Using a snowflake data model and autocompletion to support diagnostic coding in acute care hospitals
Noussa-Yao Joseph, Boussadi Abdelali, Richard Monique, Heudes Didier, Degoulet Patrice
PURPOSE: Efficient and adequate coding is essential for all hospitals to optimize funding, follow activity, and perform epidemiological studies. OBJECTIVE: We propose an autocompletion method for optimizing diagnostic coding in acute care hospitals. METHODS: Using a terminology snowflake model integrating SNOMED 3.5 and ICD-10 codes, autocompletion algorithms generate a list of diagnostic expressions from partial input concepts. RESULTS: A general autocompletion component has been developed and tested on a set of inpatient summary reports. Concepts expressed as strings of three or four characters return a noisy list of diagnostic labels or codes. Concepts expressed as groups of strings return lists that are semantically close to the labels present in hospital reports. The most pertinent information lies in the length of the expressions entered. CONCLUSION: Autocompletion can be a complementary tool to existing coding support systems.