Risk Assessment for Personalized Health Insurance Based on Real-World Data
Abstract: The way one leads their life is considered an important factor in health. In this paper we propose a system to provide risk assessment based on behavior for the health insurance sector. To do so we built a platform to collect real-world data that enumerate different aspects of behavior, and a simulator to augment actual data with synthetic. Using the data, we built classifiers to predict variations in important quantities for the lifestyle of a person. We offer a risk assessment service to the health insurance professionals by manipulating the classifier predictions in the long-term. We also address virtual coaching by using explainable Artificial Intelligence (AI) techniques on the classifier itself to gain insights on the advice to be offered to insurance customers.
Health insurance products are today static in terms of customers’ health evaluation. Any personalization is based on a risk assessment of static data from the medical record of the customer and questionnaires they answer at the contract setup phase. Personalized health insurance products need to be dynamic, employing a continuous risk assessment of the customer.
Medical history of insurance customers can be scarce, and anyway only partly deter- mines health. There are several studies that provide evidence about the relation between lifestyle and health. A study on diabetes prevention (Grey 2017) suggests that lifestyle is important for the outcomes in youths and adults. It correlates with the fact that obesity in adults has risen from less than 5% to more than 40% in some states, with an increase seen in type II diabetes over the last 20 to 30 years. Another study (Joseph-Shehu et al. 2019) shows that a good health-promoting lifestyle, especially health responsibility, physical activity and stress management, is a determinant of overweight and obesity, a major risk factor for cardiovascular diseases, type II diabetes and some forms of cancer. Behavioral, environmental, occupational and metabolic risk factors have been analyzed, leading to the 2017 global burden of disease study (Stanaway et al. 2018).
Risk assessment is an integral part of the insurance industry (Blackmore 2016b), but it is usually static, done at the beginning of a contract with a client. While the continuous estimation of risk factors is well-known in medicine, it is not widely used to personalize insurance products. Such personalized products start appearing as digital risk assessment platforms based on data start transforming insurance (Blackmore 2016a), and have been explored in the car insurance sector. The importance of vehicle-based risk assessment is discussed in Ref. (Gage et al. 2015). Usage-based insurance utilizes driver behavior analysis based on big data, as discussed in Ref. (Arumugam and Bhargavi 2019). Similarly, pricing innovations in German car insurance are addressed via telematics driving profile classification in Ref. (Weidner et al. 2017).
Unlike medical history where a snapshot in time yields information, lifestyle and behavior cannot be assessed momentarily, since they involve people’s habits and their continuous change. As such, personalized health insurance products require the continuous monitoring of customers’ lifestyle and behavior. This can be achieved with software tools for the collection of data chosen so that they capture important aspects of lifestyle and behavior. In this work we rigorously define with insurance experts the data to be collected, and we employ the Healthentia system (Innovation Sprint 2020) for data collection. Given the collected behavioral data of their customers, risk assessment services can be provided to health insurance professionals by training machine learning (ML) predictors for important health parameters. The usage of ML in insurance is not new. ML has been used to analyze insurance claim data (Bermúdez et al. 2020, Burri et al. 2019). The work in Ref. (Qazvini 2019) explores how the vehicle insurance coverage affects driving behavior and hence insurance claims. Instead of an analysis of data at the end of the insurance pathway, after the event, this paper focuses on the continuous analysis of data at the source (the customer) to modify the insurance pathway by personalizing the insurance product.
Insurance companies benefit from personalized dynamic product offerings, as they can be competitive with lower prices for low-risk customers. However, to obtain their customers’ consent to monitor them, insurance companies also need to persuade their customers about the benefits for them. Customers will potentially consent to two types of rewards: On the one hand, monetary rewards stem from receiving personalized offers with reduced premiums due to the lower risk of their healthy behavior. On the other hand, coaching for well-being is an indirect reward that can be offered by employing explainable AI techniques in the classifiers utilized by the risk assessment service.
The structure of this paper is as follows: Section 2 addresses data collection, iden- tifying what and how to measure. In Section 3, classifiers on well-being are used both to assess risk and to establish personalized advice for well-being coaching. This work leads to the introduction of personalized health insurance products by the relative pilot of the INFINITECH project (Infinitech H2020 2020), as discussed in Section 4. Finally, the conclusions are drawn in Section 5.
By Aristodemos Pnevmatikakis, Stathis Kanavos, George Matikas, Konstantina Kostopoulou, Alfredo Cesario and Sofoklis Kyriazakos from the source https://www.mdpi.com/2227-9091/9/3/46