PyHealth — healthcare AI toolkit
⚠️ AUTHORIZED USE ONLY — build, test, and validate clinical ML models with PyHealth 2.x: EHR datasets (MIMIC-III/IV, eICU, OMOP), clinical prediction tasks (mortality, readmission, drug recommendation), medical coding (ICD, NDC, ATC), and clinical deep learning models (RETAIN, SafeDrug, GAMENet, Transformer)
- What
- ⚠️ AUTHORIZED USE ONLY — build, test, and validate clinical ML models with PyHealth 2.x: EHR datasets (MIMIC-III/IV, eICU, OMOP), clinical prediction tasks (mortality, readmission, drug recommendation), medical coding (ICD, NDC, ATC), and clinical deep learning models (RETAIN, SafeDrug, GAMENet, Transformer)
- Cost
- Free
- Needs
- Python 3.12 or 3.13, install-free — run with `uv run --python 3.13 --with "pyhealth>=2.0.2,<2.1"` (no virtualenv to create; on a machine without a GPU, take torch from the CPU-only index, see step 5) — PyHealth 2.0.2 pins numpy 2.2 / torch 2.7 / transformers 4.53; GPU recommended for deep models; your own data — the MIMIC-IV, eICU-CRD and MIMIC-III demos and PyHealth's synthetic MIMIC-III are open and need no access; full MIMIC-III/IV and eICU need YOUR PhysioNet credentialed access and data use agreement (never copy restricted records into prompts or repos)
- Install
- Copy the installer prompt below into your Muse — your agent does the rest.
Curated by Skill Harbor — @alterlab-ieu's alterlab-pyhealth skill, listed here with credit to its creator (part of the AlterLab Academic Skills suite): a complete agent reference for clinical machine learning with the PyHealth 2.x healthcare AI toolkit. ⚠️ AUTHORIZED USE ONLY / ⚠️ USAGE AUTORISÉ UNIQUEMENT — Full MIMIC-III/IV and eICU data require the user's own PhysioNet credentialed access and a data use agreement (the MIMIC-IV, eICU-CRD and MIMIC-III demos and PyHealth's synthetic MIMIC-III are open and need no access); never copy restricted records into prompts, notebooks, or repositories the agreement does not cover. The skill covers the modular 5-stage pipeline (data loading, task definition, model selection, training with the Trainer, validation), version-gotcha documentation for the PyHealth 2.0 API rewrite (tasks as instantiable classes, explicit table lists, schema-driven models, metric strings without `_score` suffix), clinical use cases (ICU mortality prediction, safe medication recommendation, readmission prediction, sleep staging from EEG, medical code translation, clinical text to ICD coding), a complete runnable mortality-prediction workflow, and a serious clinical-validation section — calibration, conformal prediction, fairness metrics, interpretability, TRIPOD+AI reporting. Honest caveats: the models are research-grade risk estimates for qualified clinicians, not medical devices — deployment as a medical device falls under device regulation (e.g. FDA SaMD, EU MDR); deep models need thousands of patients and GPU; PyHealth 2.0.2 pins its stack (Python 3.12-3.13, numpy 2.2, torch 2.7) and runs install-free via uv run, its pins isolated per run. MIT licensed. Skill Harbor never reviews the code, review it yourself before use. Discovered via skills.sh.
Version:
Quick answer
PyHealth — healthcare AI toolkit is a free Muse skill that is a reference for clinical machine learning with PyHealth 2.x, covering the pipeline of loading, task definition, model selection, training, and validation. It covers ICU mortality prediction, medication, readmission, and ICD coding, with a validation section. Authorized access only, since MIMIC and eICU data need PhysioNet access and agreement.
