key: cord-0804239-3wm833dm
authors: Das, Abhery; Singh, Parvati; Bruckner, Tim A.
title: State lockdown policies, mental health symptoms, and using substances
date: 2021-08-12
journal: Addict Behav
DOI: 10.1016/j.addbeh.2021.107084
sha: 5a90f1cd2d5af87b592d28b99866b6de3e05fe41
doc_id: 804239
cord_uid: 3wm833dm

OBJECTIVE Previous literature finds an increase in depressive symptoms, substance use, and suicidal ideation following the COVID-19 pandemic in the US - suicides do not appear to increase. We examine whether 1) state lockdown policies in the US precede an increase in mental health symptoms; and 2) the extent to which using substances amplifies or attenuates the relation. METHODS We specified, as our exposure variable, the timing of state-level lockdown orders. We used, as the outcome variable, the 4-item Patient Health Questionnaire (PHQ-4) that measures anxiety and depression symptoms. We utilized the Understanding America Study (UAS), a nationally representative sample of 7,597 adults across 50 states in the US, surveyed biweekly between March 10, 2020 and November 11, 2020. Linear fixed effect analyses controlled for time-invariant individual factors, as well as employment status, household income, and previous mental health diagnosis. RESULTS Regression results indicate an increase in PHQ-4 scores of approximately 1.70 during lockdown, relative to no lockdown (p < 0.05). Relative to no lockdown, an increase in alcohol use corresponds with a 0.08 unit decrease in PHQ-4 scores during lockdown (p < 0.05). CONCLUSION State lockdown policies precede greater mental health symptoms. Increases in consuming alcohol attenuates the relation between state lockdown policies and mental health symptoms. Results may portend greater addiction following the pandemic warranting further investigation into utilization of substance use treatment.

The COVID-19 pandemic spurred state lockdown policies and stay-at-home orders in 43 of the 50 states in the US. 1 Implemented between March and April 2020, these policies aimed to restrict community activities and decrease the spread of COVID-19 by reducing population movement and contact with persons outside of the home. 1 Concerns of economic strain and significant changes to daily life such as social isolation, possible virus contraction, losing a loved-one, and home-schooling children may have exacerbated feelings of uncertainty and fear. 2 These factors have led scholars to speculate on the potential mental health consequences of viral outbreaks, such as COVID-19, as a 'parallel pandemic. ' 3 Large, cross-sectional studies report an increase in depression, substance use, and suicidal ideation following the COVID-19 pandemic in the US. 4, 5 However, overall suicide rates do not appear to rise in the US and even show a decrease in other countries. [5] [6] [7] [8] Concerns about the psychological fallout from the pandemic prompted urgent calls to examine the relation between mental health and substance use to better inform public policy, practice, and recommendations. 9 Within the self-medication hypothesis of addictive disorders, substances may temporarily relieve painful feelings or control emotions when absent or confusing. 10 The hypothesis complements other bio-genetic and sociocultural theories on the etiology of substance use disorders. It aims to address the emotional and psychological dimensions of addiction. 10 The hypothesis posits that 1) substance abuse may relieve psychological suffering; and 2) individuals may prefer particular substances depending on their psychoactive properties. 10 The cross-sectional literature finds greater adverse mental health symptoms and substance use after state lockdown orders relative to before the order. A nationally representative study in the US, for instance, finds a three-fold increase in depressive symptoms. 4 Studies examining state lockdown policies report adverse depressive symptoms following policy implementation. 11, 12 Researchers also report greater alcohol consumption following adoption of social distancing mandates. [13] [14] [15] [16] [17] Others report a positive association between psychological distress and substance use. [18] [19] [20] The above work, while suggestive, has three key limitations. First, cross-sectional studies do not follow individuals over time and therefore cannot track the time course of symptoms and substance use behaviors among individuals who newly experience the lockdown. To the extent that survey respondents before and after the lockdown differ systematically in ways that affect mental health and substance use, serial cross-sectional surveys may suffer from selection bias. Cross-sectional work may also be confounded by fundamental differences, across states, in mental health resources and other structural factors that could affect mental health symptoms. Additionally, prior work cannot address the important policy question of whether state-level changes to lockdown policies precede changes in mental health symptoms. The above work ignores the fact that state lockdown policies have changed over time, including cases in which many states have since lifted their policies. 21 Second, although some research focuses on consuming alcohol during COVID-19, literature on using different substances remains scarce. In addition to reports of increased alcohol consumption, cannabis sales increased dramatically at the beginning of the COVID-19 pandemic. 13, 14, 22 Sales then stabilized at figures 40% higher than 2019 numbers. 22 Previous literature has also grouped substance use with other avoidant or maladaptive behaviors -such as disengagement -masking their individual influence on mental health symptoms. Given that substance use can become addictive, 23 examining them individually may uncover potential disorders that follow.

