US researchers are using artificial intelligence (AI) and machine learning for early detection of dementia.

Research scientists from Regenstrief Institute, Indiana University-Purdue University Indianapolis (IUPUI), Indiana University and the University of Miami are conducting a digital detection of dementia study.

This real-world evaluation is testing the use of an AI tool they developed for early identification of Alzheimer’s disease and related dementias in primary care, the setting where most adults receive healthcare. Individuals identified as cognitively impaired will be referred for diagnostic services.

The AI tool, called a passive digital marker, is a machine-learning algorithm the researchers developed, trained and tested. The tool uses natural-language processing to cull unstructured information combined with structured data from a patient’s electronic health record. This could include mention of memory issues, a notation of vascular concerns, comorbid conditions or other factors potentially linked to dementia.

“Between fifty and eighty per cent of dementia cases are unrecognised by the healthcare system in the USA,” said Malaz Boustani, Regenstrief Institute and Indiana University School of Medicine faculty member and senior author of the study in the peer-reviewed journal Trials. “And, if you include patients living with mild cognitive impairment, that number might actually climb to higher than eighty per cent of cases that are not recognised. In this new study, we are evaluating the practical use of our tool when used alone and when used with an accompanying patient-reported outcomes survey.”

Unfortunately, Boustani said the lay public believed there was nothing that could be done if a family member had Alzheimer’s disease.

“But that is not true,” said Boustani. “Over the past twenty years we have developed, validated and have been operating a comprehensive collaborative care model for dementia that reduces the disease burden for the patient, reduces caregiver stress and reduces inappropriate hospitalisations, keeping people living at home longer and lowering overall costs to them and to the healthcare system.”

Few primary care practices are designed for the timely detection of Alzheimer’s disease. The limited time primary care clinicians have to spend with patients, the need to focus on the health problems that brought the patient to the clinic, as well as the stigma of dementia are the major reasons for lack of recognition of the condition, according to Boustani. In addition, he said there was no demand from the public for dementia diagnoses, most likely driven by the stigma of dementia, lack of public knowledge about the benefits of early recognition of Alzheimer’s, and issues related to health insurance coverage.

The first aspect of the study is a clinical trial, already underway in Indianapolis enrolling patients seen in primary care clinics at federally qualified health centres affiliated with Eskenazi Health. The participants are expected to be predominantly people who are black and reside in urban areas. The second clinical trial starts early this year at the University of Miami primary care clinics. The participants are expected to be predominantly Hispanic and include a high percentage of rural residents.

Each trial is the same and has three arms: usual primary care approach; the specially designed passive digital marker relying on artificial intelligence; and the novel passive digital marker plus a specially designed patient reported outcomes survey.

Patients in all three arms will be followed for two consecutive years to determine how many new cases of Alzheimer’s disease were recognised and documented in patients’ electronic health records.

The 7200 anticipated study participants (3600 in Indiana and 3600 in Florida) must be 65 years old or older, have had at least one visit to a primary care practice within the past year and be able to communicate in English or Spanish. Electronic health record data from at least the past three years must be available to be enrolled in the study. Individuals residing in a nursing home or with serious mental illness are not eligible.

“Passive digital markers are developed by using data from electronic health records,” said Zina Ben Miled, a Regenstrief Institute affiliated scientist and an IUPUI associate professor, who developed the passive digital marker tool along with Boustani. “The benefit of this approach is that the information has already been collected and requires no extra effort from the patient or the provider. Through machine-learning algorithms and natural language processing, we can use these data to identify people who may be at risk for Alzheimer’s disease without the need for invasive and costly tests.”

Cognitive neurologist James Galvin, a professor of neurology at the University of Miami, developed the patient reported outcomes tool.

“Through this project, we hope to prove that these approaches can be successfully implemented in the real world, leading to benefits for both patients and their families,” said Galvin. “Combining a patient-reported outcome with a passive digital marker is an innovative and highly sensitive way to detect mild cognitive impairment, Alzheimer’s disease and related disorders, yet low burden to patients and clinicians for ease of use.”

Both the passive digital marker tool and the patient-reported outcomes tool are low cost. The study aims to confirm their suitability for use in primary care practices as well as their scalability.

It is estimated that the prevalence of dementia in primary care in the USA is about six per cent of patients but only two per cent of patients are recognised by the healthcare system as having the condition. The researchers believe using their passive digital marker, possibly with the patient-reported outcomes tool, will potentially double the number of cases identified, improving the lives of more individuals living with dementia and their caregivers.

“We are not just recognising patients with cognitive impairment and saying see you later,” said Boustani. “We are also providing computerised decision support to the primary care clinician that will help them refer these patients for confirmatory dementia screening. If the screening is positive, they can receive care based on our successful, evidence-based collaborative dementia care model.”

The study is supported by a five-year grant from the National Institutes of Health’s National Institute on Aging. Authors, in addition to Boustani, Ben Miled and Galvin, are Randall Grout, Paul Richard Dexter, and Nicole Fowler, all of Regenstrief Institute and IU School of Medicine; Arthur Owora, IU Bloomington School of Public Health; and joint first authors Michael Kleiman, University of Miami, and Abbi Plewes, IU School of Medicine.