Four hidden roads discovered by AI lead to Alzheimer’s disease

UCLA researchers have mapped four different pathways that lead to Alzheimer’s disease, revealing how devastating conditions develop through sequential chains of health problems rather than isolated risk factors.

The study, published in ebiomedicine, analyzes electronic health records from nearly 25,000 patients to identify predictable patterns that can alter early detection and prevention strategies.

Unlike previous studies that examined individual conditions such as depression or diabetes, the analysis tracked how diseases followed each other over time, thus gradually developing progress in Alzheimer’s. The results of the study show that the risk of some diagnostic sequences is significantly higher than that of individual conditions.

Disease detective work

The team examined longitudinal health data from the University of California Health Data Warehouse after patients underwent medical trips backward from the diagnosis of Alzheimer’s. Using machine learning algorithms and network analysis, they identified recurring patterns in 5,762 patients who contributed 6,794 unique disease trajectories.

“We found that multi-step trajectory can indicate greater risk factors for Alzheimer’s disease,” explains Mingzhou Fu, first author of UCLA’s medical informatics. “Understanding these pathways can fundamentally change the way we take early detection and prevention.”

The analysis reveals four main ways of progress:

• Mental Health Pathways: Depression and anxiety lead to cognitive decline, mainly affecting women and Hispanic patients
• Correctional disease pathways: Brain dysfunction conditions show the fastest development of Alzheimer’s disease and death
• Mild cognitive impairment pathways: Traditional gradual cognitive decline overlaps with other approaches
• Vascular disease pathways: Cardiovascular disease leads to dementia risk, longest history

Beyond coincidence

Perhaps most striking is the discovery that approximately 26% of diagnostic progress exhibited consistent orientation sorting. For example, hypertension often precedes the onset of depression and then increases the risk of Alzheimer’s, which shows potential causality rather than just accidental.

“Recognizing these sequential patterns, rather than focusing on the diagnosis of isolation, can help clinicians improve the diagnosis of Alzheimer’s disease,” notes Timothy Chang, Ph.D., assistant professor of neurology at UCLA Health.

Researchers applied complex causal reasoning algorithms to distinguish between true causal relationships and mere statistical associations. The encephalopathy cluster showed the highest proportion of causality at 42.9%, indicating that disease progression was clearer compared to other pathways.

Verification across the United States

Trajectory patterns remain strong when tested in all of our research programs (a diverse population of country representatives). Nearly 90% of patients in this independent population can be assigned to one of the four previously identified pathways, confirming that these findings go beyond the academic medical centers in California.

This validation is critical to demonstrate that multi-step trajectory predicts the risk of Alzheimer’s more accurately than a single diagnosis. Of the nine tested trajectory patterns, seven were significantly associated with the development of Alzheimer’s disease in the national cohort.

The meaning exceeds academic interests. Healthcare providers can potentially use these patterns to enhance risk stratification, thereby identifying patients at high risk of disease progression. More importantly, recognizing the harmful sequences may cause target interventions to disrupt dangerous progress before advancing to Alzheimer’s disease.

For the 6.7 million Americans currently with Alzheimer’s (the number is expected to nearly double by 2050), understanding these pathways can provide tailor-made strategies for prevention strategies. When the patient first shows signs of following a high-risk trajectory, the clinician no longer waits for memory loss to appear.

This study represents the end result of a transition from viewing Alzheimer’s disease as a disease, as a variety of different biological pathways developed over the years or decades.

Related