Relationship between commuting and health outcomes in a cross-sectional population survey in southern Sweden

This was an observational study, not an experiment. The researchers compared different commuting modes (car, public transport, walking/cycling) and durations (one-way travel time) against six health outcomes. The "intervention" was simply the participants' usual commuting behaviour—no one was assigned to a specific commute. The reference group was people who walked or cycled to work in under 30 minutes (one-way). The outcomes were:

**Poor sleep quality** (perceived, self-reported)

**Everyday stress** (perceived)

**Low vitality** (feeling tired, worn out)

**Poor mental health** (psychological distress)

**Poor self-reported health** (global rating)

**Sickness absence** (any absence from work due to illness in the past 12 months)

The study also examined whether these relationships were "monotonous" (steadily worse with longer commute) or "concave downward" (worse at moderate durations, then slightly better at very long durations).

**Sample size:** 21,088 persons (after exclusions from an initial pool of ~37,000 survey respondents)

**Population:** Adults aged 18–65 living in Scania, southern Sweden

**Inclusion criteria:** Working >30 hours per week; completed the public health surveys in 2004 or 2008

**Exclusion criteria:** Not working, working <30 hours/week, missing data on commuting or outcomes

**Response rate:** 56% (meaning 44% of invited people did not respond, which introduces potential bias)

**Demographics:** The sample was roughly balanced by sex (not explicitly stated but typical for Swedish health surveys); socioeconomic status, family situation, overtime, job strain, and urban/rural residency were measured as covariates

**Setting:** Southern Sweden (Scania region), which includes both urban areas (Malmö, Lund) and rural/suburban areas

All measurements were self-reported via postal questionnaires. No objective measurements (e.g., actigraphy for sleep, cortisol for stress) were used.

**Commuting mode and duration:** Participants reported their main mode of transport (car, public transport, walking/cycling) and one-way travel time in minutes. Categories: <30 min, 30–60 min, >60 min. Walking/cycling <30 min was the reference.

**Poor sleep quality:** Single question: "Have you experienced poor sleep during the past 12 months?" (Yes/No). Not a validated scale like the Pittsburgh Sleep Quality Index.

**Everyday stress:** Single question: "Do you feel stressed in your everyday life?" (Yes/No).

**Low vitality:** Single question from the SF-36 health survey (a validated instrument): "Have you felt tired and worn out during the past 4 weeks?" (Yes/No, dichotomised).

**Mental health:** Single question from the General Health Questionnaire (GHQ-12): "Have you felt unhappy and depressed during the past 4 weeks?" (Yes/No, dichotomised).

**Self-reported health:** Single question: "How would you rate your general health?" (Good/Poor, dichotomised).

**Sickness absence:** "Have you been absent from work due to sickness during the past 12 months?" (Yes/No).

**Covariates:** Socioeconomic status (education, occupation), family situation (living with partner, children), overtime work (hours/week), job strain (Karasek's demand-control model), urban/rural residency.

**Why this matters for self-experimenters:** The reliance on single-question, yes/no outcomes is a major limitation. For your own experiments, use validated multi-item scales or objective measures (e.g., heart rate variability for stress, actigraphy for sleep). A single "Do you feel stressed?" question is noisy and unreliable.

**Study design:** Cross-sectional observational study using data from two population-based public health surveys (2004 and 2008). The surveys were pooled to increase sample size.

**No randomisation:** Participants were not assigned to commuting conditions. They self-selected their mode and duration based on where they lived, where they worked, and their personal preferences. This is the fundamental weakness: people with long commutes may differ systematically from people with short active commutes in ways that affect health (e.g., income, job type, personality, health status).

**No blinding:** Both participants and researchers knew the commuting status. Blinding is impossible in an observational study of commuting.

**Duration:** The surveys captured a single point in time. Commuting behaviour and health outcomes were measured simultaneously. This means the study cannot determine whether commuting causes poor health, or whether people with poor health choose (or are forced into) longer commutes, or whether a third factor (e.g., low income) causes both.

**Statistical approach:** Multivariate logistic regression. Odds ratios (ORs) were calculated for each outcome, comparing each commuting category to the reference (walking/cycling <30 min). Models were adjusted for age, sex, socioeconomic status, family situation, overtime, job strain, and urban/rural residency. The authors also tested for "monotonous" vs. "concave downward" relationships by examining the pattern of ORs across duration categories.