Third, previous research does not examine whether, and to what extent, substance use may amplify or attenuate the relation between COVID-19 related stressors and mental health symptoms. Scholars report increases in substance use and mental health following the state lockdown policies and a positive relation between psychological distress and substance use. However, substance use may affect the relation between stress and psychopathology. It may buffer the relation or destabilize psychological adjustment to stress.

We address these limitations and analyze whether 1) state lockdown policies in the US precede an increase in mental health symptoms; and 2) the extent to which using substances amplify or attenuate any discovered relation between state lockdown policies and mental health symptoms. We examine the Understanding America Study (UAS), a nationally representative sample of adults across 50 states in the US, surveyed biweekly (over 16 waves), between March 10, 2020 and November 11, 2020. Importantly, these individuals provide continual participation over a long study period, which permits a rigorous longitudinal analysis of the evolution of mental health symptoms and substance use behaviors. Results from our study may hold particular relevance to understanding how state lockdown policies and using substances relate to mental health symptoms in the US.

Our study comprised respondents from UAS, a probability-based Internet panel of adults in the US. The study drew respondents from the universe of US postal addresses and provided them with internet and/or a tablet, if necessary. Beginning March 10, the UAS issued its first longitudinal survey on COVID-19. UAS then issued subsequent rounds beginning April 1, 2020, and every two weeks thereafter, to the same panel of respondents. UAS compensated panel members $20 for their participation in 30-minute surveys. 24 Additional details regarding methodology appear on the UAS website. 24 We used data from 16 waves of the study between March 10, 2020 and November 11, 2020 from respondents across all 50 states in the US. Of the 7,962 panel members eligible for the survey, our analytic sample comprised 7,597 who provided complete data on the variables of interest (95% completeness).

We used, as the mental health outcome variable, the 4-item Patient Health Questionnaire (PHQ-4) that measures anxiety and depression. 24, 25 The PHQ-4 asked respondents the number of days in the last two weeks, in which they have mental health symptoms. 25 The PHQ-4 total score is calculated across questions and ranges from 0 to 12, in which a higher score indicates more anxiety and depression. 25 Researchers confirm the reliability and measurement validity of the PHQ-4 scale. 25 Within the UAS dataset, the PHQ-4 has high internal consistency and scale reliability (Cronbach's alpha= 0.910).

We specified, as our exposure variable, the timing and duration of state-level lockdown orders. We retrieved start and end dates for state lockdown and stay-at-home orders, issued by US state governors, from Ballotopedia. 21 Ballotopedia, a non-profit organization, provides information on American politics and elections as a digital encyclopedia. With a staff of 100 professional editors, writers, and researchers, Ballotopedia regularly updates articles as new information becomes available. 21 Ballotopedia provided official state lockdown orders outlining mandates and closures, issued by governors in 43 of the 50 states that enacted policies. 21 Mandates included information on closure of non-essential businesses, travel restrictions, and public health practices. 21,26 Previous literature also uses these data, thus supporting reliability. 27, 28 We linked these data to the UAS based on whether respondents completed waves of the survey when their state enacted lockdown policies or not. We specified a binary indicator for whether respondents completed each of the 16 surveys during their own state's lockdown policy or not (binary indicator; 0 = no lockdown; 1 = during lockdown). We categorized respondents in the seven states that did not implement lockdown policies (Arkansas, Iowa, Nebraska, North Dakota, South Dakota, Utah, and Wyoming) as '0' for the entire study period.

