Computer Assisted Assignment of ICD Codes for Primary Admission Diagnostic in ICUs
| Autor: | Carlos Rojas, Giovanny Quiazúa, Francisco Gómez, Darwin Martinez, Javier Ordoñez, Cesar O Enciso-Olivera |
|---|---|
| Rok vydání: | 2017 |
| Předmět: |
medicine.medical_specialty
business.industry Real-time computing Psychological intervention medicine.disease Task (project management) 03 medical and health sciences 0302 clinical medicine Intervention (counseling) Intensive care Epidemiology Medicine 030212 general & internal medicine Medical emergency Medical diagnosis business Recovery approach 030217 neurology & neurosurgery Natural language |
| Zdroj: | Communications in Computer and Information Science ISBN: 9783319665610 |
| DOI: | 10.1007/978-3-319-66562-7_16 |
| Popis: | The intensive care units (ICUs) provide a constant monitoring and specialized support to patients with acute critical conditions, assuring timely interventions to rapid changes. A major determinant of the patient care in ICUs is the primary admission diagnosis. A typical diagnosis includes a nosological entity or syndrome name, with the possibility to describe the clinical condition and the patient health state. This diagnosis is the starting point to establish intervention plans and to devise epidemiological studies. In the ICU physicians are in charge to define this diagnosis. Diagnoses in ICUs are commonly described in natural language. However, a common practice is to assign a normalized code from the international classification of diseases and related health problems (ICD). Unfortunately, this codification task is time expensive and requires highly specialized medical knowledge. In this work, we introduce a text mining system to automatically recover ICD codes for diagnosis of admission in ICUs. The system is based on a novel hierarchical recovery approach which is well suited to representation used in the ICD code. The proposed approach was evaluated by using a set of 1206 codified descriptions written in Spanish language corresponding to diagnoses in an real ICU. The results suggest that this approach may account for a considerable percentage of the diagnoses codified by the expert in the ICU. In particular, the F1 measure was 0.21 ± 0.06 with a mean precision average of 0.3. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|