When Machine Learning Helps Reveal Hidden Dementia

 You'll likely need this because of your chances of dementia post stroke. Which then means your  COMPETENT? DOCTOR NEEDS EXACT DEMENTIA PREVENTION PROTOCOLS! 

DOES YOUR INCOMPETENT? DOCTOR NOT HAVE THESE?

DOES YOUR DOCTOR HAVE EXACT DEMENTIA PREVENTION PROTOCOLS? NO? So, your doctor is incompetent? 

1. A documented 33% dementia chance post-stroke from an Australian study?   May 2012.

2. Then this study came out and seems to have a range from 17-66%. December 2013.`    

3. A 20% chance in this research.   July 2013.

4. Dementia Risk Doubled in Patients Following Stroke September 2018 

The latest here: 

When Machine Learning Helps Reveal Hidden Dementia

Patients at primary care clinics received more dementia diagnoses after implementation of a machine-learning tool designed to flag potential cases, according to a study published in JAMA Network.

The findings point to a potential solution for diagnosing dementia earlier in the disease process, said Malaz A. Boustani, MD, MPh, professor in the Department of Medicine at the Indiana University School of Medicine in Indianapolis, who led the study.

“Changing or transforming primary care to accommodate the needs of people living with unrecognized dementia or mild cognitive impairment requires a scalable and sustainable solution with minimum time and minimum cost,” Boustani said.

An estimated 6 million Americans have dementia, and more than one third of those older than 55 will eventually develop the condition, according to the National Institutes of Health.

Boustani and his colleagues conducted a randomized clinical trial of more than 5300 adults aged more than 65 (mean age, 71 years; 62.2% women) at nine federally qualified health centers in Indianapolis over a 2-year period beginning in July 2022. Patients did not have a previous diagnosis of mild cognitive impairment, dementia, or severe mental illness.

Clinics were randomized to one of three approaches. One group of 1724 patients received usual care with no routine screening for Alzheimer’s disease and related dementias (ADRD). Another 1300 patients were seen at clinics that used a machine learning algorithm that scanned electronic record data for indicators of dementia risk. A third group of 2301 patients received care at clinics that made diagnosis based on the algorithm plus a patient-reported survey of 10 questions on cognitive abilities, daily tasks, behavior, and mood.

Among clinics using the AI tool, clinicians would receive a notification if a particular patient showed signs of risk for dementia, and suggest ordering a memory test, referral to a specialist, and talking to the patient about any concerns.

Boustani said he and his colleagues validated the diagnosis by analyzing more than 2000 cases of ADRD and more than 11,000 people without those conditions in Indiana. They then divided the population into two cohorts, one of which was used to train the program and the second was used for validation, producing accuracy of close to 80%, Boustani said.

Clinics randomized to the algorithm and survey approach showed 31% higher odds of new ADRD diagnoses (adjusted OR [aOR], 1.31; 95% CI, 1.05-1.64) than usual care clinics (12-month incidence, 12.4%). Clinics using the algorithm alone had a lower incidence of diagnosis than usual care (12-month incidence, 10.3%; aOR, 0.84; 95% CI, 0.63-1.11).

After 12 months, 36.7% of patients in clinics using the survey and algorithm approach had undergone a dementia-related diagnostic test compared to 27.8% of patients in clinics using algorithms alone and 29% of those in usual care clinics.

Chelsea Cox, MPH, MSW, social worker and doctoral candidate at the University of Michigan School of Public Health in Ann Arbor, Michigan, said primary care clinicians are often overburdened and do not have time to screen for ADRD.

“The tools that were employed in this research study remove some of those barriers to help facilitate not only from the patient side, being able to report concerns about memory and cognition, but from the clinician and primary care provider side, to be able to very efficiently determine whether somebody’s at risk and should be referred for additional cognitive and neuropsychological testing,” Cox said.

Various study authors reported receiving grants, consulting fees, and personal fees, along with having equity interest and holding patents from the National Institutes of Health, the Academy for Continued Healthcare Learning, Cognivue, and Pfizer, among others. No other disclosures were reported.