status classification, diagnosis, surgical procedure code, department, and type of anesthesia. It also includes vital signs in the operating theatre, general wards, and intensive care units (ICUs), laboratory results from six months before admission to six months after discharge, and medication during hospitalization. Complications include total hospital and ICU length of stay and in-hospital death. This pipeline extracts the INSPIRE dataset (from physionet, https://physionet.org/content/inspire/) into the MEDS format.

Usage:

pip install INSPIRE_MEDS
export DATASET_DOWNLOAD_USERNAME=$PHYSIONET_USERNAME
export DATASET_DOWNLOAD_PASSWORD=$PHYSIONET_PASSWORD

meds-extract-run spec=INSPIRE output_dir=$MEDS_OUTPUT_DIR download_dest_dir=$RAW_INPUT_DIR

This one command gives the complete build. It downloads the raw INSPIRE files into $RAW_INPUT_DIR, builds a small derived table (the inspire_subject_table pre-stage; see Birth and death), runs the eight-stage MEDS-Extract pipeline, which also describes the codes (see Code metadata), and then fills the official ICD titles and the vocabulary parents into metadata/codes.parquet with suMEDS (the sumeds_enrich post-stage, installed with this package; see Enriching code metadata with suMEDS). The MEDS dataset is written to $MEDS_OUTPUT_DIR. To reuse a download you already have:

meds-extract-run spec=INSPIRE output_dir=$MEDS_OUTPUT_DIR do_download=false input_dir=$RAW_INPUT_DIR

Run meds-extract-run --help for every option. Rerunning the same command on an existing output only runs the stages that have not finished (each finished stage leaves .logs/<stage>.done in the output).

MEDS_extract-INSPIRE is deprecated. It was the complete build up to 1.0.0. It is kept for one more release as an alias that prints the equivalent meds-extract-run command and runs it (MEDS_extract-INSPIRE ROOT builds into ROOT/MEDS_cohort from ROOT/raw_input, as before, and --no-enrich becomes skip_stages=[sumeds_enrich]). It no longer accepts suMEDS options after a --; set them as stage options instead.

Configuration

The whole pipeline is one file, src/INSPIRE_MEDS/messy.yaml, registered under the MEDS_extract.pipelines entry-point group. Its etl: block names the inspire_subject_table pre-stage (see Birth and death) and the sumeds_enrich post-stage, and the release's license and PhysioNet page, which MEDS-Extract writes into metadata/dataset.json.

Pseudo-timestamps

INSPIRE ships no wall-clock times. Every time column is an offset in minutes, anchored per patient at that patient's own first hospital admission — min(admission_time) is 0 for all 99,886 subjects. Each patient's anchor is then placed at one shared literal origin, the midpoint of the study window (2011-01-01 .. 2020-12-31) = 2016-01-01.

So only relative differences within a subject are meaningful. The absolute dates in the MEDS output are not real dates, and comparing them across subjects is meaningless.

Null handling

A null component of a composite code drops the row since MEDS-Extract 0.7, where 0.6.x rendered it as the literal UNK and kept the row. Components are therefore coalesced with ?? 'UNK' wherever a null must not drop the row.

Two are deliberately left uncoalesced. route on medications is never null in the release (0 of 10,854,338 rows), so a coalesce would be dead code. The demographic codes drop rather than mint an UNK category, because a missing demographic is not an observed value.

Birth and death

MEDS_BIRTH and MEDS_DEATH are properties of a subject, but INSPIRE records what they derive from on operations, which is one row per operation:

• age is the age on the operation date, quantised to 5-year bins, so a patient with several operations in one admission can straddle a bin boundary and carry two ages.
• inhosp_death_time appears only on the rows of the admission during which the patient died, while allcause_death_time is invariant across all their rows.

MESSY cannot reduce a table over its own rows — self-joins are rejected and dftly is strictly row-wise — so the inspire_subject_table pre-stage (src/INSPIRE_MEDS/pre_MEDS.py, registered under the MEDS_transforms.stages entry-point group) reduces operations to one row per subject, and the config joins it back. That is the only Python step this package ships. meds-extract-run runs it before convert_to_parquet; it writes subjects.parquet into its own directory ($MEDS_OUTPUT_DIR/inspire_subject_table), which MEDS-Extract searches before the raw input, so the checksum-verified download is never modified.

Code metadata

metadata/codes.parquet carries code, description, parent_codes and code_template. Up to 1.0.0 only the lab and vital codes had a description (from the release's parameters.csv) and no code had a parent. The ETL now describes every code family whose meaning INSPIRE itself documents, except the diagnoses, and writes no parent codes: the ICD-10-CM category titles and the vocabulary parents come from suMEDS, whose sumeds_enrich stage runs last (see Enriching code metadata with suMEDS). Nothing in the event data changes.

