Health insurance reads patient records with AI

Team Konfuzio

Administrative management and future potential:

Manual health checks for statutory and private health insurance companies were yesterday - today the keyword is artificial intelligence!

Health insurance status quo  

Wrong decisions in the health examination when changing or taking out health insurance can have far-reaching consequences. Based on the medical history, the risk of an illness is estimated and the corresponding premium amount is calculated, or private health insurers may refuse to accept an application [1]. Applications may be falsely rejected or accepted. Can automated data processing by AI-supported systems simplify decision-making bases, reduce sources of error and relieve all parties involved?

Information from documents is usually processed manually. The increasing competition in the insurance industry is causing a wave of digitalization. Rule-based procedures such as OCR and RPA robots offer only small digitization advantages in automated data processing [2]. Increasingly, learning systems using artificial intelligence are helping people to make decisions and to manually process and extract data from documents such as preliminary risk enquiries for occupational disability insurance [5].

ADDED VALUE AND SOLUTION THROUGH KI

AI cannot and should not completely replace human cognitive processes. But AI can support and relieve human beings. Artificially intelligent systems make it easier for the person in charge to examine critical cases. Thus, AI enables exact evaluation and individual health checks.

There are already solution models for this: Intelligent algorithms that combine deep learning-based image processing and natural language processing [6]. They know how to not only structure unstructured information, but also to understand and logically link the contents of texts. With the help of trained data sets, the cognitive systems learn and the AI continuously improves. 

The problem can be illustrated using the example of the occupational disability risk pre-enquiry: Automatically processed data can make it easier to identify people with pre-existing conditions in prior risk inquiries. In health insurance, AI enables the pre-filtering of illnesses that could lead to an exclusion. If the BU applicant has an "asthma spray", it is assumed that he belongs to the risk group "asthma". However, if this patient only has the spray for emergencies and does not use it regularly, an exclusion from the BU insurance would still occur, even though the patient does not clearly belong to a risk group. In general, in Germany up to 25% of all prior risk inquiries with pre-existing conditions are either rejected or only conditionally accepted due to exclusions and surcharges [3]. Approximately 75% of all BU inquiries with pre-existing conditions are accepted without aggravation. AI can recognize this information from the BI request and structure the relevant information for further examination by humans.

Illustration of the acceptance rate for BU requests
Illustration of the acceptance rate for BU requests

But it's not just occupational health risk inquiries that can be processed by health insurers' AI. Personal data, medical histories, hospital and medical bills can be quickly analyzed and prepared for further human processing by AI and without human labor. Post-processing times and errors are reduced, saving everyone involved a lot of time, money and effort. Components of BU competence define to 30% from workflow management, standardization of processes, cover letters and questionnaires [4]. 

Pie chart assumption in the event of prior illness in the case of occupational disability insurance
Occupational disability Acceptance rate for pre-existing conditions

CONCLUSION 

Artificial intelligence offers health insurers promising opportunities in health assessments. Thanks to artificially intelligent and automated processes, AI helps to identify individual risks and sort the acquired pre-existing conditions from the occupational disability inquiries into the non-critical 75% and the critical 25%. This subsequently facilitates the manual processing of the critical 25% by humans. Ultimately, decisions can be made more easily on the basis of reliable information, processes can be accelerated and process deviations in risk assessment can be reduced.

SOURCES

[1] Knowledge-PKV.de (June 2020): What health questions do you have to answer when applying for private health insurance?

[2] McKinsey & Company (June 2017): Artificial intelligence in health insurance: smart auditing with self-learning software. S.2.

[3] MORGEN & MORGEN Press Release (May 2018): M&M Rating Occupational Disability: Psyche issue continues to gain importance. S.4.

[4] Morgen & Morgen BU Rating (May 2020): Excellent performance of NLV: Market Management Life Private. S.7.

[5] Capgemini (June 2019): Artificial intelligence in healthcare: The health insurance of the future

[6] AOK (December 2018): Scientific Advisory Board for Digital Transformation: comments on an AI strategy for a statutory health insurance company. S.5

LIST OF ILLUSTRATIONS

Fig.1: Pexels - Typewriter, after: Andrea Piacquadio, as of 02/2020

Fig.2: Occupational disability - acceptance in the case of pre-existing conditions, according to: Morgen & Morgen, as of 05/2018

More Articles

Skyline of the banks in Frankurt

AI in banking: AI ideas for value-driven use

How can banks save a lot of time in the back office with AI? KYC data is extensive and complex. Even for well-trained employees...

Read article
Data Engineer at work

AI Quality via DIN SPEC 92001-2

DIN 92001 defines a quality meta-model over the lifecycle of AI and identifies performance, stability, functionality and traceability of AI....

Read article

IT-Tage 2020

From 7 to 10 December 2020, the IT Days will be held as a remote conference for the first time. The conference is aimed at...

Read article

    Are you looking for more information?

    You are also welcome to send us an e-mail to [email protected] , call us via +49 6441 8994005 or book a meeting.
    Arrow-up