Summary
This guide walks you through Linkr end-to-end: workspace, project, OMOP database import, cohort building, analytical dataset creation and then a dashboard. Expect 15 to 20 minutes with a demo dataset.
Before you start
Make sure you have a Linkr instance ready, either online or local. See the quickstart or local install.
For this walkthrough we’ll use the MIMIC-IV demo OMOP dataset (100 patients, OMOP CDM v5.4), shipped as demo data — no import needed.
How the demo data arrives
It depends on the mode. On the online demo and in a client-only build, the demo workspace is already there on first launch. On an instance with a backend, the setup wizard’s third step installs it — and if you chose “Start empty”, you can install it at any time from the Catalog page. Running client-only locally, run npm run data:fetch once to fetch it.
What about connecting to a real hospital warehouse?
Client-only mode works on local files (extracts, demo datasets). To query a relational warehouse (PostgreSQL, SQL Server, Oracle…) directly, you need full-stack mode. See Deployment modes.
Long vs wide format
An OMOP database is in long format: one row per clinical event (one measurement, one diagnosis, one prescription…). That’s the opposite of an Excel spreadsheet, which is in wide format (one row per patient). Linkr is specifically designed to turn long into wide without writing SQL. For more on this distinction, see Data organization — long vs wide format.
Guided walkthrough
Opening the demo and a project is described in the quickstart. This guide picks up from data exploration.
Open the demo database
Once the demo data is in place, you'll find the MIMIC-IV demo OMOP dataset (100 patients) in Data Warehouse → Databases, marked as active (used by default on all warehouse pages in the project).
MIMIC-IV Demo (OMOP)
DuckDB / OMOP CDM 5.4
31 Parquet files
Explore concepts
In Data Warehouse → Concepts, browse the concepts present in the database, following the dictionary it declares (on an OMOP database: Condition, Drug, Measurement…). Click a concept to view its statistics (patient count, distribution…). This helps identify the codes you'll use in the cohort.
| Concept ID | Concept Name | Concept Code | Standard Concept | Records | Patients |
|---|---|---|---|---|---|
| 3013682 | Lactate [Moles/volume] in Blood | 32693-4 | Yes | 1,842 | 78 |
| 3016723 | Creatinine [Mass/volume] in Serum or Plasma | 2160-0 | Yes | 2,184 | 96 |
| 3000963 | Hemoglobin [Mass/volume] in Blood | 718-7 | Yes | 1,576 | 92 |
| 132797 | Sepsis | 91302008 | Yes | 312 | 41 |
| 201826 | Type 2 diabetes mellitus | 44054006 | Yes | 268 | 34 |
| 1321341 | Norepinephrine | 7512 | Yes | 487 | 52 |
Build a cohort
In Data Warehouse → Cohorts, create a new cohort by stacking criteria (age, sex, concepts, period, length of stay…) into a logical tree (AND, OR, NOT). Linkr generates the corresponding SQL automatically, which you can inspect and copy. Example: adult patients (age ≥ 18) with a sepsis diagnosis admitted between 2015 and 2020.
Explore a dataset
In Lab → Datasets, you can analyse data in wide format (one row per patient, one column per variable) — the format clinicians and statisticians are used to.
| patient_id | age | sex | sofa_score | length_of_stay | mortality |
|---|---|---|---|---|---|
| P-0001 | 67 | M | 8 | 4.2 | Alive |
| P-0002 | 54 | F | 12 | 9.7 | Deceased |
| P-0003 | 71 | M | 5 | 2.1 | Alive |
| P-0004 | 49 | F | 7 | 6.4 | Alive |
| P-0005 | 82 | M | 11 | 14.3 | Deceased |
| P-0006 | 35 | F | 3 | 1.8 | Alive |
Visualise the distribution of each column, spot missing values and explore your datasets in a few clicks.
From long to wide format: where Linkr creates the link
Today, turning long format (the OMOP warehouse) into wide format (one row per patient) cannot yet be done through the graphical interface: the transformation has to be written in the built-in IDE (Python or R, against the loaded dataset). This is exactly where Linkr creates the link: the same app brings together, side by side, the low-code tools clinicians use and the programming environment data scientists need. Each works at their preferred level of abstraction, on the same project, with the same data — combining the strengths of both worlds in service of health data.
Run an analysis, build a dashboard
On your dataset, run built-in analyses (Tableau descriptif, Indicateur clé, Constructeur de graphiques) without writing code, then compose a dashboard from the resulting widgets. The dashboard is saved automatically inside the project.
Table 1 — cohort characteristics
| Variable | Overall (n=400) | Survivors (n=312) | Deceased (n=88) | p |
|---|---|---|---|---|
| Age (years) | 63.4 ± 16.8 | 61.2 ± 16.1 | 71.5 ± 15.4 | <0.001 |
| Sex — Male (%) | 54.5 | 53.8 | 57.0 | 0.62 |
| SOFA score | 6.8 ± 3.2 | 5.9 ± 2.7 | 10.1 ± 3.5 | <0.001 |
| Length of stay (d) | 5.3 (2.1–9.7) | 4.6 (1.9–8.4) | 8.1 (3.2–14.6) | <0.001 |
Describe your cohort with a Table 1 — no code required.
Save the project
In Versioning → Export, download the project as a ZIP archive. The ZIP follows a canonical structure (scripts/, cohorts/, datasets/, etc.) and can be re-imported into another Linkr instance. On an instance with a backend, the same Versioning tab also lets you link the project to a GitLab or GitHub repository and push your changes to it.
Export
Download your project as a ZIP archive with a git-friendly folder structure.
Data files are excluded by default for security. Enable this only if you need to share the actual dataset contents.
Recap
In six steps you have:
- Opened the MIMIC-IV demo database.
- Explored the database’s concepts.
- Built a visual cohort with generated SQL.
- Explored a dataset in wide format.
- Run an analysis and assembled a dashboard.
- Saved everything through ZIP export.
You did not write a single line of code. If you’re comfortable with Python or R, you can go further in the built-in IDE (the project’s IDE page) to write custom analyses directly on the dataset.
Next steps
- Workspaces and projects — how Linkr organises content, and what belongs to what.
- The data pipeline — understand the Data Warehouse → Pipeline → Lab flow.
- Dashboards — build a complete presentation, tab by tab.
- Deployment modes — worth reading if you’re considering a shared instance with a backend.