openIMIS AI-based Claim Adjudication

Digital Square D1 proposal: Claim categorization using Artificial Intelligence: a proof of concept


  • Swiss TPH (@Dragos Dobre @Simona Dobre @Siddharth Srivastava

  • SolDevelo (@Kamil Madej (Deactivated) @Damian Borowiecki )


The goal of this project is to develop an automatic claims categorization module for openIMIS based on state-of-the-art Artificial Intelligence (AI) algorithms, standards, and methodologies which will drastically reduce the manpower, resources and time required to review a reimbursement claim. This process is centered on the claim response process. Here are some details about this process.

Current openIMIS claim adjudication process

A claim is what a doctor or health service submit to the patient insurance company so that they are paid for the medical services and/or items (medical prescriptions). Thus, a claim can be composed of one or more medical services and items. After the visit to the Health facility, the claim will be submitted to the insurance by the health facility administrator (called also Claim Administrator). After submission, a claim processor – based on a rule engine that takes in consideration openIMIS configuration (insurance product, medical items and services, insuree, policy, etc.) – will check it for completeness, accuracy and whether the service/item is covered by the patient insurance.  If all the services and items contained in a claim are rejected by the Rule Engine, the HF will receive a negative claim response for the respective claim. All the items and services that were statically validated by the Rules engine may be subject to a manual evaluation by a Medical Expert, which can accept, partially accept or reject the selected items/services. As a claim is composed by one or several medical services and/or items, the output of the process can be:

  • Accepted item/service after static validation and manual evaluation;

  • Rejected item/service after static validation or manual evaluation.

If all the items and services related to a specific claim are accepted, the respective claim can be considered as accepted, whereas if part of the items/services are accepted, the claim will be considered as partially accepted. If all the items and services related to a claim are rejected, the HF will receive a negative response for that claim.

The openIMIS implementation in Nepal receives up to 14 000 claims per day, as it can be observed in Fig 1. From all these claims, the team composed of five Medical Officers are able to review only around 1,000 claims per day.

Fig. 1 Evolution of the number of claims/day through the openIMIS system in Nepal

As the Medical Officers can not physically check all the submitted claims, the Claim module was implemented in openIMIS in order to select claims (accepted or partially accepted by the Rules Engine) to be further reviewed by the Medical Officers. This manual claim adjudication process is illustrated in Fig. 2.

Fig. 2 Manual claim adjudication workflow

Future openIMIS AI-based automatic claim adjudication process (vision)

For the AI claim categorization algorithm, the input is represented by a service or item, which has been statically validated by the Rules engine. The output of such a model will be the acceptance or rejection of the item or service. This AI module will allow the Medical Officer to concentrate only on those claims that really need to be reviewed, such as inconsistent or erroneous. Furthermore, in order to detect false positive or false negative cases misclassified by the algorithm, a Quality Assurance module will then select AI-categorized claims to be review by the Medical Officers. This process is illustrated in Fig. 3.

Fig. 3 Claim adjudication workflow with AI algorithm

Project realization

The following wiki pages present the project realization:



This page in other languages







Did you encounter a problem or do you have a suggestion?

Please contact our Service Desk

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.