From Cortex to Context: Rethinking Late-Onset Epilepsy Risk

Abstract

Incidence and Risk Factors of Epilepsy Among Older Adults in the US Medicare Population.

Dong W, Cabulong A, Vu L, Al-Kindi SG, Warner DF, Schiltz NK, Fein HL, Ghearing GR, Sajatovic M, Koroukian SMJAMA Neurol. 2026;83(2):161-170. doi:10.1001/jamaneurol.2025.4347. PMID: 41212547; PMCID: PMC12603946.

Importance: Geographic variation in epilepsy incidence among older adults may reflect contextual risk factors and point to opportunities for targeted prevention. However, privacy constraints and sparse case counts have historically limited small-area analyses. Objective: To map incident epilepsy among older adults at the smallest geography permissible by privacy constraints and identify contextual social and environmental determinants associated with high-incidence areas. Design, Setting, and Participants: This cohort study examined Medicare administrative claims from 2016 to 2019 for all counties in the contiguous United States. A random sample of 4 999 999 Medicare Fee-for-Service beneficiaries 65 years or older with non-Hispanic Black and Hispanic beneficiaries oversampled at rates of 1.50 and 1.75 times their representation in the study population. Beneficiaries with incident epilepsy were identified by claims criteria and codes from the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, in 2019 and had no epilepsy claims during the period 2016–2018. Data were analyzed from January to March 2025. Exposures: Area-level social and environmental determinants of health (SEDH) obtained from publicly available sources and linked to beneficiaries’ residences. Main outcomes and Measures: The outcome was area-level epilepsy incidence rate in 2019. To comply with data privacy requirements, the Max-P regionalization method was used to aggregate 3108 counties into 692 “MaxCounties,” each containing at least 11 incident cases. Incidence rates per 100 000 persons were mapped. Associations between SEDH variables and epilepsy incidence were estimated using random forest and multivariable logistic regression. Results: Among 4 817 147 beneficiaries, 20 263 incident epilepsy cases were identified in 2019 (mean [SD] age, 78.7 [7.5] years; 54.6% women). Incidence rates across MaxCounties varied more than 10-fold (range, 141-1476 per 100 000). In random forest models, higher incidence was associated with insufficient sleep, heat index, physical inactivity, uninsured rate, proportion of non-Hispanic Black residents, and obesity prevalence. In multivariable regression, MaxCounties in the highest tertile for insufficient sleep had nearly double the odds of high epilepsy incidence compared to the lowest tertile (odds ratio [OR], 1.99; 95% CI, 1.10-3.60). Lack of household vehicle access was similarly associated with high incidence (OR, 1.93; 95% CI, 1.16-3.25). Conclusions and Relevance: Our findings highlight the spatial heterogeneity of epilepsy burden in the US Medicare population and underscore the importance of contextual SEDH factors, such as sleep, mobility, and infrastructure, in shaping disease patterns. These insights may help guide targeted public health interventions and resource allocation.

Commentary

Neurology has long sought disease within the individual: lesions, proteinopathies, genes, a paradigm that epilepsy has reinforced through focality and network-based ictogenesis. But what if we have been looking in the wrong place? In this compelling work, Dong and colleagues present a captivating reframing of late-onset epilepsy (LOE)^1,2: not as a purely biological phenomenon, but as one deeply embedded within the social, environmental, and structural contours of where people live. ³

Drawing on nearly 5 million Medicare beneficiaries and an innovative spatial analytic approach, the authors reveal a striking >10-fold variation in LOE incidence across the United States, from 141 to 1476 per 100 000 depending on location. ³

Perhaps the most visually arresting finding of the study is that the highest-incidence regions cluster across the American South. This “epilepsy belt” echoes the contours of the Stroke Belt, ³ where vascular disease, poverty, structural inequity, and historical neglect intersect, implicating shared pathways of vascular and structural brain injury. Poststroke epilepsy, the most common cause of LOE, likely explains part of this geography but not all of it. ⁴

The clustering of LOE in the South implicates a broader constellation of contextual adversity. Heat index emerges as a high-importance predictor, with regions experiencing more days of extreme heat demonstrating higher LOE incidence. This pattern aligns with a growing body of literature linking climate to neurological vulnerability. ⁵ Heat, through its effects on hydration, sleep, cardiovascular strain, medication pharmacokinetics, and environmental safety, ⁶ may collectively lower seizure thresholds. In an era of accelerating climate change, this raises an intriguing question: might LOE, in part, be a climate-sensitive neurological disease?