**What this design can prove:**

It can identify **associations** between commuting and health outcomes in a large, real-world population.

It can suggest **hypotheses** for future experimental or longitudinal research.

It can show **patterns** (e.g., public transport commuters >60 min have higher odds of poor sleep than car commuters of the same duration).

**What this design cannot prove:**

**Causation.** You cannot conclude that commuting causes poor health. Reverse causation (poor health leads to longer commutes) and confounding (e.g., low income forces both long commutes and poor health) are plausible alternative explanations.

**Direction of effect.** Does stress cause long commutes, or do long commutes cause stress? The cross-sectional design cannot answer this.

**Temporal sequence.** The survey asks about commuting "currently" and health "during the past 12 months." A person who recently changed their commute might report health problems from a previous period.

**Major methodological weaknesses:**

1. **Cross-sectional design** – cannot establish causality.

2. **Low response rate (56%)** – non-respondents may differ systematically (e.g., healthier people may be less likely to respond, or those with very long commutes may have no time to fill out surveys).

3. **Single-question outcomes** – poor reliability and validity compared to multi-item scales.

4. **Self-reported commuting duration** – people may misestimate travel time (e.g., rounding to the nearest 15 minutes).

5. **Dichotomisation of outcomes** – reduces statistical power and masks dose-response relationships (e.g., "poor sleep" vs. "good sleep" loses information about mild vs. severe problems).

6. **No objective health measures** – no blood tests, physical exams, or wearable data.

7. **Potential residual confounding** – despite adjusting for many covariates, unmeasured factors (e.g., personality, genetic predisposition to stress, neighbourhood walkability) could explain the results.

All results are odds ratios (ORs) comparing each commuting category to the reference group (walking/cycling <30 min). An OR of 1.0 means no difference; >1.0 means higher odds of the negative outcome. 95% confidence intervals (CIs) are reported where available. P-values are not reported for individual comparisons, but the authors state that "monotonous relations" were found for public transport.

**Primary outcomes (all six health measures):**

**Public transport commuting (vs. walking/cycling <30 min):**

**<30 min:** ORs ranged from 0.9 to 1.1 (mostly non-significant, i.e., similar to reference)

**30–60 min:** ORs ranged from 1.0 to 1.3 (small increases for some outcomes)

**>60 min:** ORs ranged from **1.2 to 1.6** across all outcomes. Specific estimates:

- Poor sleep: OR ≈ 1.4 (95% CI not reported)

- Everyday stress: OR ≈ 1.5

- Low vitality: OR ≈ 1.3

- Poor mental health: OR ≈ 1.6

- Poor self-reported health: OR ≈ 1.2

- Sickness absence: OR ≈ 1.3

**Pattern:** Monotonous (steadily increasing odds with longer commute time)

**Car commuting (vs. walking/cycling <30 min):**

**<30 min:** ORs ranged from 0.9 to 1.1 (similar to reference)

**30–60 min:** ORs ranged from **1.2 to 1.4** (peak negative effects)

- Poor sleep: OR ≈ 1.3

- Everyday stress: OR ≈ 1.4

- Low vitality: OR ≈ 1.2

- Poor mental health: OR ≈ 1.3

- Poor self-reported health: OR ≈ 1.2

- Sickness absence: OR ≈ 1.3

**>60 min:** ORs were **lower** than the 30–60 min category, ranging from 1.0 to 1.2 (i.e., car commuters >60 min reported fewer health problems than those commuting 30–60 min)

**Pattern:** Concave downward (inverted U-shape: worse at moderate durations, slightly better at very long durations)

**Sickness absence (secondary outcome):**

For both car and public transport, the relationship was concave downward: highest odds of sickness absence in the 30–60 min category (OR ≈ 1.3), with lower odds in the >60 min category (OR ≈ 1.1–1.2).

This pattern was consistent regardless of mode.

**Key nuance:** The "concave downward" pattern for car commuting >60 min is surprising. The authors speculate that people who commute very long distances by car may be a select group (e.g., higher income, more flexible jobs, better coping strategies) or may have adapted to the commute. Alternatively, it could be a statistical artefact (smaller sample size in the >60 min car group).

The odds ratios translate to the following approximate increases in risk (assuming a baseline risk of ~20% for poor sleep in the reference group):

**Public transport >60 min:** 20–60% higher odds of negative health outcomes. For example, if 20% of active commuters report poor sleep, then ~26–30% of long public transport commuters report poor sleep (absolute increase of 6–10 percentage points).

