Health Analytics


Report Abstract

Demographic trends, availability of health personnel and new treatment methods challenge the sustainability of healthcare. How can future health service be designed for more preventive, patient-centered, and cost-effective care, and reduce burden on health system? How can we utilize artificial intelligence (AI) for providing equal accessibility and quality of health services to all citizens?

Health analytics is a process of deriving insights from health data to make informed healthcare decisions. While such aspects of health analytics as the use of statistical models, data mining, and clinical decision support have existed for decades, only recent availability of enormous volume of data from various sources and increased processing power have made it readily available to support integrated decision making. Machine learning, data and process mining and natural language processing are the main topics in the report.

Data integration technologies provide completely new opportunities for advanced analytics. Health analytics has evolved from being descriptive to being predictive and prescriptive.

  • Descriptive analytics looks at what has already happened.
  • Predictive analytics tries to say something about what is going to happen. This involves projecting the trends and patterns in historical data and real-time data to predict, for example, the future development of health situation of a patient or patient groups in the population.
  • Prescriptive analytics is a step further. By using medical knowledge, it can evaluate alternative treatments and find the optimal one in a given situation. Prescriptive analytics is key in personalized medicine.

HIMSS[1] has developed a systematic model, the Adoption Model for Analytics Maturity[2] (AMAM), used to evaluate maturity of healthcare organizations within the analytics field. On the second most advanced stage of the AMAM model, there are predictive analytics and risk stratification. Prescriptive analytics and personalized medicine are on the most advanced stage; to get to this level, genomic data must be available.

Good analyses require access to relevant high-quality data. A significant challenge for health data is its heterogeneity and complexity. The healthcare processes generate a large amount of unstructured data, such as medical images and free text. Therefore, in order to exploit the full potential of health data, the machine-learning analytics tools can be utilized. Moreover, for preserving the meaning of data, it is necessary to harmonize it with both a common data format and terminology.

The governmental white paper “One Citizen - One Record” [1] promotes that “data should be available for quality improvement, health monitoring, management and research”. The Norwegian e-health strategy (2017-2022) [2] has “better use of health data” as one of the strategic focus areas. The Norwegian Directorate of e-health is developing a national health analytics platform, which will “... simplify access to health data and facilitate advanced analytics across health registries, source data, health records and other sources of health information.” It is further stated: “The Directorate of e-health aims to contribute to testing and development of new technologies in health analytics, such as artificial intelligence, machine learning, algorithms and predictive analytics.” In addition, the Norwegian Technological Council[3] has its own project focusing on AI and the welfare state [3].

[1] https://www.himss.org

[2] http://www.himssanalytics.org/amam

[3] The Norwegian Technological Council is an independent public body that provides advice to the Norwegian Parliament and the Norwegian Government on new technologies