openIMIS-AI - 1. Input data model

This page presents the analysis of AI algorithm’s input data model. First, we analyse the openIMIS data structure and fields representing the information used by the Medical Officer to adjudicate a claim. Second, we look at the openIMIS-FHIR mapping and extract the required FHIR resources to feed the AI algorithm during the learning, development and test processes. Because the openIMIS FHIR module do not map all openIMIS fields to FHIR resources, we are limited by the current FHIR-openIMIS R4 mapping.

openIMIS Claim Review Web Page

Open

openIMIS Database Tables

The following database diagram displays openIMIS tables and their relations. The analysis is centered on Claim and identifies other entities in relation to a Claim and used in the adjudication process.

Open

FHIR Resources Definition

From all the fields that the Medical Officers are using when reviewing claims, we mainly need to identify entities in relation to the Claim. From some entities (Diagnosis, HF), we might not require more than the identification fields.

The fields that will be taken into consideration from openIMIS database will be limited by the current openIMIS to FHIR mapping. More particular, the ClaimResponse will be used for the development of the AI algorithm and for production use. This resource is required to have the adjudication results for already categorized claims by a medical officer for development stage and to have the static validation response (automatic checking step) when categorizing the claims through the AI model.

In the following table, we list FHIR and openIMIS resources that will be considered as the input data model for the development of the AI algorithm and model.