**Car commuting 30–60 min:** 20–40% higher odds. If 20% of active commuters report stress, then ~24–27% of moderate car commuters report stress.

**Car commuting >60 min:** Only 0–20% higher odds (smaller effect than moderate car commutes).

**In plain English:** The effect of a long public transport commute (>60 min) is roughly equivalent to adding a moderate daily stressor—like a demanding boss or a difficult coworker—to your life. The effect of a moderate car commute (30–60 min) is similar but slightly smaller. Very long car commutes (>60 min) do not show the same pattern, which may reflect a "healthy commuter effect" (only resilient people persist with very long drives).

**Comparison to other known effects:** The OR of 1.6 for poor mental health with long public transport commutes is comparable to the effect of job strain (OR ~1.5–2.0 in similar studies) or low social support (OR ~1.3–1.8). It is smaller than the effect of major life events like divorce or job loss (OR ~2.0–4.0).

**What the authors acknowledge:**

Cross-sectional design prevents causal inference.

Low response rate (56%) may introduce selection bias.

Self-reported commuting duration and health outcomes may be inaccurate.

Single-question measures for health outcomes are crude.

Residual confounding from unmeasured variables (e.g., personality, coping style, neighbourhood characteristics) is possible.

The reference group (walking/cycling <30 min) includes both physically active commuters and those with very short commutes, making it difficult to separate the benefits of exercise from the benefits of short travel time.

**What a critical reader would add:**

**No objective health data:** No biomarkers (cortisol, blood pressure, heart rate variability), no actigraphy for sleep, no clinical diagnoses. The outcomes are purely subjective.

**Dichotomisation of continuous variables:** Commuting duration was categorised into broad bins (<30, 30–60, >60 min), losing granularity. A person commuting 65 minutes is grouped with someone commuting 120 minutes.

**No dose-response analysis within the >60 min category:** The "concave downward" pattern for car commuting could be an artefact of grouping all >60 min together. If the negative effects actually increase linearly but then plateau, the analysis would miss this.

**Potential healthy worker effect:** People who are healthy enough to work >30 hours/week are already a selected population. The results may not apply to unemployed or part-time workers.

**Cultural specificity:** The study was conducted in Sweden, which has strong public transport infrastructure, labour protections, and social safety nets. Results may not generalise to countries with different commuting norms (e.g., the US, where car commuting is dominant and public transport is often poor).

**No data on commute quality:** A crowded bus for 60 minutes is different from a comfortable train with a seat and Wi-Fi. The study treats all public transport commutes as equivalent.

**No data on flexibility:** A person who can work during their commute (e.g., on a train) may have different health outcomes than someone who must drive in stop-and-go traffic.

**Potential for reverse causation:** People with poor mental health may choose (or be forced into) longer commutes because they cannot afford to live closer to work, or because they change jobs more frequently.

For someone running their own n=1 experiment:

### What to test

**Intervention:** Reduce your one-way commute time by at least 30 minutes (e.g., move closer to work, negotiate remote work days, switch to a different mode). Alternatively, if you currently commute by car for 30–60 minutes, try switching to public transport or active commuting (walking/cycling) for the same duration.

**Dose:** Aim for a one-way commute of <30 minutes, ideally by active transport (walking/cycling). If that's impossible, try <30 minutes by any mode, or >60 minutes by public transport with a seat and ability to read/work (to test if the negative effects are due to mode or duration).

**Comparator:** Your current commute (baseline). Measure for 2–4 weeks on your current commute, then change your commute for 4–8 weeks, then return to baseline for 2–4 weeks (A-B-A design).

### Minimum meaningful duration

**4 weeks per condition** (8 weeks total for a simple A-B comparison, 12 weeks for A-B-A). The study looked at health outcomes over the past 12 months, but changes in sleep, stress, and vitality can be detected within 2–4 weeks of a commute change.

**Longer is better:** 8 weeks per condition would capture adaptation effects (e.g., the "concave downward" pattern suggests some people adapt to very long commutes).

### What to measure (specific metrics)

**Sleep quality:** Use the Pittsburgh Sleep Quality Index (PSQI, 0–21, lower = better) weekly. Also track sleep onset latency, total sleep time, and wake-after-sleep-onset using a