Alzheimer's disease (AD) is the most common neurodegenerative disorder that starts late. Identifying individuals at risk for AD with the increasing incidence of AD is important for early intervention. In a new study, scientists at Ohio State University have created a machine learning model to classify risk factors to determine how strong their association is with the ultimate manifestation of AD. This is the first study to create machine learning models with genetic risk scores, non-genetic information, and electronic health record data. Using the models, scientists graded risk factors for two populations from UK Biobank: white family individuals aged 40 or older and a subset of adults who were 65 or older.

Age is the most significant risk factor for AD in the general population, accounting for one-third of total risk by age 85, according to the Alzheimer's Association. 

Xiaoyi Raymond Gao, lead author of the study and associate professor of ophthalmology and visual sciences and biomedical informatics at Ohio State University College of Medicine, said, "We all want to live healthy lives, and income can be an important factor in making decisions about what you can eat, where you can live, your level of education, and access to care - all of which can contribute to Alzheimer's disease."

Of the 457,936 individuals in the UK Biobank sample, 2,177 had AD, while 455,759 did not, and 88,309 were 65 or older.

The following non-genetic risk factors were found to be different between individuals with and without AD: high systolic and low diastolic blood pressure, diabetes, low household income, and education, recent falls, hearing loss, and having a mother with a history of AD were more common among adults with AD.

High blood pressure diagnosis, urinary tract infection, episodes of depression, fainting, vague chest pain, numbness, and significant weight loss were among the top 20 risk factors for the entire adult sample. High cholesterol and irregular gait were two additional risk factors in the top 20 for adults aged 65 and older. These findings highlight the impact of including condition codes from electronic medical records in models.

Gao said, "Machine learning can identify relationships between all of these traits or variables, select important ones, and assign some special ones that are heavily involved in the risk of Alzheimer's disease compared to others. Generally, being overweight is not good, but we also see here that a low body mass index is not good. High blood pressure is generally not good, but we see here that low diastolic blood pressure is not good either. The models revealed some interesting patterns."

Researchers built the model in two stages. To identify genetic changes linked to the onset of the disease and overall risk of Alzheimer's, researchers first studied the genome-wide association of Alzheimer's disease using data from the Alzheimer's Genetics Consortium. They constructed two polygenic risk scores, which combined genetic effects into a single score for each individual, using different sets of variations. These scores estimated the same risk for each individual. The team used over 11,000 condition codes from electronic health records, along with traditional risk factors such as age, gender, education, body mass index, and blood pressure, collected from participants' records in the UK BioBank. Specific individuals were selected for the study.

The team also used an algorithm in the model's output analysis to ensure that the variables associated with risk in the analysis were weighted appropriately. 

Scientists noted, "We already have a genetic risk for established diseases, but information on how other health and socioeconomic factors affect our risk for Alzheimer's - as well as glaucoma, which the cow also studied - empowers us to take protective measures." 

The cow said, "If people know more about the risk factors, they can adjust their lifestyle. There is no cure for Alzheimer's or glaucoma to stop their progression, so building models to predict these diseases would also help in developing drugs and effective and low-cost screening programs."