JAMA sub-issue: AI AI accurately predicts who will develop dementia within two years

Written by | Edited by Mia | Typography by Wang Duoyu | Shui Chengwen Dementia does not actually refer to a specific disease, but is a medical term used to describe a group of symptoms with "decreased memory, thinking, and social abilities
.
"
According to WHO statistics, there are more than 55 million dementia patients worldwide, with nearly 10 million new cases every year
.

Dementia ranks seventh in the top ten global causes of death and is one of the leading causes of disability and dependence among older adults worldwide
.

Memory loss is the most common early symptom of dementia, and earlier identification of those at risk of developing dementia could help prioritize interventions to prevent the disease from progressing
.

However, identifying patients at high risk of dementia remains a challenge for clinicians.
Could an artificial intelligence be developed to accurately predict the probability of future dementia in memory clinic patients? Recently, researchers from the University of Exeter School of Medicine in the United Kingdom published a research paper titled: Performance of Machine Learning Algorithms for Predicting Progression to Dementia in Memory Clinic Patients in JAMA Network Open, a sub-journal of JAMA
.

The research developed a machine-learning clinical decision aid that can uncover hidden patterns in examination data, analyze people at high risk of developing dementia in the next 2 years in memory clinics, and help reduce clinical misdiagnosis rates of dementia
.

 For dementia, there are two clinical decision aids available to assist in assessing the medium- and long-term incidence of dementia in different populations
.

Among them, the Cardiovascular Risk Factors, Aging, and Incidence of Dementia (CAIDE) risk score is designed to predict the risk of dementia in middle-aged people after 20 years, and the Brief Dementia Screening Index (Brief Dementia Screening Indicator, BDSI) was designed to predict the risk of dementia 6 years later in elderly patients
.

The study included 15,307 patients who visited memory clinics of more than 30 National Alzheimer's Coordinating Centers (NACC) in the United States between 2005 and 2015
.

These patients do not yet have dementia, although they have some problems with memory and brain function
.

Using machine learning for dementia diagnosis and risk prediction, researchers achieved the first prediction of dementia incidence in memory clinics over a short clinically relevant period (2 years)
.

 The data showed that 1 in 10 memory clinic patients (1568) were diagnosed with dementia within the next two years during the study time frame from 2005 to 2015
.

Machine learning can work efficiently, using patient information routinely provided in the clinic, such as memory and brain function, performance on cognitive tests and specific life>
.

Compared to BDSI and CAIDE, the machine learning algorithm was able to predict these newly diagnosed dementia cases with a 92% accuracy rate
.

The researchers also found for the first time that about 8 percent (130 people) of dementia diagnoses appeared to be wrong because their diagnoses were later overturned, yet machine learning models were able to identify these inconsistent diagnoses to a large extent
.

"This is the first analysis of potential misdiagnoses in the NACC Uniform Data Set (UDS), and machine learning as a clinical decision aid has the potential to reduce false positives by up to 84%," the team said
.

 Study co-author Dr Janice Ranson, a researcher at the University of Exeter, added: "Dementia is a very frightening disease
.

Embedding machine learning in memory clinics could help ensure diagnoses are more accurate and less likely to be misdiagnosed.
caused unnecessary distress
.

" The team now plans to conduct follow-up studies to assess the practical use of the machine learning approach in the clinic to assess whether it can be generalized to improve dementia diagnosis, treatment and care
.

Paper link: https://jamanetwork.
com/journals/jamanetworkopen/fullarticle/2787228 Open for reprinting, welcome to forward to Moments and WeChat groups