Population Ageing: Predicting Older Persons at Risk of Declining Health

Population Ageing: Predicting Older Persons at Risk of Declining Health

 

Background

Citizens of Copenhagen, while living longer than ever, have lower life expectancy compared to the country as a whole and many live with one or more chronic diseases. Combined with the ageing population and lack of equal opportunities, there is an urgent need for coordinated health services targeting high risk groups, particularly older persons.

Currently, everyone aged 75 years old are invited to an information meeting on preventive home visits and when they reach the age of 80, everyone is offered annual preventive home visits from the municipality. For everyone else aged 65-79, the municipality offers preventive home visits for selected high risk citizens.

However, there is no established strategy yet on how to accurately identify older persons at risk of declining health. Here the advancements of register data and machine learning methods provide a unique opportunity to establish a novel, data-driven, population-based prevention strategy to identify older persons at high risk of disease, disability, and thereby needing long term care who would benefit from tailor-made interventions.

 

Objective

To develop iterative self-learning predictive algorithms to accurately detect older persons at risk of disease, disability thereby needing long term care and at increased risk of dying in the population of adults aged 65 and over living in Copenhagen.

 

Methods

First, we will develop predictive algorithms by analysing patterns of health progression over time using register data from Statistics Denmark.
Second, we will improve the predictive algorithms by adding new types of predictors using data from various additional sources e.g. screening tools and data gathered by Copenhagen Municipality.

 

State-of-the-art

The machine learning methods have been developed and will soon be implemented on an initial dataset consisting of everyone aged 65 years and over living in Denmark during the period of 2012-2016. In this initial dataset we have common predictors that are easily available in Statistics Denmark and two types of outcomes outcomes: the number of hours of personal care provided by nurse and mortality.

 

Expected outcomes and perspectives

Rigorous scientific analysis of health and registry data will improve our ability to identify citizens at risk of declining health, facilitating a shift from the traditional, demand-driven care towards a novel, data-driven prevention strategy. Its scalability and flexibility makes this innovation applicable to other populations within Copenhagen and beyond.

Stay tuned and subscribe for updates.

For further information, contact us at info@dataforgood.science.