A literal code has no components to join a dictionary on, so its description is a constant _self block on the event that emits it (MEDS-Extract 0.8.1 or later), e.g. description: '"Patient entered the operating room (INSPIRE operations.orin_time)"'.

Enriching code metadata with suMEDS

Vocabulary work is maintained once, in suMEDS, rather than in every ETL. This package ships the rules for it, src/INSPIRE_MEDS/sumeds_rules.yaml, registered in suMEDS's sumeds.rules entry-point group under the same name as the MEDS-Extract spec, INSPIRE, which is how the spec's sumeds_enrich: {rules: INSPIRE} finds them; --rules INSPIRE works for any suMEDS command. The sumeds_enrich post-stage enriches metadata/codes.parquet in place, after the ETL's own descriptions; the table is replaced only once the enriched one is complete, and it stays valid MEDS code metadata. By default the ICD data comes from suMEDS's ontology packs, fetched from GitHub on first use and cached; no Athena download is needed.

Enrichment is on by default and soft-fails: if it cannot finish, for example on a server without internet access, metadata/codes.parquet is left as the ETL wrote it and the build still succeeds. The warning, with the command that enriches it later, is in $MEDS_OUTPUT_DIR/.logs/pipeline.log. metadata/dataset.json tells whether a build was enriched: sumeds_enrichment.enriched is true or false, next to the rules, options and coverage used, and the error if it failed.

jq .sumeds_enrichment.enriched "$MEDS_OUTPUT_DIR/metadata/dataset.json"

Finish the enrichment later. MEDS-Extract marks the stage done even when it soft-fails, so rerunning meds-extract-run does not try again. Run suMEDS-enrich on the build instead; it reuses the rules and options recorded in dataset.json:

suMEDS-enrich "$MEDS_OUTPUT_DIR"
# Air-gapped: from a copy of the sumeds-ontologies packs directory instead of GitHub
suMEDS-enrich "$MEDS_OUTPUT_DIR" --ontology-source /path/to/sumeds-ontologies/ontologies --offline

Skip it. skip_stages=[sumeds_enrich] builds without the enrichment and without a record. That build, or one made with an earlier release of this package, has no record to reuse, so add --rules INSPIRE to suMEDS-enrich later:

meds-extract-run spec=INSPIRE output_dir=$MEDS_OUTPUT_DIR do_download=false input_dir=$RAW_INPUT_DIR \
	skip_stages=[sumeds_enrich]

Stage options for one run. meds-extract-run passes overrides= to every stage, so set a sumeds_enrich option (the suMEDS-enrich options, spelled with underscores) with ++stage_cfg.: the ++ adds the key to the stages that do not declare it, which ignore it. strict makes a failed enrichment fail the run (no .done marker, so the same command retries the stage):

# Fail the run if the enrichment fails
meds-extract-run spec=INSPIRE ... "overrides=['++stage_cfg.strict=true']"
# Also validate the ATC codes against your own OHDSI Athena download
meds-extract-run spec=INSPIRE ... "overrides=['++stage_cfg.athena_csv=/path/to/athena']"
# Use only already cached packs (no network), from a local pack directory
meds-extract-run spec=INSPIRE ... "overrides=['++stage_cfg.offline=true','++stage_cfg.ontology_source=/path/to/packs']"

SUMEDS_ONTOLOGY_SOURCE=/path/to/packs in the environment sets the pack source too.

Clusters without internet on compute nodes. Fill a shared cache once, from a login node, then run offline on the compute nodes with the same XDG_CACHE_HOME:

export XDG_CACHE_HOME=/shared/project/cache
suMEDS-enrich --prefetch --rules INSPIRE # login node: downloads the pinned pack release
meds-extract-run spec=INSPIRE output_dir=$MEDS_OUTPUT_DIR do_download=false input_dir=$RAW_INPUT_DIR \
	"overrides=['++stage_cfg.offline=true']" # compute node: no network requests

ICD-10-CM and ICD-10-PCS codes are validated against suMEDS's versioned ontology data (sumeds-ontologies, pinned release), built from the CMS and CDC files. The rules add:

