Married but Disconnected: Landmark Study Maps Five Social Lives of Older Americans

Marriage has long been treated as a protective shield in old age, a reliable marker that someone is socially embedded and therefore healthier, sharper, and likely to live longer. A new analysis of more than ten thousand older Americans suggests that this assumption deserves a serious rethink. Researchers at the University of Southern California’s Davis School of Gerontology have identified five distinct patterns of social connection among adults aged sixty-five and older, and one of them stands out as a quiet warning: a large group of married older adults who, despite living with a partner, are profoundly disconnected from the wider social world. The study, published in BMC Geriatrics using data from the Health and Retirement Study, offers one of the most detailed portraits yet of how social lives are actually configured in later life, and why the conventional categories of isolated versus connected may be hiding as much as they reveal.

The research team, led by Dokyung Yoon with Elizabeth Zelinski and Teal Eich, took a deliberately different approach from most previous work in the field. Rather than treating each aspect of social life as a separate variable and asking how it correlates with health outcomes, they used a person-centered method known as latent profile analysis. This statistical technique searches for hidden subgroups within a population, groups of people who share a similar pattern across many measured characteristics at once. The logic is intuitive: two people with identical marital status may have utterly different social realities, and averaging across them can obscure the very configurations that matter for health. By modeling people rather than variables, the researchers could capture the full architecture of an individual’s social world in a single classification.

To build that architecture, the team drew on seven indicators spanning the two major dimensions that social scientists use to define connection. The structural dimension describes the observable scaffolding of a social life: whether someone is married or partnered, whether they live alone or with others, how frequently they contact children, other family members, and friends, and whether they participate in social activities. The functional dimension captures the subjective quality of that scaffolding: how much social support a person perceives themselves to have, and how lonely they report feeling. The distinction matters because the two dimensions can diverge dramatically. A person can be surrounded by relatives yet feel emotionally abandoned, or live entirely alone yet feel deeply supported by a few close ties. Structural measures alone miss the first case; functional measures alone miss the second.

The sample was drawn from the 2012 and 2014 waves of the Health and Retirement Study, a nationally representative longitudinal survey of Americans over fifty that has become the workhorse of aging research. The analysis focused on 10,145 community-dwelling adults aged sixty-five and older, with an average age of 74.48 years. From these participants, the latent profile analysis extracted five distinct profiles, each representing a characteristic constellation of the seven indicators. The largest group, accounting for forty percent of the sample, was married and moderately socially connected. A further twenty-seven percent were single but still moderately connected. Nine percent formed a group the researchers labeled most socially connected, combining partnership, frequent contact, participation, high perceived support, and low loneliness. Seven percent fell into the most socially disconnected profile, lacking both structural ties and functional support. And eighteen percent, nearly one in five older adults in the sample, occupied the profile that gave the study its most striking finding: married but socially disconnected.

This married and disconnected group is the study’s central surprise. Members of this profile had the structural marker most often equated with social protection, a spouse or partner, yet scored on the disconnected end of the functional spectrum, reporting elevated loneliness and weaker perceived support alongside limited contact with family, friends, and community activities. In earlier studies that relied on single indicators, these individuals would likely have been classified as socially embedded and aggregated into a protected category. The person-centered approach makes them visible, and their sheer size, nearly a fifth of the sample, suggests that a substantial population of older adults is being missed by screening approaches that ask only about marital status or living arrangements. The finding echoes a growing recognition in gerontology that loneliness and isolation are overlapping but non-identical conditions, and that neither can serve as a proxy for the other.

The researchers then asked what distinguished the profiles on a battery of sociodemographic, psychological, environmental, and health-related measures. Compared with the most socially connected group, members of every other profile were more likely to have less education and fewer economic resources. Psychological characteristics tracked connection as well: the less connected profiles showed greater neuroticism, more depressive symptoms, and worse cognitive scores, while scoring lower on extraversion. Even the perceived character of the surrounding neighborhood differed, with less connected groups reporting lower neighborhood cohesion and, notably, lower neighborhood disorder. The pattern held across profiles in a graded fashion, with the most socially connected group anchored at the favorable end of nearly every comparison and the most socially disconnected at the unfavorable end.

The association with cognition deserves particular attention. Social disconnection has been repeatedly linked in prior research to faster cognitive decline and increased dementia risk, and the present study adds weight to that literature by showing that worse cognitive scores characterize not just the extremely isolated but also the moderately connected and the married-but-disconnected groups relative to the most connected. The cross-sectional design of the analysis means causality cannot be established; poor cognition may constrain social engagement just as disconnection may erode cognitive reserve. But the consistency of the gradient across five distinct profiles, each defined by a different blend of structural and functional indicators, strengthens the argument that the full pattern of someone’s social life, not any single component, is what relates to brain health in aging.

Methodologically, the study demonstrates why latent profile analysis is gaining traction in social gerontology. Variable-centered approaches, such as regression models that estimate the effect of loneliness or marital status on an outcome, implicitly assume that the same relationship holds for everyone. Person-centered approaches relax that assumption and ask whether the population contains qualitatively different types. The five-profile solution here replicated much of what earlier studies had found, which the authors note as a strength, while also surfacing a profile that those studies had missed. That combination, replication plus discovery, is exactly what one wants from a new analytical lens: it confirms the robust features of the landscape while revealing terrain that older maps rendered invisible.

The practical implications are considerable. Interventions targeting loneliness in older adults are proliferating, from social prescribing programs to one-to-one befriending services, but they are typically designed for people identified as isolated or lonely by simple criteria. A married older adult with infrequent contact outside the household and a pervasive sense of loneliness may never be flagged, yet the data suggest such people number in the millions when national samples are considered. Screening tools that capture both structural and functional dimensions, and that treat the two in combination rather than in isolation, would better match the actual heterogeneity of older adults’ social lives. The study also hints at where interventions might look beyond the individual: neighborhood cohesion and economic resources both differentiated the profiles, suggesting that community-level and material factors shape the social worlds people can build.

What emerges from this research is a picture of social connection in later life as genuinely heterogeneous, structured into recognizable types rather than spread along a single continuum from isolated to connected. Most older adults in the sample were neither fully connected nor fully disconnected but occupied intermediate positions, and the specific configuration mattered. The married and disconnected profile, in particular, challenges the field to stop treating a wedding ring as evidence of a social life. As populations age across the developed world, and as health systems grapple with the documented consequences of disconnection for depression, cognition, morbidity, and mortality, studies like this one provide the finer-grained map that both researchers and practitioners have been missing. The next step, which the authors’ framework makes possible, is to track these profiles over time and determine which configurations predict which health trajectories, and for whom.

Subject of Research: Latent social connection profiles and their health correlates in older adults

Article Title: Identification and characteristics of social connection latent profiles in older adults: Findings from the Health and Retirement Study

Article References: Yoon, D., Zelinski, E., & Eich, T. (2026). Identification and characteristics of social connection latent profiles in older adults: Findings from the Health and Retirement Study. BMC Geriatrics. https://doi.org/10.1186/s12877-026-08404-8

Image Credits: AI Generated

DOI: 10.1186/s12877-026-08404-8

Keywords: social connection, loneliness, social isolation, older adults, latent profile analysis, Health and Retirement Study, cognitive function, depressive symptoms, marriage, gerontology, social support, neighborhood cohesion