We obtained data on substance use from survey questions on the UAS. 24 We measured using alcohol, cannabis, and other recreational drugs with the following question: Out of the past 7 days, what is your best estimate of the number of days that you did each of the following activities? 24 Responses included 0-7 days for 1) drank alcohol; 2) used cannabis products such as marijuana; and 3) used recreational drugs other than alcohol or cannabis products. 24

We first tested whether mental health symptoms increase following state lockdown policies. We used Ordinary Least Squares (OLS) regression with individual-level fixed effects specification to examine whether state lockdown policies correspond with mental health symptoms. Results from the Durbin-Wu-Hausman hypothesis test, commonly used to decide between fixed or random effects analyses in panel data, support the use of a fixed effects model. Our individuallevel fixed effects analytic strategy exploits the longitudinal "repeated measures" nature of the survey and compares mental health over time, within the same individual. This approach controls for time-invariant attributes that may correspond with mental health symptoms. Individual features such as preference for help-seeking, and sociodemographic attributes that do not vary over time are controlled using individual fixed-effects. 29 This strategy limits confounding from unobserved stable attributes that may influence mental health outcomes after COVID-19. We clustered standard errors by state to account for confounding that may arise from non-independence of individuals within states. Clustered standard errors, also known as the Liang-Zeger standard errors, corrects for possible correlation in modeling residuals.

Employment status and household income that changes over time, within an individual, may confound our analysis in that unemployment and economic strain may worsen mental health symptoms. 30 To control for this possibility, we utilized questions from the UAS regarding employment status and household income as control variables in our analysis. Additionally, we control for previous mental health diagnosis as that may influence mental health symptomology. We include indicators for month to control for month-specific factors (e.g., seasonality), or exogenous events, such as the George Floyd protests, which may also influence mental health symptoms. As a sensitivity analysis, we estimated whether lockdown corresponded with mental health symptoms after adjusting for outliers. We included PHQ-4 scores within the 5 th and 95 th percentile for this analysis. We also conduct an exploration of whether duration of lockdown (in weeks), among states that enacted lockdown policies, corresponds with greater mental health symptoms to assess a 'dose-response' relationship.

Second, we examined whether using substances (drinking alcohol, using cannabis products, or using recreational drugs) amplifies or attenuates any relation between state lockdown policies and mental health symptoms (separately for each substance). We used the same individual fixed-effects specification as described above, but now added an interaction term (using substances*lockdown policy), on the additive scale, as the key variable to examine the interaction. 31 This interaction term permits estimation of whether any relation between lockdown levels (i.e., lockdown, no lockdown) and PHQ-4 scores varies depending on the level of substance use. It also measures whether using substances strengthens or changes the relation between state lockdown policies and mental health symptoms. Third, we stratified analyses by heavy and non-heavy drinking to examine whether any relation between lockdown levels and PHQ-4 scores varies depending on the level of alcohol use. We removed all individuals who did not consume alcohol, for this analysis.

For all analyses, we used robust standard errors to adjust for heteroscedasticity in residuals and clustered standard errors by state. We performed all analyses using Stata SE version 16.0. The University of California, Irvine, Institutional Review Board deemed this study exempt owing to the use of publicly available, de-identified data.

Over the study period, 7,597 respondents participated for the full 16 waves. These respondents reflect the age, gender, racial/ethnic, and socioeconomic diversity of adults in the US (Table 1) . PHQ-4 scores among survey respondents average 1.94 across 16 waves of the survey (Table  1 ). Figure 1 shows average PHQ-4 scores from respondents across each wave of the survey. Scores initially increase after Wave 1 (i.e., greater reported anxiety and/or depression), but steadily decline in the subsequent waves. This general pattern indicates an acute rise in mental health symptoms in late March and an abrupt drop immediately thereafter, to the month of June.