• Diagnoses (inspire-icd10cm-categories). INSPIRE documents its diagnoses as ICD-10-CM, cut to the first three characters, with a list of sensitive categories removed (icd10_excluded.csv). A code that is an ICD-10-CM category gets the CMS title and the parent ICD10CM//<category>. Codes that ICD-10-CM does not have (likely codes from the Korean KCD that were not converted) get neither.
• Operations (inspire-icd10pcs-prefixes). INSPIRE documents icd10_pcs as the first five characters of an ICD-10-PCS code (section, body system, root operation, body part, approach), mapped by hand from the operation name. A valid five-character prefix gets the parent ICD10PCS//<prefix> and its shorter prefixes (OMOP holds partial codes as ICD10PCS Hierarchy concepts).
• Medications (inspire-atc-medications). The parent is the dataset's own ATC code, ATC//<atc_code>, when it has the shape of an ATC code of any level. ATC is licensed, so suMEDS only tags it, unless your Athena download validates it and adds its ancestors.
• Same-dataset parents (inspire-medication-atc-prefixes). A medication code also gets the nearest shorter ATC code given by the same route that INSPIRE contains, e.g. MEDICATION//N02BE//IV for MEDICATION//N02BE01//IV. Diagnosis and operation codes all sit at one level (three and five characters), so they have no same-dataset parents.

suMEDS fills only missing descriptions, never removes or rewrites existing metadata, writes parents as VOCABULARY//CODE, and names them in parent_names, a map from parent code to name. A build with skip_stages=[sumeds_enrich] has none of these parent codes or the diagnosis descriptions.

Parallelism

To convert the full release faster, give meds-extract-run a runner file; its parallelize: block becomes every stage's default (this uses the hydra-joblib-launcher package, the local_parallelism extra):

pip install "INSPIRE_MEDS[local_parallelism]"
printf 'parallelize:\n  n_workers: 8\n  launcher: joblib\n' >runner.yaml
meds-extract-run spec=INSPIRE output_dir=$MEDS_OUTPUT_DIR do_download=false \
	input_dir=$RAW_INPUT_DIR stage_runner_fp=runner.yaml

Summary statistics

tests/regression/full_summary_stats.json is a meds_summary_stats fingerprint of the MEDS dataset this ETL builds from the full INSPIRE 1.4.2 release with meds-extract-run spec=INSPIRE, so its code-metadata coverage includes the suMEDS enrichment. It covers subject, event and measurement counts, digests of the code vocabulary, per-subject, value and time distributions, code-metadata coverage, the splits and the schema. It names no code and reports no cell of fewer than 20 subjects and no minimum or maximum, which is what makes it safe to publish. Because it is committed next to the ETL, its diff shows what an ETL change did to the output, and two ETL versions can be compared without access to the data.

Regenerate it in every pull request that changes the output, from a build of that branch:

# Build the release with this checkout, then summarize it
scripts/summary_stats.sh full --input-dir $RAW_INPUT_DIR
# Or summarize a build of this checkout you already have
scripts/summary_stats.sh full --meds-dir $MEDS_OUTPUT_DIR

Both print how the statistics changed against the committed baseline before overwriting it. Add --compare to check a build without writing anything; it exits 1 when tests/regression/policy.yaml flags an error. Rebuilding the same commit reproduces the baseline exactly, so any finding is a change in the output. CI cannot rebuild the credentialed release, so the Summary stats workflow only warns when a pull request changes src/ (the config, the pre-stage or the suMEDS rules) without regenerating the baseline. A MEDS-Extract, suMEDS or polars upgrade can also move the output, so check it with --compare.

Citation

Please cite the following original paper when using the INSPIRE dataset:

@article{PhysioNet-inspire-1.4.2,
  author = {Lim, Leerang and Lee, Hyung-Chul},
  title = {{INSPIRE, a publicly available research dataset for perioperative medicine}},
  journal = {{PhysioNet}},
  year = {2026},
  month = jun,
  note = {Version 1.4.2},
  doi = {10.13026/1eay-yc85},
  url = {https://doi.org/10.13026/1eay-yc85}
}

Also cite the ETL framework used to extract the dataset:

@software{van_de_Water_INSPIRE_MEDS_ETL_2025,
author = {van de Water, Robin Philippus},
doi = {10.5281/zenodo.14891940},
license = {MIT},
month = feb,
title = {{INSPIRE\_MEDS ETL}},
url = {https://github.com/rvandewater/INSPIRE_MEDS},
year = {2025}
}

For citing MEDS in general:

@article{mcdermott2026meds,
  title={MEDS—An Emerging Data Standard and Ecosystem for Health AI Research},
  author={McDermott, Matthew BA and Steinberg, Ethan and Fries, Jason A and van de Water, Robin P and Pang, Chao and Rockenschaub, Patrick and Renc, Pawel and Oh, Jungwoo and Stankevi{\v{c}}i{\=u}t{\.e}, Kamil{\.e} and Xu, Justin and others},
  journal={NEJM AI},
  volume={3},
  number={6},
  pages={AIra2501253},
  year={2026},
  publisher={Massachusetts Medical Society}
}