Attention ! Gitlab fournissant maintenant nativement des certificats Let’s Encrypt aux domaines personnalisés des Gitlab Pages, nous avons coupé notre service qui le faisait automatiquement pour vous.

Il est impératif, pour que votre domaine personnalisé continue à avoir un certificat Let’s Encrypt à jour, d’activer la fonctionnalité native dans les paramètres de votre projet. Cette activation remplacera votre certificat actuel par un nouveau certificat Let’s Encrypt géré par Gitlab.

Voir les détails sur https://docs.framasoft.org/fr/gitlab/gitlab-pages-le.html

Commit fdf70d52 authored by Matthieu's avatar Matthieu
parents 46d50b64 2a3189bc
......@@ -10,7 +10,19 @@ Once developed, these algorithms will then be transmitted to partner hospitals f
# CHALLENGE
All details of a patient's hospitalization are kept by the hospital teams. In some cases, the patient is not well identified (name misprint, human error...) and as a result his history is fragmented into several medical records: this is the concept of duplicated patient. This could cause harm to patients and result in significant costs. In addition, these duplicated individuals affect the quality of data and the results of impact analysis algorithms. Currently, duplicate candidate patients are searched manually and the process could be improved by algorithms that identify them. Ultimately, the identified duplicate patients could be merged with a great benefit for care, cost and research.
The challenge is to provide the best opensource algorithm to identify duplicated patients in a hospital in a given database.
The challenge is to provide the best opensource algorithm to identify duplicated patients in a hospital in a given database. This can be related to an *entity resolution* task and some interesting methods can be found in those used in *record linkage*.
## Example
My name is Paul DURANT and my birthdate is 02/01/1990.
Here is two examples of duplicate patients :
- my name may be misprint Paul DURAND
- my birth date may be reversed : 01/02/1989.
Then the system treats these as separate patients.
Sometimes databases do not have the identifying information present (ex : the centralized French healthcare system called #SNDS).
We need to invent a way to find duplicate patients with procedures or diagnoses...
# DATA
......@@ -38,7 +50,7 @@ Once a database has been converted to the OMOP CDM, evidence can be generated us
### csv format
- Header of the columns
- UTF8 character set
- UTF-8 character set
- Comma field separator
### 6 clinical csv files in OMOP format :
......@@ -49,6 +61,9 @@ Once a database has been converted to the OMOP CDM, evidence can be generated us
- CONDITION_OCCURRENCE.csv
- OBSERVATION.csv
Important remark : you have to use the standard concepts (_concept_id) for all the tables except for PROCEDURE_OCCURRENCE.
In PROCEDURE_OCCURRENCE usese local concepts - *_source_concept_id (concept_id) or *_source_value (concept_code) - provide in CCAM_CONCEPT_EXTENSION.csv
### 1 extension vocabulary csv file in OMOP format :
- CCAM_CONCEPT_EXTENSION.csv
......@@ -89,6 +104,18 @@ The csv to provide is:
"4,5,6"
```
## METRICS
We will evaluate this entity resolution task with classical [precision/recall/F1](https://en.wikipedia.org/wiki/Precision_and_recall) statistics:
$$precision = \frac{tp}{tp+fp}$$
$$recall = \frac{tp}{tp+fn}$$
$$F1 = 2 \cdot \frac{precision \cdot recall}{precision + recall}$$
Where tp are the true positive links, fp, the false positive links (links that are found in your submitted `PERSON.csv` but not present in the gold standard) and fn the false negative links (links that are not in you submitted `PERSON.CSV` but are in the gold standard).
# WHO
We aim to bring together clinicians, data scientists and innovators in healthcare to address current problems in healthcare with data analytics technologies.
You want to learn medical interoperability on real-life dataset. You have a personal computer and a little time in July and August. This challenge is for you!
......@@ -98,7 +125,7 @@ For the glory
# SUBMISSION TIMELINE
- Start Date: 10/07/2019
- OMOP-CDM presentation and challenge presentation: 22/07/2019
- [Video](https://video.ploud.fr/videos/watch/c18fd8b8-5e02-4e15-8db4-ba0e9293ff7e) of OMOP-CDM and challenge presentation (french) : 22/07/2019
- Team submission Deadline: 30/07/2019
- first algorithm submission Deadline: 01/08/2019 (once a day)
- final algorithm submission Deadline: 20/08/2019
......@@ -106,5 +133,6 @@ For the glory
# RESOURCES
- [forum](interchu.zulipchat.com) : forum for personal registration and team submission
- [omop wiki](https://github.com/OHDSI/CommonDataModel/wiki)
- [The book of OMOP](https://ohdsi.github.io/TheBookOfOhdsi/CommonDataModel.html#design-principles)
- online terminologies' explorer : [omop-mapper](omop-mapper.fgh.ovh) or [athena](athena.ohdsi.org)
- [synPUF data visualization](http://www.ohdsi.org/web/atlas/#/datasources/SYNPUF1K/dashboard)
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment