What the Science Remote-Work Study Actually Gives You for a Cost Case

A study published in Science on 2026-03-01 estimates that the rise of remote work explains roughly a third of the increase in isolation and mental distress recorded between 2011–2019 and 2022–2024. Remote workers who live alone saw their days-spent-alone rise at roughly ten times the rate of workers who live with others. That is a specific, load-bearing number, and it will show up in board decks within weeks. The question for anyone building a cost case for connection programming is what it actually licenses you to say.

The honest answer: less than you’d like, more than nothing.

What the study supports

The study is population-level and correlational across a multi-year window, not a workplace intervention trial. It tells you that remote work has driven a measurable share of a national rise in isolation and distress, including use of mental health services and prescriptions. It does not tell you what happens if your organization runs a return-to-office mandate, a peer-connection programme, or anything else. It is evidence about a labour market trend, not a program evaluation.

That distinction matters because most cost cases fail by skipping it. They cite a population trend, then present a proposed intervention as if the trend itself were proof the intervention will work. The Science paper is strong evidence that remote arrangements carry a connection cost. It is not evidence about what closes that gap.

Building the case in the right order

1. Establish the baseline cost, not the programme benefit. Use the isolation figures to justify that a problem exists and is growing, not to size a return. Gallup’s 2024 workplace data adds a usable comparison point here: fully remote employees report loneliness at 25%, against 16% for fully on-site staff — a gap consistent with what the Science paper describes as a mechanism, not just a correlation.

2. Attach a cost only where the literature has actually measured one. Cigna’s 2020 workplace survey found lonely workers miss work twice as often for illness and five times as often for stress-related absence. That is a real, citable absenteeism pathway. Cigna’s 2025 follow-up puts loneliness among workers at 52%, which tells you the scale of exposure, not the size of any intervention effect.

3. Do not borrow mortality statistics for a workplace ROI slide. Holt-Lunstad’s 2015 meta-analysis on isolation and early mortality is real and serious, but a mortality hazard ratio has no place in a cost-per-employee calculation. It answers a public health question, not a staffing question. Funders and finance directors notice when a health statistic is doing work it wasn’t designed for, and it undermines the parts of the case that are solid.

4. Separate isolation from loneliness in your own language before someone else does it for you. Workplace isolation and loneliness are related but distinct constructs with different correlates — a 2024 study of healthcare workers under remote arrangements found perceived social support moderates the relationship between remote work and well-being, meaning the same working pattern produces different outcomes depending on what support already exists. A 2025 nationally representative study of remote work and loneliness among U.S. employees reinforces that the relationship isn’t uniform across the workforce. If your proposal treats “remote” as a single risk category, a sharp reader will ask why your intervention is aimed at isolation when your baseline data measures loneliness, or vice versa.

What this means in practice: use the Science study to justify that the problem is real, growing, and partly attributable to work arrangements. Use Cigna’s absenteeism figures to attach a rough cost. Do not use either to promise a specific return from a specific programme — that number does not exist yet.

The failure mode: the borrowed number

Call this the borrowed-number problem. A programme manager takes a statistic from a study about the problem and presents it as if it described the solution. The mortality hazard ratio becomes “our programme protects against a 30% higher death risk.” The one-third isolation figure becomes “our programme could recover a third of lost productivity.” Neither claim survives a second reading of the source, and once a funder catches one borrowed number, they discount the rest of the deck — including the parts that were accurate.

The safer structure is two separate sections: what the problem costs (backed by prevalence and absenteeism data), and what you propose to test (backed, if at all, by intervention evidence — which for workplace connection programming is thin). Say that plainly rather than papering over it with adjacent statistics.

Evidence status, at a glance

Claim	Evidence status
Remote work has increased isolation at population scale	Strong — Science, 2026, multi-year national data
Remote workers report more loneliness than on-site staff	Reasonably strong — consistent across Gallup 2024, Cigna 2020/2025
Lonely workers cost more in absenteeism	Established for the Cigna sample, not independently replicated at that magnitude
A specific workplace connection programme will reduce these costs	Weak — no controlled workplace trial evidence cited here supports a return figure
Isolation and loneliness respond to the same interventions	Not supported — the constructs are distinct and respond differently

What this does not solve

None of this tells you whether a given programme will work for your workforce, because the intervention evidence at population or workplace scale barely exists yet. It also says nothing about the people the cost case never reaches: contractors, shift workers, and staff who were never fully remote and so don’t show up in a remote-versus-on-site comparison at all. A cost case built only on the Science study will systematically miss them, along with anyone whose isolation has nothing to do with where they log in from.