Tracking changes in health with machine learning

17 August 2021• NEWSITEM

Studies conducted at the Leiden University Medical Center (LUMC) – Campus Den Haag indicate the Dutch often perceive their health state as constant, despite possible improvements or deteriorations. Brian Doornenbal of LUMC and Renz Bakx of Salut have used machine learning to identify key patterns associated to certain changes in health.

Researchers suggest that self-rated health (SRH) can provide important data to predict health events in an individual’s life. In a population health study led by Doornenbal, Postdoctoral Researcher at the department of Public Health and Primary Care at the LUMC, Dutch persons were found to generally perceive their health state as perpetual. Out of 2154 participants assessed during an 11-year period, over 85% felt that their good health was steadily maintained. The remaining group had a poor perception of health for an extended period of time, which was generally accompanied by medication use, a higher BMI and unemployment rate - factors considered difficult to rapidly change.

Prolonging life in good health

The Ministry of Health, Welfare and Sport (VWS) states that by 2040 the Dutch population should expect to have life prolonged by five years in good health. Health disparities between the lowest and highest socioeconomic groups within the Netherlands should also be reduced by 30%. This goal will not be met without challenges considering the increasing number of patients with chronic conditions, an aging population and the prevailing COVID-19 pandemic. According to Doornenbal: “In addition, since the Dutch often perceive their health state as perpetual, it may limit overall changes in population health”.

Experiencing changes in health differently

LUMC and Salut collaborate on several projects concerning SRH. For this study, Doornenbal and Bakx were amongst the first to apply a machine learning technique based on speech technology to identify and gather data on common health change patterns. Their findings have been published in Preventive Medicine Reports. Doornenbal explains: “Recognizing long term changes in health can be difficult. Not always does the change in one’s perception of health occur simultaneously to the actual change. Individuals can also experience the same change in very different ways, such as how they respond to a viral infection. Consequently, recognizing key patterns associated to certain health changes is a complex process. The machine learning technique we used during our research has helped us identify how health changes, as well as subtle - yet typical - change patterns.”

Four key patterns

From 2009 to 2018, a representative group of the Dutch population - consisting of 2154 participants -  reported how healthy they felt. Four key patterns concerning changes in health were identified. Nearly two-thirds of the participants often felt their good health was stable. Two smaller groups indicated feeling very healthy or excellent. However, more than 14% of the participants felt only moderately healthy. “Self-rated moderate health is particularly worrying since sense of health is a strong predictor of illness and other adverse health situations”, says Doornenbal. If individuals feel unhealthy for extended periods of time, they are at an increased risk of experiencing illness for longer. “Once you are sick, it is difficult to feel healthier”, he notes.

Follow-up research

Population Health (or public health) is one of the LUMC’s Societal Outreach Themes. Doornenbal and Bakx’s study offers interesting results in this regard. While examining changes in self-rated health perception over long periods of time, the researchers identified crucial differences between groups of people who felt healthy and unhealthy. In follow-up studies, Doornenbal and Bakx will analyse how little changes in lifestyles can influence one’s health in the long-run. Applying the same machine learning technique on data gathered through a health app, the researchers will also focus on health changes over a shorter period of time. 

1 / 1

Stay informed on the latest news of the Leiden University Medical Center, and sign up for our biweekly newsletter or subscribe to our LUMC Magazine.