States enacted and lifted lockdown policies at various times ( Figure 2 ). New York and California, for instance, enacted policies relatively early. Texas and Georgia enacted policies later, and the duration of lockdown policies in these two states lasted less than one month. This variation in initiation and termination of lockdown policies provides important spatial-temporal variation in exposure across UAS participants.

Respondents averaged 1.46 days consuming alcohol, 0.51 days using cannabis products, and 0.12 days using recreational drugs in the past week. Consuming alcohol, cannabis, and recreational drugs initially increases after Wave 1 and declines in later waves (Figure 3 ). Table 2 presents results from OLS fixed effects regression analyses predicting PHQ-4 scores as a function of state lockdown policies. Regression results indicate an increase in PHQ-4 scores of approximately 1.70 during lockdown as opposed to not during lockdown (p<0.05). This finding indicates that on a scale of 0-12, PHQ-4 scores increase by 1.70 during lockdown as opposed to no lockdown.

For a sensitivity analysis, we tested our model after removing outliers of PHQ-4 scores below the 5 th percentile and above the 95 th percentile. We find increases in PHQ-4 scores of 1.20 during lockdown as opposed to no lockdown following outlier adjustment (Appendix , Table A1 ). We also find attenuation for each category of household income and previous mental health diagnosis (Appendix , Table A1 ). These results indicate that very high or very low PHQ scores may have driven a portion of the overall effect size of the lockdown coefficient and that of the covariates reported in Table 2 .

As an exploration, we measure whether greater duration of lockdown (in weeks) corresponds with increased mental health symptoms. Among individuals living in states that enacted lockdown policies, we find no relation between duration of lockdown (in weeks) and PHQ-4 scores. This indicates that a 'dose-response' relationship may not exist (Appendix , Table A2 ). Table 3 (Models A-C) shows results predicting PHQ-4 scores as a function of the interaction between lockdown policies and consuming alcohol (Model A), cannabis products (Model B), and recreational drugs (Model C). Relative to before lockdown, increase in alcohol use corresponds with a 0.08 unit decrease in PHQ-4 scores during lockdown as opposed to no lockdown. This finding indicates that consuming alcohol slightly attenuates the relation between state lockdown policies and mental health symptoms. Although we fail to reject the null for cannabis and recreational drugs, the positive effect sizes indicate that cannabis and recreational drugs amplifies the relation between state lockdown policies and mental health symptoms (Table 3 , Model B and C).

We then stratified our analyses by heavy and non-heavy alcohol use to gauge whether the "buffering" role of alcohol use during the lockdown varies by the severity of alcohol consumption (Appendix , Table A3 ). Results suggest that the overall relation between PHQ-4 scores and the lockdown by level of alcohol consumption concentrates among non-heavy alcohol drinkers (Appendix , Table A3 ).

State governors issued COVID-19 related lockdown policies in 43 of the 50 states in the US. These policies likely spurred substantial lifestyle changes and economic uncertainty for many. Using substances may relieve painful emotions or exacerbate mental disorder. In this study, we examined 1) whether state lockdown policies preceded increases in mental health symptoms; and 2) the extent to which any relation between state lockdown policies and mental health symptoms varied by the level of substance use. We find that state lockdown policies correspond with an increase in PHQ-4 scores after the lockdown. Our results also indicate that consuming alcohol attenuated this relation. Alcohol consumption offers a very modest but statistically detectable 'protective effect' in that increase in frequency of alcohol consumption corresponds with fewer anxiety and depressive symptoms (i.e., lower PHQ-4 score).

