AI for Senior Citizens

We are now living longer, and the number of people worldwide aged 65 and over is expected to grow from 703 million in 2019 to 2.2 billion in 2080, according to the World Population Prospects Report published by the United Nations last year. The proportion of the global population that is elderly is also on the rise, almost doubling from 5.5% in 1974 to 10.3% last year, and it is projected to grow to 20.7% by 2074.

A consequence of aging is that we are more likely to have medical problems. At the same time, the healthcare system in many countries is already stretched due to a lack of workers.

“There are just not enough doctors and nurses to deal with a growing elderly population,” said Massimiliano Zecca, a professor of healthcare technology at Loughborough University in the U.K.

In the U.S, for example, a severe shortage of doctors is expected by 2034, with between 37,800 and 124,000 physicians lacking, partly fueled by the growing number of seniors, according to a recent report by the Association of American Medical Colleges (AAMC).

Artificial intelligence (AI) is being harnessed as a way to bridge the gap. Instead of replacing medical professionals, it can provide support and extend their capabilities both in hospital settings and at home.

However, researchers are still investigating how different solutions can best benefit patients.

“There are lots of challenges with deploying AI well, and with developing AI solutions that don’t misbehave,” said Muhammad Mamdani, a professor at Canada’s University of Toronto and vice president of data science and advanced analytics at the hospital network Unity Health Toronto. “I think we need a very thoughtful, careful approach.”

Dozens of AI tools that can help improve elderly care are already being deployed in hospitals. Some examples are diagnostic aids that analyze images from medical scans, such as a platform called RapidAI that can help determine if a stroke patient needs further treatment by automatically detecting obstructed blood vessels in the brain.

Mamdani and his colleagues have developed a tool called Chartwatch that monitors hospitalized patients and acts as an early warning system to flag unexpected deteriorations in health. The idea came from a clinician who said he struggled to identify when a patient would be in such trouble, which is typically assessed based on experience and by making sense of available data.

Chartwatch can predict if a patient is going to die or go to the ICU in the next 48 hours. The tool uses a machine learning algorithm trained on health data from more than 20,000 patients to learn how about 150 different variables, from vital signs to different lab test values, relate to the state of a person’s health. The system then monitors these variables in a patient every hour to predict if their condition is becoming high-risk, sending an alert to the medical team if that is the case.

“It’s almost impossible for the human mind to be able to actually go through all of that data, understand all the temporal changes, and then make sense of what will happen,” said Mamdani, “so the computer does that for the [physicians and nurses].”

Chartwatch’s predictions are not always accurate, but clinicians make mistakes, too. Mamdani and his colleagues realized, however, that there were no figures showing how good medical staff were at assessing whether a patient’s condition would deteriorate. They decided to collect predictions from more than 3,000 clinicians and organize the information they received into a study, which was recently published in the journal Critical Care Explorations.

“What we were able to show is that when a physician says a patient is going to die or go to the ICU, they are right less than one-third of the time,” said Mamdani.

Chartwatch’s performance initially was found to be comparable to that of clinicians, but its success rate rose considerably after its algorithm was improved. Mamdani said the tool can now identify over 80% of cases where a patient is at risk of dying or requiring ICU care, which is more than twice as good as physicians. However, Mamdani and his colleagues recommend taking a conservative approach and checking on a patient if either the AI or the clinician predicts they are at risk.

The tool, which is currently being used at two hospitals in Toronto, is helping to save lives. The team followed over 13,000 patients admitted to a ward for acute needs at St. Michael’s Hospital, comparing the number of in-hospital deaths during two time periods before and after Chartwatch was implemented, for a recent study published in the Canadian Medical Association Journal (CMAJ). They found that using the AI system resulted in a 26% drop in unexpected deaths.

Mamdani said many patients in such wards are elderly and suffer from several medical conditions that often increase their risk of dying.

“We’ve been getting a lot of benefit out of [Chartwatch],” he says. “We would love to see the solution go to other hospitals.”

