What Cross-Sectional Studies of Older Adults Cannot Tell Us

The National Academies’ 2020 consensus report put the figure at roughly one quarter of adults aged 65 and older living in a state of social isolation. AARP Foundation’s 2018 survey of 3,020 adults aged 45 and older found one in three reporting loneliness on the 20-item UCLA Loneliness Scale. Both numbers are widely quoted as if they describe a stable feature of aging. Neither can, on its own, say whether isolation or loneliness in later life is rising, falling, or simply persisting in the same people over time. That is not a flaw in either study. It is a limit built into the design.

The snapshot problem

A cross-sectional survey asks a group of people, once, how connected they feel and how large their networks are. The AARP survey did this well: a large sample, a validated instrument, and a genuinely useful finding that 33% of respondents who had spoken to their neighbors in the past week were lonely, against 61% of those who never had. That is a real association, and a striking one.

But a single timepoint cannot distinguish between several very different underlying stories. It could mean that lonely people are less likely to talk to neighbors. It could mean that not talking to neighbors makes people lonely. It could mean that some third factor — poor health, recent widowhood, a move to a new home — produces both the isolation and the reluctance to engage. The correlation is solid. The causal arrow is invisible in a single snapshot, and the same data would look identical under any of these explanations.

The National Academies report is explicit that its one-quarter figure for isolation among adults 65 and older is a prevalence estimate, not a trend line. It says how many people are isolated at a given moment, drawn largely from studies that measured isolation once. It does not say how long any individual has been isolated, whether isolation preceded a health decline or followed it, or whether the same quarter of the population is isolated from one year to the next.

What a longitudinal design would add

A longitudinal study follows the same people over time, measuring isolation or loneliness repeatedly rather than once, alongside health outcomes as they occur. This changes what can be claimed in several concrete ways.

First, it establishes sequence. If loneliness is measured at time one and cognitive decline or mortality is observed at time two, three, or four, a researcher can say the loneliness preceded the outcome — which a cross-sectional design, by definition, cannot. The meta-analyses that Julianne Holt-Lunstad has published draw much of their strength from pooling studies with this structure. The 2010 review, covering 308,849 participants across 148 studies, found that stronger social relationships were associated with a 50% greater likelihood of survival, an effect size on par with established risk factors like smoking. The 2015 review reported an odds ratio of 1.26 for loneliness and 1.29 for social isolation on early mortality, with the effects holding after adjusting for baseline health. Those numbers carry weight precisely because many of the underlying studies tracked people forward rather than asking once.

Second, a longitudinal design can separate a stable trait from a transient state. Cacioppo’s work frames loneliness as an aversive signal, evolved to motivate reconnection, comparable to hunger. If that framing is right, loneliness should function like a state that fluctuates — rising after a bereavement, falling after a move closer to family — rather than a fixed characteristic of a person. A single survey wave cannot tell the difference between someone who has been lonely for a decade and someone who became lonely three weeks ago. Repeated measurement can, and the distinction matters enormously for what kind of intervention would help.

Third, and most relevant to health policy, a longitudinal design lets researchers ask what happens to the same people as the health care system changes around them. The American Journal of Geriatric Psychiatry’s 2020 commentary on the National Academies report calls for routine clinical assessment of isolation and loneliness. Whether that assessment, once implemented, actually changes outcomes is a question only a study that follows patients before and after such screening could answer. A repeated cross-sectional survey — even a well-designed one, repeated annually with a fresh sample each time — would show whether population-level prevalence moved, but not whether any individual clinician’s identification of an isolated patient led to a different trajectory for that patient.

What is missing from the current evidence base

Much of what drives public attention to this issue — the AARP figure, the National Academies’ quarter-of-older-adults estimate — is cross-sectional or draws on studies that are. The mortality risk literature is stronger precisely because Holt-Lunstad’s reviews pool cohort studies that tracked participants forward. But a comprehensive prospective study specifically of older adults, using a consistent instrument at repeated intervals, linked to both self-reported loneliness and objectively measured network isolation, does not yet exist at the scale the mortality meta-analyses achieve for other outcomes.

What would strengthen the evidence is not a larger single survey but a cohort followed for a decade or more, with isolation and loneliness measured separately and repeatedly, allowing researchers to see who moves between states and what precedes those moves. Until that exists, the widely cited prevalence figures should be read as photographs of a single moment, not as a description of how any individual’s later years actually unfold.