When trained on routine health data and observation notes gathered by homecare aides, AI can be used to anticipate medical emergencies in the elderly one to two weeks ahead of an incident. The advance insights can both guide preventive care and save on unnecessary hospital and transportation costs.
Researchers proved out the concept in a pilot study published by PLOS One this week.
The team, representing multiple agencies and institutions in France, conducted a prospective study of 301 homebound seniors aged 75 or older who were regularly visited by homecare workers.
Following each visit, the aides entered their respective patients’ functional indicators across 27 metrics into a smartphone app. The metrics included things like ability to stand, eat and move about, as well as signs of mood changes, loneliness and other factors that might combine to increase the overall risk of an emergency.
The researchers aggregated this data, which totaled nearly 10,000 readings and observations, and analyzed it alongside records of ER treatments provided for the 301 participants.
From this data store, the team used machine learning to create an algorithm that could predict ER visits within seven days and 14 days over a 10-month period.
Checking the algorithm’s performance against the ER records, they found the tool predicted ER visits with 70% accuracy for the 7-day window and with 67% accuracy at 14 days.
These results were encouraging enough that the researchers concluded their exploratory work could be used to develop a survey system for real-time identification of risks.
Noting a high level of willingness to participate by the homecare aides recruited for the study, the authors chalk up this success to the simplicity of the smartphone-based process.
“For those elderly persons using in-home care, the involvement and empowerment of homecare aides in health monitoring is a promising way to improve the management of chronic disease and medical events without increasing health expenditures cost,” they write. “To our knowledge, researchers have not reported such an approach for homecare, and our study shows that homecare aide involvement is feasible even over a long period of time.”