Elderly people can also use AI tools at home to assist with their health-related needs. As of August 2024, the US Food and Drug Administration (FDA) has approved nearly 1,000 AI-based medical devices, many of which can be purchased by consumers. KardiaMobile, for example, is a tiny bar-shaped device with two sensors that can be plugged into a smartphone; it measures electrical activity in the heart to detect pulse or rhythm irregularities.

There are also tools to help senior citizens keep track of their food intake to detect changes in eating habits that could affect their health and to optimize their diet, such as a system called Meal Vision designed for use in nursing homes. Some fitness apps are specifically designed for seniors too, providing personalized exercise programs that take into account mobility levels, such as SilverSneakers GO.

Chatbots such as OpenAI’s ChatGPT are increasingly being used to learn about symptoms or medical conditions. About 17% of American adults use them at least once a month to get information about their health, according to a recent survey from KFF, a San Francisco-based non-profit health policy research organization.

“It was a much higher percentage than I had expected,” said Ateev Mehrotra, a professor at Brown University’s School of Public Health, of the results of the KFF study. “They’re [just] so new.”

Mehrotra thinks many elderly people use chatbots, or have someone use them on their behalf, simply because they are much more likely to have a medical issue compared to younger adults.

“I think people erroneously believe that older adults are less likely to use these tools,” he said. “We’ve seen that with telehealth and other technologies.”

Mehrotra thinks chatbots can provide useful health information, as long as people realize they could be wrong. Previously, search engines or ‘symptom tracker’ apps would have been used instead, and he considers chatbots to be a step up.

“Their diagnostic and triage accuracy is not superior to a physician, but it’s superior to the tools that were previously available to us,” he said.

AI embodied in robots could also support the health of senior citizens at home by helping with mobility or providing therapy, for example. However, although several designs have been proposed over the years, especially in Japan, few have taken off. Some have been too expensive or impractical to implement, while others simply weren’t appealing to users, or suffered from a short battery life. Instead of making a caregiver’s job easier, robots can also create extra work since they need to be charged, cleaned, and maintained.

Zecca thinks part of the problem is that engineers often create robots as a technical challenge, without the users in mind. He and his colleagues have taken a different approach for a system they are developing with a robot and sensors to help frail senior citizens at home.

“The solutions came up through discussions with elderly people so that they want to use these kinds of things in their houses,” said Zecca.

Managing frailty can prevent hospitalizations and limit the need for social care. Exercise has been shown to help, but it needs to be tailored to the individual: certain movements or the wrong quantity could be harmful. By monitoring elderly people’s physical abilities at home, the system becomes an extension of the healthcare system. Doctors can access data to provide personalized guidance, and instantly receive alerts if there are worrying changes.

“A doctor can see what is happening now, not six months later when the patient goes back to see [them],” said Zecca.

The system, called I’M-ACTIVE, uses machine learning algorithms to make sense of data collected by sensors placed around a person’s home. One sensor embedded in a kettle would be used to assess upper body movements, for example.

A simple and low-cost robot, which resembles a taller version of a Roomba, the popular robot vacuum cleaner, would collect information while moving around one’s home, and could be interacted with in various ways. It would transmit information between the doctor and the patient and provide an easy-to-use interface where a person can access their data.

Privacy and security concerns about this data will need to be addressed. Zecca’s colleagues recently published a review paper in Applied Sciences that specifically delves into these issues in relation to social robots for assisting older people. It looks at situations where data leaks could occur, such as unauthorized users gaining access. They propose using blockchain technology, where data is structured into blocks in a chain of coded information, to make it more secure. However, research will be required to figure out how to use the technology, which needs a lot of computing power, without impacting the performance of a robot.

Advancements in communication should help make robots more appealing to senior citizens. Zecca is particularly excited about integrating chatbots with robots to make interactions more seamless and natural. Thanks to the large language models (LLMs) that power chatbots, they can now quickly understand what a person is saying and generate a relevant and engaging reply.

“You can talk directly to the system and the system can talk back to you,” he said. “That’s obviously enabling a lot of different interactions that were not possible until very recently.”

Sandrine Ceurstemont is a freelance science writer based in London, U.K.

AI for Senior Citizens

Further Reading