Characterising associations between mental distress, mobility, and COVID-19 restrictions: a U.S. study

This U.S. study reveals that higher levels of self-reported anxiety and depression during the first wave of the COVID-19 pandemic were significantly associated with slower recovery of pre-pandemic mobility, indicating that psychological distress influences population movement behaviors beyond the effects of formal restrictions and mortality rates.

The Gist

The Big Picture: The "Double Lockdown"

Imagine the United States during the first wave of the pandemic (Spring/Summer 2020) as a giant, bustling city that suddenly had to hit the brakes.

The government put on the first brake: official rules like "Stay at home" and "Close the shops." This is what we call policy stringency. Everyone knows this stopped people from moving around.

But this study discovered there was a second, invisible brake that the government didn't put on. It was the collective feeling of anxiety and depression among the people. The researchers found that even when the official rules were relaxed, people who were feeling more stressed and sad stayed home more than those who were feeling okay.

The Ingredients: What Did They Mix?

To figure this out, the researchers acted like detectives mixing three different data sources into a giant pot of soup:

1. The "Footprints" (Mobility Data): They looked at anonymized data from millions of mobile phones. Think of this as a giant map showing where people were walking, driving, or shopping every day compared to the same time the year before.
2. The "Mood Ring" (Mental Health Data): They used a massive online survey where people answered questions like, "Have you felt anxious in the last five days?" This gave them a daily "mood score" for every state.
3. The "Rulebook" (Policy & Deaths): They tracked how strict the government rules were (the Stringency Index) and how many people were dying from the virus.

The Experiment: The "Fixed-Effect" Filter

Imagine you are comparing two neighbors, Alice and Bob.

• Alice lives in a state with strict rules and high anxiety.
• Bob lives in a state with loose rules and low anxiety.

If you just look at them, you can't tell if Alice stays home because of the rules or because she's scared.

To fix this, the researchers used a statistical trick called a "Fixed-Effects Model." Think of this as a time-traveling camera. Instead of comparing Alice to Bob, the camera follows just Alice over time.

• It asks: "When Alice's anxiety went up this week, did she move less than she did last week, even though the government rules stayed the same?"
• By doing this for every state, they could isolate the effect of mood from the effect of rules.

The Findings: The "Sadness Brake"

The results were clear and surprising:

• The Official Rules Worked: When the government said "Stay home," people stayed home. (The first brake worked).
• The Mood Brake Worked Too: Even when the government said, "Okay, you can go out now," people in states with high levels of anxiety and depression still didn't go out as much as people in states where everyone felt calmer.

The Analogy:
Imagine a car driving down a hill.

• Policy Stringency is the driver pressing the brake pedal.
• Mental Distress is the car's engine stalling because it's out of gas.
• Even if the driver takes their foot off the brake (relaxes the rules), the car (the population) won't speed up if the engine (the people's mental health) is sputtering.

The Surprising Twist: Why Did High Death Rates Mean More Movement?

The study found a weird thing: States with more COVID deaths actually had more movement.

• Why? The researchers think this isn't because people were reckless. It's likely because people in those states couldn't stay home. They had to go to work to pay bills, they couldn't work from home, or they had to care for sick family members. They were "forced" to move despite the danger. It's like a runner who is injured but has to keep running because the finish line is their paycheck.

The Takeaway: Why This Matters

This study tells us that human behavior isn't just about following rules. It's about how we feel.

If a government wants to manage a crisis (like a pandemic or a natural disaster), they can't just look at the laws they pass. They have to look at the emotional weather of the population.

• If people are terrified and depressed, they will stay home even if you tell them to go out.
• If people are hopeful and supported, they might be more willing to follow rules.

In short: To get a population to move (or stay still) effectively, you have to fix the roads (policy) and fix the drivers' mental state (support and communication). You can't just drive the car; you have to tend to the engine.

Technical Summary

1. Problem Statement

Large-scale epidemics trigger significant psychological distress and alter human mobility patterns. While previous research has established that Non-Pharmaceutical Interventions (NPIs) like lockdowns reduce mobility and increase mental health issues, the specific relationship between population-level mental distress and effective mobility recovery remains under-explored.

Existing studies often rely on policy stringency indexes (e.g., Oxford Stringency Index) to explain mobility changes. However, these indexes fail to capture "spontaneous adaptation" and "adherence fatigue"—the phenomenon where individuals voluntarily reduce movement due to fear or psychological burden, even when formal restrictions are relaxed. This study aims to quantify the association between self-reported mental distress (anxiety and depression) and deviations from pre-pandemic mobility levels, controlling for policy stringency and mortality.

