2.1. Study Design and Selection of Participants
A cross-sectional web survey asked about COVID-19 health impacts, risk factors, food access, and concerns related to the social determinants of health. The survey was adapted from the validated National Food Access Research Team survey [24
]. Questions about food access and security were adopted directly from the National Food Access and COVID Research Team (NFACT) survey and validation analysis of the survey showed an alpha value of 0.70 [24
]. Additions to the survey included validated questions for anxiety and depression and a set of questions about the COVID-19 impact on employment and health adopted from COVID-19 surveys in the PhenX COVID-19 Toolkit, a battery of consensus measures for “Phen
otypes and eX
posures” vetted and assembled by expert working groups funded by the National Human Genome Research Institute [26
]. A quota-based non-proportional sample of 415 individuals in New York State, excluding counties in the New York City metropolitan area due to the vastly different context of city and state communities, was recruited by Qualtrics during the phased reopening stages of NY State on Pause [33
]. Quotas were set to oversample minority (50% Hispanic, 50% African American) individuals with low income or low education (50%) and male respondents (50%) to ensure a sufficient sample size to analyze the experiences of individuals at high risk of adverse COVID-19 risk and consequences. Research shows that in 2018, 89% of white, 88% of Hispanic, and 87% of Black Americans and 81% of Americans with an annual income of less than USD 30,000 used the internet [35
]; therefore, a web survey is appropriate for reaching the target population. Participants were recruited from survey panels maintained by Qualtrics [36
]. Panel members were eligible to participate if they were aged 18 or older and lived in New York State, excluding New York City. Potential participants were asked about their race, ethnicity, education, income, and gender to target filling the quotas set for the study. Participants that met the inclusion criteria but fell outside of the quotas needed for the sample were ineligible to continue. A total of 1,274 people began the survey, and 475 people were excluded due to ineligibility, such as not living in a recruitment county, not fitting in the quotas, or not consenting to participate. Another 325 participants were removed due to poor quality responses, such as speeding, straight-lining, or providing nonsense responses [37
]. A response rate is not available with panel data collection through Qualtrics because the research team does not have access to how many people were invited to participate in the survey. Data are only available on how many people started and completed the survey.
Race was assessed by asking respondents to select all self-identified races from a list of 13 options. Respondents were coded into race groups of white, Black or African American, other, and more than one race. Hispanic ethnicity was assessed by asking respondents to indicate if they identify as having Hispanic, Latino, or Spanish origins. Respondents were dichotomously categorized as Hispanic or not Hispanic. The variable for race and ethnicity for the present analysis was computed by classifying each respondent as non-Hispanic white, non-Hispanic Black or African American, Hispanic, or other or more than one race (white, Black or African American, Hispanic, other or multiracial).
Direct health impacts of COVID-19 were assessed by asking respondents if they knew anyone who had tested positive, been quarantined, been hospitalized, or died due to the virus and about current mental health. Respondents were classified as having a direct COVID-19 impact if they checked “self” for any of the impact categories and impact on family or friends if they checked “family” or “friend” for any category. Likely depression was assessed by the PHQ-2, a two-item screener for depressive disorders [31
]. A cut point of three was used to classify respondents with likely major depressive disorder (83% sensitivity and 90% specificity) [31
]. Anxiety was assessed with the GAD-2, a two-item screener for generalized anxiety disorder [30
]. A cut point of three was used to classify respondents with likely generalized anxiety disorder (86% sensitivity and 83% specificity) [30
Secondary health impacts were evaluated using the social determinants of health framework (Table 1
). Secondary health impacts include economic stability, education, healthcare, neighborhood and built environment, and social and community contextual factors. Economic stability included measures for income, reduced work and concerns about job security, housing, debt, and food access. Income was assessed by asking participants to select the income range that best described their income in 2019 before taxes (eight categories from less than USD 13,000 to greater than USD 150,000 were categorized as less than USD 25,000, USD 25,000–50,000, and greater than USD 50,000). Reduced work was assessed by asking respondents to check all that apply from a list of job impacts including working more hours, working less hours, furloughed, laid off, working from home, unemployed before the pandemic, or no changes (categorized as reduced work or no reduced work). Concerns about job security, housing, debt, and food security were assessed by asking respondents to indicate on a Likert scale how often (never, sometimes, most of the time, always) they have been concerned about the set of issues. For each issue, responses were categorized as ever (sometimes, most of the time, always) or never (never).
Education was assessed on a Likert scale by asking about how often (never, sometimes, most of the time, always) they were concerned about schooling and responses were categorized as ever (sometimes, most of the time, or always) or never (never). Healthcare was assessed by asking about concern about healthcare access and about health insurance status. Respondents were asked on a Likert scale how often (never, sometimes, most of the time, always) they were concerned about healthcare access since the pandemic began and responses were categorized as ever (sometimes, most of the time or always) or never (never). Health insurance status was assessed by asking respondents if they have public, private, or no health insurance. Health insurance was categorized as insured (public or private insurance) and uninsured (no insurance).
Neighborhood and built environment were assessed on a Likert scale by asking respondents about the frequency (never, sometimes, most of the time, always) of behaviors in the built environment including standing too close to others while getting food, going to bars and restaurants, making reduced grocery shopping trips, and need for more public transportation access during the pandemic. Responses were categorized as ever (sometimes, most of the time, always) or never (never) for each behavior. Access to public transportation was assessed by asking respondents how helpful (not helpful, somewhat helpful, helpful, very helpful) it would be for their household to have more access to public transportation and responses were categorized as helpful (somewhat helpful, helpful, or very helpful) or not helpful (not helpful).
Social and community context includes perceptions of response to the pandemic by different levels of government (city government, state government, federal government, public health such as the Centers for Disease Control and Prevention (CDC)) and communications about protecting households. Respondents were asked to indicate their level of agreement with a set of statements about response to the pandemic (strongly disagree, disagree, somewhat disagree, somewhat agree, agree, strongly agree). Respondents were classified as agreeing the response was effective if they indicated somewhat agree, agree, or strongly agree.