... | ... | @@ -9,11 +9,15 @@ Each hospitalization is coded as follow: |
|
|
|
|
|
|
|
|
## Conditions in the clinical datawarehouse
|
|
|
We have approxymately 15.000 different conditions coded in ICD-10 of which 75% had already a mapping to a standard concept.
|
|
|
We have approxymately 15.000 different conditions coded in ICD-10 of which 75% had already a mapping to a standard concept.
|
|
|
|
|
|
![conditions_distribution](./images/cumulative_percentage_conditions.png)
|
|
|
|
|
|
When we look at the frequency distribution of conditions, we see that 1.111 ICD-10 codes represent 80% and 2.217 represent 90% of the most frequent conditions at the Bordeaux University Hospital.
|
|
|
|
|
|
## Mapping strategy
|
|
|
We selected the unmapped source codes among the 90% most frequent conditions and mapped them manually using Usagi.
|
|
|
We selected the unmapped source codes among the 90% most frequent conditions (N = 496) and mapped them manually using Usagi.
|
|
|
|
|
|
![usagi_conditions_mappings](./images/Usage_conditions_90_mappings.png)
|
|
|
|
|
|
We filtered by domain "condition" but we realized afterwards that the unmapped ICD-10 codes are mainly not actual conditions but finding like "history of drug allergy" etc... |