The self-medication hypothesis of addictive behaviors offers a plausible explanation for this finding. 10 Individuals may consume alcohol as means of regulating negative emotions. This notion is further supported by the observation that the main detectable association concentrates among non-heavy drinkers. We caution against the interpretation that promotion of alcohol use may reduce anxiety and depressive symptoms. Rather, our findings indicate that individuals may consume alcohol to modestly regulate negative mood during state lockdown restrictions. We also note that our analyses describe relatively short-term relations (from March to November 2020) between alcohol consumption and mental health symptoms. These relations may change following prolonged exposure to the underlying stressor. 32 Additionally, our findings reflect mental health responses to lockdown policies that also appear in other countries such as Italy, United Kingdom, and Austria. [33] [34] [35] Studies conducted in these countries report great adverse mental health during lockdown restrictions, as opposed to before or after lockdown. [33] [34] [35] Similarly, studies on COVID-19 lockdown policies in the US also report a modest increase in mental health symptoms. One population-representative study finds a 0.085 standard deviation change in adverse mental health following lockdown orders, particularly among women. 11 Another study reports a minimal increase in depressive, rather than anxiety, symptoms among college students following Stay-At-Home Orders in Louisiana. 36 Although our study finds no significant differences in whether cannabis and recreational drugs moderate the relation between lockdown policies and mental health symptoms, the positive effect size indicates the potential for amplifying the relation. Previous studies, however, find conflicting results as to whether cannabis use precedes mental health symptoms. 37, 38 Our study also indicates that alcohol use modestly attenuates the relation between state lockdown policies and mental health symptoms, specifically among non-heavy drinkers. Prior works also finds that light-or moderate-alcohol users had fewer depressive symptoms in the presence of stress, suggesting that alcohol either attenuated the effect of stress on depression or suppressed the influence of the stressor. 39 Future work may benefit from better understanding differences based on the type of stressors.

Strengths of our study include the use of a nationally representative panel of respondents spanning all 50 states in the US. The longitudinal nature of the data allows for measurement of within-individual changes over time. This study design notably departs from previous COVID-19 and mental health literature which compares different adults over time or provides crosssectional information. Our analysis, by contrast, controls for selection bias due to individual differences in sampling and measures the evolution of health behaviors of the same individuals during the study period. The analysis also controls for confounding by trend, place, and time. Methods further control for time-varying factors such as household income and employment status, which may influence mental health symptoms. Taken together, our use of high-quality population-based data, longitudinal study design, and fixed effects analytic approach offer strong internal validity.

Limitations include that UAS questions regarding substance use do not specifically capture motive. This would allow for a better understanding regarding whether individuals may use substances for coping, interpersonal conflict, or fewer alternative activities. In addition, substantial macrosocial shocks may provoke or uncover psychiatric conditions by development of new disorders or exacerbation of pre-existing disorders. 40 We have controlled for previous mental health diagnosis in our analyses, but cannot differentiate between types of psychiatric diagnosis. Clinic-based studies with detailed medical history information, therefore, would complement population-representative surveys such as the UAS. We encourage such future research if and when more detailed and complete patient-level data become available.

Additionally, our study cannot account for social ties (i.e., friends, family members) that may also influence mental health symptoms during state lockdown restrictions. 41 Further research would benefit from evaluation of how social ties may have influenced substance use and mental health. Although we utilized rigorous methods (i.e., longitudinal data and fixed effects regression analysis), our data do not allow us to rule out the possibility that social networks may have partially accounted for the protective relation between alcohol use and adverse mental health following state lockdown restrictions.

Increased mental health symptoms following lockdown and the attenuation of symptoms consuming alcohol may portend greater addiction following the pandemic. 10 Although we fail to reject the null for cannabis and recreational drugs, the increase in cannabis sales and opioidrelated emergency room visits observed in other studies may contribute to greater substance use disorder. 22,42 The Affordable Care Act expanded treatment for substance use disorders through insurance coverage, enhanced parity, and opportunities to integrate primary care and psychiatric treatment. 43 In addition, rapid expansion of mental health and substance use treatment in Community Health Centers may also have provided low-cost and free care to medically underserved communities most affected by economic hardship. 44 However, punitive drug policies in the US have played a role in the incarceration, rather than treatment, of those with substance use disorders -especially in minority communities. 45 As the policy landscape changes to include legalized medical and recreational marijuana, 46 this shift may allow for use of expanded substance disorder treatment following the Affordable Care Act. The role of these policies and changes in the Federal healthcare landscape warrants further research. Whereas our study focuses on lockdown policies during COVID-19, examination of specific indicators such as local or regional unemployment rates may help uncover contributing factors that exacerbate mental disorder and/or using substances. In November 2020, unemployment in the US reached 10.7 million. 47 With additional economic uncertainty as COVID-19 cases rise and states implement further mandates, unemployment may precede increases in adverse mental health symptoms and suicidal behavior. Literature on this topic forecasts that a 5.7% worldwide unemployment rate would precede a predicted increase in more than 9000 suicides, globally. 48 This forecast, however, should be reconciled with the somewhat counterintuitive results of fewer (or no change in) suicides than expected following the COVID-19 pandemic, [6] [7] [8] as well as acute decreases in opioid-related ED visits following economic downturns. 49 Examination of potential psychiatric and substance use sequelae of the COVID-19 recession, as well as development of theory regarding collective responses to large societal shifts, need further investigation. 

