Using A Deep Neural Network Model to Predict Elderly Care

A team of researchers has developed an AI model that offers a cost-effective approach to elderly care.

Spendings on healthcare services for the elderly has been increasing since 1965.

Back in 2010, the average medical expenditures for Americans between the age of 65 and older were 2.6 times the national average. Similarly, people within this age group accounted for over 66 percent of U.S. medical spendings in the same year.

So, it’s not surprising that healthcare providers would be considering methods to cut costs without compromising elderly care. One way to do this involves an effective and fair allocation of healthcare funds.

That means healthcare providers must be able to predict how their services would be used accurately. That way, they would save a large sum of money by not allocating funds unnecessarily.

That’s where artificial intelligence models come in.

Assistant Professor at Aalto University and FCAI, Pekka Marttinen noted:

“Without a risk adjustment model, healthcare providers whose patients are ill more often than average people would be treated unfairly.”

Deep learning models can analyze previous behavior to predict the future. And researchers in Finland have developed one that does just that.

The new risk adjustment model can predict how often older adults seek treatment in a healthcare center and how the situation changes over time. Here’s how it works.

Using a Risk-Adjustment Model to Predict Elderly Care

Countries like the Netherlands, Germany, and the U.S. are already using a form of a risk-adjustment model. However, this is the first proof-of-concept on how deep neural network could improve the accuracy of such models.

To train their model, the researchers used data from the Register of Primary Health Care Visits of THL. The data consists of out-patient visit information for every Finnish citizen between the age of 65 or above.

It was the first time researchers were using the database to train deep machine learning models, and this led to remarkable results.

As it turned out, you don’t need an enormous dataset to get a reliable result from a deep model training. Instead, a relatively small dataset could provide a more accurate prediction.

Since acquiring a large amount of medical data is challenging, the result was a welcome development for the team.

Marttinen points out:

“Our goal is not to put the model developed in this research into practice as such but to integrate features of deep learning models to existing models, combining the best sides of the bot.”

The researcher further explained that health professionals could one day use the model to support decision making and allocate funds more reasonably.