top of page
Search

Using AI to improve personalised care: a case study

Updated: Aug 18, 2023


We recently collaborated with a Belgian health insurance company to assist them with using AI to unlock data-driven insights. The company’s aim was to identify the most vulnerable members in order to offer them personalised care. Through the creation of a health vulnerability dashboard - with each individual assigned a ‘vulnerability score’ - this innovative project leveraged AI-powered predictive modelling to create a new standard of healthcare. The goal was to apply vulnerability scoring for the entire region of Flanders, the Dutch-speaking northern portion of Belgium.


Vulnerable population groups in Belgium


Understanding Vulnerability

Vulnerability is the degree to which individuals are susceptible or unable to cope with adverse life events such as health conditions, disease, lost jobs or an accident. An individual’s vulnerability is a function of sensitivity (the degree to which one’s health depends on surrounding conditions), exposure to hazards due to an occupation, personal habit or environmental condition and adaptation capacity (e.g. social support). The vulnerability of a population to a health risk depends on a variety of factors including the level of material resources, local environment, effectiveness of government policies, access to quality healthcare, etc. Therefore, individual, community and geographical factors all play a role in determining vulnerability.

The Power of AI in Assessing Vulnerability

The insurance company had already developed a vulnerability dashboard, built on 19 criteria, which acted as a foundation for the project. Their criteria included age, citizenship and increased compensation for people with lower income, among other criteria. We brought our expertise to the table by combining these factors with open-source, complex datasets such as average household size, average income and health seeking behaviour.

By employing AI and Bayesian networks, our model could identify hidden relationships among variables, including cause-effect patterns and associations for disease progression. This allowed for a more accurate and high-resolution vulnerability assessment, in their respective neighbourhoods.


By employing AI and Bayesian networks, our model could identify hidden relationships among variables, including cause-effect patterns and associations of vulnerability. This allowed for a more accurate and high-resolution vulnerability assessment, even at the neighbourhood level.


Vulnerability Scales: A Comprehensive Picture

AI's ability to analyse diverse datasets played a crucial role in computing vulnerability levels for both individuals and their local environment. Some of the data sources used to make these predictions included contextual variables like socio-demographics, the prevalence of diabetes, chronic disease, average visits to the dentist and women going to the gynaecologist.


Every individual was assigned a vulnerability score based on various factors affecting him or her. In addition, contextual variables linked to that individual’s specific environment were included. This combined score created a comprehensive health vulnerability scale. Individual vulnerability factors were defined as high, medium or low risk. High risk factors included unemployment, older age (over 70), single parent families and not having a phone number.



Mapping Vulnerability: Empowering Timely Responses

The vulnerability scores generated through AI and data analysis were visualised on a dashboard, providing valuable insights into geographical areas requiring additional support. This visual representation helped the insurance company gain a deeper understanding of vulnerable populations, enabling them to respond proactively with targeted interventions and resources. The model’s capacity to learn from new data meant that it stayed up-to-date and precise.


Shaping the Future of Healthcare

The outcomes of the project have been very positive. Our AI-driven approach helped create a vulnerability scale at the individual and contextual level, hence revealing which individuals were most vulnerable. Feedback from the insurance company has been overwhelmingly positive, indicating their readiness to embrace AI and predictive modelling for future challenges.


The project's success has paved the way for a future where AI and data-driven insights will play an increasingly significant role in revolutionising healthcare. As more industries embrace these technologies, the potential for positive impact on people's lives and society as a whole becomes boundless.



Comments


bottom of page