Automation of Health Superintendence's claims

The challenge

The Superintendency of Health is a public agency whose goals include ensuring compliance with the rights of citizens in relation to health care providers and Isapres. Annually, about 45,000 claims of various kinds are received, which take an average of 200 days to be resolved. We identified those claims with the highest volume and which are the most susceptible to being automated. The Base Price Increase constitutes close to 50% of the claims made by citizens. The main goals for the project were:

Streamline the APB claims handling process.
Automate the recognition and validation of the documents requested from the public
Redesign the process diagram through which a complaint has to go through within the public agency
Quickly and accurately classify the subject matter and sub-subject matter of the complaints filed.

The great challenge of this project was to create a model that could read the APB letters, recognize on which page the necessary information was found, identify which Isapre the letter corresponds to, read the ID card, verify that the letter and the ID card refer to the same person, and that the person has a valid ID card.

The strategy

Neural networks were used to analyze the text patterns in each page of the letter presented in order to identify which Isapre it corresponds to, as well as to corroborate that the letter corresponds to the Base Price Increase.

NLP techniques were used to take the claims, clean them and make them ready to be processed by machine learning models in order to achieve an optimal classification.

Subsequently, the claim entered by the user is received by the insurance company in its corresponding portal where he/she can appeal the claim, all within the established legal deadlines.

In addition, a complete survey was made of the processes for the entry of claims and they were redesigned to improve the efficiency of face-to-face and digital attention.

Neural network analysis

to identify texts and automate text ingestion

Natural Language Processing (NLP) Techniques

to process claim information and subsequent classification

Survey and redesign of processes

ensuring ad-hoc operation transformation

The achievements

The result was a reduction in processing times for Base Price Increase claims, as well as the freeing up of the different analysts within the public health agency to dedicate their time to other cases and tasks of greater added value. Additionally, this was accompanied by a process of optimization and automation of the associated processes.

A classifier was also generated that is capable of taking any claim and classifying them in one of the classes designated by the Superintendence of Health with an accuracy of over 85%.