Install
Copy the install package below, then paste it into Muse🧪 Install ran in our test sandbox · Nov 11, 2026📌 Pinned version: commit 6fa2616Step 0 — before installing anything: read this install prompt, summarize in a few lines what this build does and what installing it involves, then ask me to confirm before you proceed. Prerequisites: Python 3.12 or 3.13, install-free — run with `uv run --python 3.13 --with "pyhealth>=2.0.2,<2.1"` (no virtualenv to create; on a machine without a GPU, take torch from the CPU-only index, see step 5) — PyHealth 2.0.2 pins numpy 2.2 / torch 2.7 / transformers 4.53; GPU recommended for deep models; your own data — the MIMIC-IV, eICU-CRD and MIMIC-III demos and PyHealth's synthetic MIMIC-III are open and need no access; full MIMIC-III/IV and eICU need YOUR PhysioNet credentialed access and data use agreement (never copy restricted records into prompts or repos) Install "PyHealth — healthcare AI toolkit" for me. It gives my agent @alterlab-ieu's PyHealth 2.x reference: EHR dataset loaders (MIMIC-III/IV, eICU, OMOP), clinical prediction task classes (mortality, readmission, length of stay, drug recommendation), medical coding utilities (ICD/NDC/ATC/CCS via InnerMap/CrossMap), clinical deep learning models (RETAIN, SafeDrug, GAMENet, Transformer, GAT/GCN), the Trainer with metrics and monitoring, calibration/conformal-prediction/fairness/interpretability guidance, TRIPOD+AI reporting, and a complete runnable mortality-prediction workflow. Part of the AlterLab Academic Skills suite. MIT licensed. Repository: https://raw.githubusercontent.com/alterlab-ieu/alterlab-academic-skills/6fa26160d28f14ba3daf22a8cb30b91ea3d46837/skills/clinical-research/alterlab-pyhealth/SKILL.md Pinned version: install the code at commit 6fa26160d28f14ba3daf22a8cb30b91ea3d46837 exactly (Skill Harbor tested this version). Do not substitute a newer version. If this commit is unreachable, stop and tell me. Updated November 5, 2026: Skill Harbor regenerated this install prompt from the current upstream file. What changed since the version the previous prompt described: installation is now install-free — `uv run --python 3.13 --with "pyhealth>=2.0.2,<2.1"` instead of creating a virtualenv with `uv venv` and installing with `uv pip install`, with a CPU-only torch index on machines without a GPU (the `<2.1` pin is because PyHealth 2.1.0, released 2026-10-17, is not yet checked against this skill); and the demo data is now open — the MIMIC-IV, eICU-CRD and MIMIC-III demos and PyHealth's synthetic MIMIC-III need no PhysioNet access, while credentialed PhysioNet access and a signed data use agreement are still required for the full MIMIC-III/IV and eICU cohorts. The skill content itself (loaders, task classes, models, Trainer, validation guidance) is unchanged. 1. Fetch the SKILL.md file (and any helper files) from the repository path into a temporary folder 2. Safety check: review the SKILL.md and scripts for anything suspicious (unexpected network calls, shell commands, credential harvesting). This repo should contain zero secrets in code — full-dataset access is through my own credentialed PhysioNet account, while the demo data is open and needs no login. Verify that holds here; STOP on any red flag and tell me. 3. Install it as a skill: copy SKILL.md and its helper files into the agent's skills directory, in a folder named "alterlab-pyhealth". 4. Verify with no network calls: frontmatter valid, files in place. 5. Report what was installed, where, and what I still need to do myself (e.g. run with `uv run --python 3.13 --with "pyhealth>=2.0.2,<2.1"`, adding the CPU-only torch index (`--index https://download.pytorch.org/whl/cpu --index-strategy unsafe-best-match`) on a machine without a GPU; the MIMIC-IV, eICU-CRD and MIMIC-III demos and PyHealth's synthetic MIMIC-III are open and need no access, while full MIMIC-III/IV and eICU still need PhysioNet credentialing and a signed data use agreement; split data by patient; treat outputs as research-grade risk estimates, not clinical decisions). GitHub is optional: if I have a GitHub account or the gh CLI, you may use it; otherwise public access is fine. Never require it unless it's in the prerequisites above. Rules: don't touch anything outside the temp folder and the install target. If anything looks off, stop and ask me.
Saved to your recent installs. Find it anytime on /connect.
🔒 Security, provenance and what this touches
The install prompt below already includes the vetting steps: your agent follows the community checklist before installing anything with executable code. Want more?
Verification certificate
Dated facts about what was checked for this listing, and how. This is not a safety certification.
- Pinned version
- 6fa2616 · 2026-10-13
- Sandbox install test
- Install ran in our test sandbox · Nov 11, 2026
- Creator identity
- Not verified
- Fingerprint watch
- Last checked · Nov 19, 2026 : no change detected in the files linked by this install prompt
- Install prompt analysis
- 2026-10-13 : No commands, network calls, packages, file writes or API keys detected.
Skill Harbor never reviews the code of a listed skill. These are dated facts about the install prompt and our install test, not a safety certification. Review a skill yourself before you install it. How we verify
Questions
How do I install a skill?
Every product page includes a copy-paste install prompt. Paste it into your Muse and it sets the skill up for you — no manual configuration.
Where does my money go?
Straight to the seller. Skill Harbor never processes payments: checkout happens on the seller’s own page, usually Stripe.
What does the ✓ next to a creator’s name mean?
It means we confirmed the identity of the person behind the listing. It says nothing about the code itself — always check a skill before installing it.