Timing of State and Territorial COVID-19 Stay-at-Home Orders and Changes in Population Movement -United States

Mental Health of Communities during the COVID-19

Prevalence of Depression Symptoms in US Adults Before and During the COVID-19 Pandemic

Mental Health, Substance Use, and Suicidal Ideation During the COVID-19 Pandemic -United States

Suicide during the COVID-19 Pandemic in Japan. Public and Global Health

Trends in Suicide Rates during the COVID-19 Pandemic Restrictions in a Major German City. Psychiatry and Clinical Psychology

Racial Differences in Statewide Suicide Mortality Trends in Maryland During the Coronavirus Disease 2019 (COVID-19) Pandemic

Multidisciplinary research priorities for the COVID-19 pandemic: a call for action for mental health science

The self-medication hypothesis of substance use disorders: a reconsideration and recent applications

The Impact of the Coronavirus Lockdown on Mental Health: Evidence from the US

Mental health during the COVID-19 pandemic: Effects of stay-at-home policies, social distancing behavior, and social resources

Drinking to Cope During COVID-19 Pandemic: The Role of External and Internal Factors in Coping Motive Pathways to Alcohol Use, Solitary Drinking, and Alcohol Problems

The Lancet Gastroenterology & Hepatology. Drinking alone: COVID-19, lockdown, and alcohol-related harm

Effect of COVID-19 lockdown on alcohol consumption in patients with pre-existing alcohol use disorder

Changes in Adult Alcohol Use and Consequences During the COVID-19 Pandemic in the US

Drinking to cope with the pandemic: The unique associations of COVID-19-related perceived threat and psychological distress to drinking behaviors in American men and women

Depression and Its Relationship with Coping Strategies and Illness Perceptions during the COVID-19 Lockdown in Greece: A Cross-Sectional Survey of the Population

Coping, Mental Health, and Wellbeing in the UK during the pandemic

An Ultra-Brief Screening Scale for Anxiety and Depression: The PHQ-4

Stay-at-Home Orders Were Issued Earlier in Economically Unfree States

States Divided: The Implications of American Federalism for COVID-19

Introductory Econometrics : A Modern Approach

Economic contraction and mental health: A review of the evidence

Estimating interaction on an additive scale between continuous determinants in a logistic regression model

Adaptive and maladaptive coping

Mental health before and during the COVID-19 pandemic: a longitudinal probability sample survey of the UK population

Comparing Mental Health During the COVID-19

Lockdown and 6 Months After the Lockdown in Austria: A Longitudinal Study. Front Psychiatry

Students' mental health problems before, during, and after COVID-19 lockdown in Italy

Increases in distress during stay-at-home mandates During the COVID-19 pandemic: A longitudinal study

Cannabis use, depression and anxiety: A 3-year prospective population-based study

The association between cannabis use and depression: a systematic review and meta-analysis of longitudinal studies

Economic predictors of depressed mood and stressful life events in a metropolitan community

Opioid Overdose-Related Emergency Department Visits and Accidental Deaths during the COVID-19 Pandemic

The Affordable Care Act Transformation of Substance Use Disorder Treatment

Psychiatric Emergency Department Visits After Regional Expansion of Community Health Centers

The Black Family and Mass Incarceration

State Medical Marijuana Laws

Opioid-related emergencies in New York City after the Great Recession