A New Workplace Isolation Study, and the Missing Effect Size

A study published February 26 examined remote working, loneliness, workplace isolation, well-being, and perceived social support among healthcare workers. It found that workplace isolation and loneliness are separate constructs with different correlates, and that perceived social support moderates the relationship between remote working and well-being. Both findings are plausible and consistent with the wider literature. Neither comes with a number attached that indicates how large the effect is.

This is not a flaw specific to this paper. It is close to the norm in a research area that has spent the last several years establishing that associations exist, without always reporting how much they matter. The distinction matters most for a policy reader, and it matters more the further the research moves from mortality data — where effect sizes are unavoidable because the outcome is binary and final — into well-being measures, where a “significant” result can span anything from a trivial shift on a Likert scale to a clinically meaningful one.

What “significant” does not tell a reader

A cross-sectional design, which this study uses, can establish that two variables move together at one point in time. It cannot establish that one caused the other, and by itself a p-value below 0.05 says nothing about magnitude. A correlation of 0.08 and a correlation of 0.48 can both be statistically significant in a large enough sample; they describe very different worlds. The 2024 study’s contribution, moderation of the remote-work/well-being relationship by perceived social support, is a real and useful finding for interpreting a growing body of work on hybrid and remote employment. But without effect sizes reported in the source material available to the Center, it is not possible to say whether that moderation is large enough to change a staffing policy or too small to act on.

Compare this with the older adult literature, where the field has largely done the work of quantification. Holt-Lunstad’s 2010 meta-analysis, covering 148 studies and 308,849 participants, found that stronger social relationships were associated with a 50% increased likelihood of survival — a figure precise enough to compare against other known risk factors. Her 2015 follow-up assigned odds ratios: 1.29 for social isolation, 1.26 for loneliness, 1.32 for living alone, as risk factors for early mortality. The American Heart Association’s 2022 scientific statement went further, breaking the risk down by outcome: a 29% increased risk of heart attack and death from heart disease, a 32% increased risk of stroke. These are not just statistically significant findings. They are findings with a magnitude a clinician or policymaker can weigh against, say, the known risks of hypertension or smoking.

Why the gap exists

Part of the difference is subject matter. Mortality and cardiovascular events are hard endpoints; well-being and job satisfaction are soft ones, usually measured on scales that do not have an intuitive real-world unit. A one-point shift on a five-point loneliness scale does not mean the same thing to every reader the way a 29% increase in heart attack risk does. But part of the difference is also a habit of reporting. The National Academies’ 2020 consensus report on isolation in older adults was explicit about prevalence — roughly one quarter of adults 65 and older are socially isolated — precisely so that health systems could size an intervention against a known population. The AARP Foundation’s 2018 survey of adults 45 and older did something similar: 33% of people who have spoken with their neighbors report loneliness, against 61% of those who never have. That 28-point gap is itself an effect size, legible without a statistics course.

A 2023 review in BMC Public Health, mapping the state of loneliness and isolation research broadly, identified inconsistent measurement as a persistent barrier to comparing findings across studies. The workplace isolation literature, still younger than the older adult mortality literature by a decade or more, appears to be repeating an early stage of that same problem: establishing that relationships exist before settling on how to report their size in a comparable way.

What this means for reading the next study

None of this is a reason to dismiss the February 26 findings. The distinction between workplace isolation and loneliness as separate constructs is a genuinely useful conceptual clarification, and one the Center has flagged before as commonly conflated even in the older adult literature, where structural isolation and subjective loneliness predict outcomes independently rather than as proxies for each other. What the study cannot yet support, on the evidence available, is a claim about how much perceived social support offsets the costs of remote work, or how that offset compares to other known levers such as scheduled team contact or manager check-ins.

A stronger version of this study, and of the workplace isolation literature more generally, would report standardized effect sizes or, ideally, absolute differences in the outcome measures between exposed and unexposed groups, the way the AARP survey reports a 28-point gap in loneliness by neighbor contact. Until that becomes standard practice outside the mortality literature, readers should treat “significant” as a floor, not a finding, and ask the number that should have come with it.