Equally evocative is the role of transportation. Households without access to a vehicle are nearly twice as likely to reside in high-incidence areas, reflecting multiple intersecting mechanisms. Transportation barriers delay diagnosis, disrupt medication access, limit follow-up, and compound the burden of unmanaged comorbidities. In this light, transportation is not merely a logistical concern; it becomes a biological one, shaping the management of LOE risks and the trajectory of epileptogenesis.

A similar pattern is observed with insurance. Regions with higher rates of uninsurance, particularly in states without Medicaid expansion, demonstrate elevated LOE incidence. ³ These observations indicate that gaps in access to care earlier in life may reverberate decades later, allowing vascular and metabolic comorbidities to accrue unchecked until they manifest as epilepsy in later life.

Among modifiable determinants, insufficient sleep stands out as the strongest predictor. Regions with the highest burden of sleep deprivation exhibit nearly double the odds of elevated LOE incidence. Its disproportionate prevalence in the South and Appalachian regions mirrors the geography of high LOE incidence. ³ Chronic sleep deprivation is a well-established modulator of neuronal excitability, precipitating seizures in susceptible individuals. ⁷ Prevention, therefore, may need to move upstream, from antiseizure medications to sleep health itself.

Additional contributors, such as obesity and physical inactivity, ³ further reinforce the central role of vascular and metabolic health in shaping LOE risk. Together, these results point to LOE as a cumulative expression of systemic and environmental vulnerability.

Paradoxically, the study also identifies protective signals. Employment in agriculture or natural resource sectors and recent interstate migration are associated with lower LOE incidence. ³ These patterns may reflect healthier baseline populations, rural lifestyle, lower population density, differing environmental exposures, or, alternatively, underdiagnosis in regions with limited access to specialty care. The “healthy migrant” effect, in which movers represent a healthier subset of the population, offers one plausible explanation, but these findings invite further scrutiny. ⁸

Taken together, these diverse exposures converge on a shared pathway: a gradual lowering of the threshold for epileptogenesis in aging brains. This framework aligns with emerging models of LOE in which vascular, inflammatory, and neurodegenerative processes interact over time, ¹ even in the absence of a discrete structural lesion.

Perhaps the study's greatest strength is its methodological innovation: Max-P regionalization, which enabled analysis beyond the constraints of traditional state-level data used in prior analyses. ⁹ However, several limitations, including the ecological nature of the analysis, warrant caution. Associations at the MaxCounty level cannot be assumed to reflect individual-level causation, and unmeasured mediators likely contribute. Moreover, reliance on fee-for-service Medicare data and administrative coding introduces the possibility of selection bias and misclassification of epilepsy, particularly in underserved populations.

The demographic patterns further underscore the complexity of interpretation. Non-Hispanic Black residents were oversampled, appropriately, given prior data on racial disparities in epilepsy incidence. ¹⁰ While the proportion of non-Hispanic Black residents was identified as an important predictor in machine learning models, it did not retain independent significance in multivariable analyses, likely reflecting unmeasured structural and socioeconomic factors. ³ Rather than reassuring, this underscores disparities that remain incompletely captured and warrant targeted future investigation.

Despite its limitations, the significance of this study lies in its reorientation of our conceptualization of LOE risk. For decades, epilepsy has been framed within the brain, anchored in circuitry and the choreography of excitation and inhibition. Yet, this work compels us to expand our field of view. The environments in which people live, shaped by sleep, heat, physical activity, transportation, access to care, and inequality, emerge as determinants that accumulate over time and ultimately manifest as late-life seizures. These factors converge through interacting vascular, inflammatory, and metabolic pathways that promote epileptogenesis. In this view, epileptogenesis is not a single event but the endpoint of a long trajectory shaped as much by context as by biology.

The most provocative implication of this work is that contextual, modifiable, community-level factors may be among the strongest correlates of risk. If LOE is shaped by the environments in which people age, then prevention cannot remain confined to the clinic. Beyond stroke, neurodegeneration, and genetic risks, prevention must extend to sleep health, climate adaptation, heat mitigation, transportation infrastructure, and equitable access to care. We must prioritize modifiable and evidence-based targets that operate at scale.

Dong and colleagues have done more than map epilepsy. They have unsettled a paradigm. Their work suggests that the brain does not age in isolation and that its vulnerabilities are inscribed not only in proteins and pathways, but in neighborhoods, climates, and systems of care.

The next era of epilepsy research may not just map the brain; it may need to map the world around it. This shift raises critical questions: How do these contextual exposures interact with established biological risks? Which of these pathways are truly modifiable, and at what stage? And can altering the environments in which people age reduce the burden of LOE? These questions define the next frontier.

Once seen, this geography of risk cannot be ignored. It fundamentally reframes LOE from a disorder of neurons to a condition shaped, accumulated, and perhaps preventable within the places people call home.