2. Methodology

Data Sources

The study utilizes a multi-source dataset covering 51 U.S. states and territories from April 1 to September 1, 2020 (the first wave of the pandemic):

• Mobility Data: Derived from anonymized mobile phone records provided by SafeGraph. The authors calculated the "active mobility" ($F_i$) for each county as the sum of trips originating from that county. This was aggregated to the state level and normalized against 2019 baseline data to account for seasonality, creating a relative change metric ($\Delta F(t)$).
• Mental Health Data: Sourced from the COVID-19 Trends and Impact Survey (CTIS) conducted by the Delphi Research Group (Carnegie Mellon University) in partnership with Facebook. The study focused on self-reported anxiety and depression levels (measured on a 4-point Likert scale) weighted by state, age, and sex.
• Control Variables:
  • Policy Stringency: The Oxford COVID-19 Government Response Tracker (OxCGRT) Stringency Index (0–100 scale).
  • Mortality: Daily confirmed COVID-19 deaths per capita (log-transformed) from The New York Times repository.

Statistical Model

The authors employed Fixed-Effects Panel Regression models to analyze the relationship between mental health and mobility.

• Model Specification:
  $$Mobility_{it} = \beta_1 MentalHealth_{it} + \beta_2 Mortality_{it} + \beta_3 SI_{it} + \alpha_i + \varepsilon_{it}$$
  Where $i$ represents the state and $t$ represents time.
• Fixed Effects ($\alpha_i$): Included to control for all time-invariant state-specific characteristics (e.g., geography, demographics, infrastructure, political leaning).
• Error Handling: Standard errors were clustered at the state level to account for serial correlation and heteroskedasticity.
• Data Cleaning: States with high multicollinearity (Variance Inflation Factor > 10) were excluded (Georgia, Maine, Oregon), resulting in a balanced panel of 48 states observed over 17 weeks (816 observations).
• Variables: Two separate models were run: one with Self-reported Anxiety and one with Self-reported Depression as the primary independent variable.

3. Key Contributions

1. Decoupling Policy from Behavior: The study demonstrates that formal policy stringency alone is insufficient to explain mobility dynamics. It isolates the specific contribution of psychological distress to mobility reduction.
2. Quantifying the "Psychological Gap": It provides empirical evidence that higher levels of anxiety and depression are statistically associated with lower mobility recovery, even when controlling for the strictness of government orders and the severity of the epidemic (mortality).
3. Methodological Integration: Successfully integrates large-scale digital phenotyping (mobile phone data) with self-reported survey data to model population-level behavioral responses during a health crisis.

4. Results

Descriptive Statistics

• Mobility declined sharply in March 2020, coinciding with the national emergency declaration and a spike in the Stringency Index.
• Despite the gradual relaxation of restrictions in the summer, mobility remained below 2019 baseline levels, suggesting factors beyond policy were keeping people at home.

Correlation Analysis

• Univariate Analysis: Spearman correlation revealed that in 75% of states, higher anxiety was significantly negatively correlated with mobility. A similar pattern was observed for depression (61% of states).
• Directionality: The vast majority of significant correlations were negative, indicating that as mental distress increased, mobility decreased.

Regression Analysis (Fixed Effects)

Both models yielded statistically significant negative coefficients for mental health indicators:

• Anxiety Model: $\beta = -7.195$ ($p < 0.01$). Higher anxiety levels are associated with significantly lower mobility outflows.
• Depression Model: $\beta = -4.801$ ($p < 0.01$). Higher depression levels are associated with significantly lower mobility outflows.
• Control Variables:
  • Stringency Index: Showed a significant negative association with mobility (stricter policies = less movement).
  • Mortality: Showed a positive and significant association with mobility. The authors interpret this not as a causal link where death causes movement, but rather as a reflection of structural constraints (e.g., essential workers in high-mortality areas could not stay home) or a lagged effect where high mobility preceded higher mortality.

5. Significance and Implications

• Behavioral Response Mechanism: The findings suggest that psychological well-being is a critical driver of adherence to public health measures. High distress can lead to "voluntary isolation" (reducing mobility beyond policy requirements) or conversely, "adherence fatigue" (though the study focuses on the former during the recovery phase).
• Policy Design: Public health strategies for future emergencies should not rely solely on restrictive measures. The study argues for integrated approaches that include mental health support, transparent risk communication, and social support to manage collective psychological states.
• Predictive Modeling: Incorporating mental health indicators into mobility models could improve the accuracy of epidemiological forecasts, as psychological states influence contact rates and transmission dynamics.

Limitations

The authors acknowledge several limitations:

• Self-Reporting: Mental health data is based on surveys, not clinical diagnoses, and may suffer from reporting bias.
• Causality: The cross-sectional design identifies associations but cannot definitively prove causality or disentangle bidirectional mechanisms (e.g., does low mobility cause depression, or vice versa?).
• Lag Effects: The model uses contemporaneous data and does not explicitly model temporal lags between mobility changes and mortality or mental health shifts.
• Granularity: The analysis is at the state level, potentially masking finer-grained local variations.