The Estonian Health Insurance Fund's AI solution for risk-based treatment management

In 2015, the Estonian Health Insurance Fund in cooperation with the World Bank started to develop and pilot a risk-based treatment model that would help increase the integration of health services. With risk-based management, GPs can identify patients with multiple chronic diseases on their list for whom additional prevention, counseling and monitoring would be most beneficial to their health and quality of life. If these patients are neglected by GP teams, this can lead to serious problems, including unnecessary deterioration in health, which not only causes health issues but also unnecessary costs to the healthcare system (avoidable hospitalizations, duplication of examinations, etc.).

 

Kratt-based solution

From 2016 to 2017, the Estonian Health Insurance Fund and the World Bank conducted a pilot project for managing treatment in Estonia; the first step of the project was the designing of a clinical algorithm for identifying risk patients in cooperation with Estonian general practitioners. In designing the algorithm, the treatment invoice data of the Health Insurance Fund was combined with the clinical institution of general practitioners.

The purpose of the machine learning project was to improve the ‘mechanically designed’ algorithm for risk patients, developed during the pilot project, to more accurately identify patients whom general practitioners could involve in the risk patient programme.

More accurate identification of patients enables general practitioners to prevent the deterioration of health of patients with cardiovascular or respiratory disease or with mental disorders, and to improve/maintain their quality of life. In the health care system, this will increase treatment quality and the effectiveness of the health care system, preventing avoidable extraordinary hospitalisations, duplication of medical tests, etc. The solution will also contribute to the empowerment of primary level medical aid.

The result

With the created solution, it is hoped to find the best algorithm for predicting which of the patients with the selected diagnoses are likely to be hospitalised. In the solution, specific conditions/diagnoses will firstly be found in the Health Insurance Fund treatment invoice database that will be involved in the algorithm.  Different models have their strengths but differences should not be major.

Results of the solution are comparable with the results of research papers presented in references: John Hopkins Adjusted Clinical Groups (the leading proprietary risk stratification tool) – Haas et al.; Risk-Stratification Methods for Identifying Patients for Care Coordination; The American Journal of Managed Care (September 2013).

The machine learning-based model is better at finding risk patients than the old mechanical model.

With the solution, it is possible to create a practical model for general practitioner practices, to forecast which patients have a higher likelihood of hospitalisation or another health risk.

Performer

 

The Estonian Health Insurance Fund has been cooperating with the World Bank since 2015. This AI solution was created by the World Bank with the involvement of the Estonian Health Insurance Fund. In addition, Costa Rica is involved in the project in order to add value to the project by using the database of medical bills there (see clinical / socio-economic data, which are not included in the EHIF databases).

©2019 kratt@mkm.ee