What Switching Off Predicts—and What It Cannot Promise

This chapter examines the strongest temporal evidence in the subject: whether psychological detachment from work measured earlier is associated with mental well-being later. The aim is to read “predicts” precisely—to see what a two-wave cohort adds to a one-time correlation, which outcomes were linked to detachment, and why a protective association can coexist with worsening mental health during a large crisis.

A two-wave cohort adds time, not automatic causation

The evidence comes from the Wellbeing of the Workforce, or WoW, prospective longitudinal cohort study of working-age adults in the United Kingdom. Participants completed a first survey during the first national COVID-19 lockdown, between April and June 2020. A second survey took place during the third national lockdown, between March and April 2021. The two waves were separated by roughly a year and occurred under different phases of the same prolonged crisis.

There were 337 respondents at the first wave, and 169 completed the second survey. The participants were employed or self-employed and working in the UK; the study did not analyze people who were unemployed or completely furloughed. Psychological detachment was measured with the detachment subscale of the Recovery Experience Questionnaire. The later outcomes included psychological well-being, anxiety, life satisfaction, and self-rated general health.

The basic logic is straightforward. Measure detachment at the first wave. Measure whether detachment changed by the second wave. Then ask whether those earlier measures are associated with later outcomes after controlling for age, gender, and ethnicity. This ordering is stronger than asking people at one moment whether they both switch off and feel well. If the variables are measured simultaneously, we cannot tell whether detachment preceded well-being, well-being shaped people’s reports of detachment, or both were influenced by a third factor. A prospective design at least puts the proposed predictor before the later outcome.

sequenceDiagram participant Context as Crisis context participant Workers as Working-age adults participant T1 as First lockdown participant T2 as Third lockdown participant Model as Regression model Context->>Workers: Work and pandemic conditions continue T1->>Workers: Measure detachment at first wave Workers->>Model: Record first-wave detachment T1->>T2: Approximately one year passes T2->>Workers: Measure change and later outcomes Workers->>Model: Provide detachment change and outcomes Model->>Model: Control for age, gender, and ethnicity Model-->>T2: Associate detachment with later well-being, anxiety, and life satisfaction Model-->>T2: General-health association is not significant in the final model Note over Model,T2: Observational evidence does not prove causation

The sequence has two important features. First, detachment is not only measured at the start; improvement in the recovery activity is also considered. That allows the analysis to ask whether people who became better able to detach later reported different outcomes from people whose detachment did not improve. Second, the study does not treat “health” as one undifferentiated result. Mental well-being, anxiety, life satisfaction, and general health are separate outcomes, and the pattern was not identical across them.

What the design can isolate

It helps to separate three questions hidden inside the word “predicts.” The first is a starting-point question: among people facing the first lockdown, did those who reported more psychological detachment also tend to report better outcomes at the later wave? The second is a change question: among people who completed both surveys, did those whose detachment improved more tend to report better later outcomes? The third is a causal question: would a person have had a different later outcome if an intervention had increased their detachment while everything else remained the same?