Sample
For these analyses, we used publicly available data from three annual rounds of the Community Health Survey (CHS) conducted in 2018 and 2019 (before the COVID-19 pandemic) and 2020 (first year of the COVID-19 pandemic). CHS is an annual cross-sectional telephone survey conducted between March and December by the NYC Department of Health and Mental Hygiene (DOHMH). Each year, the CHS sampling frame is constructed from a list of household telephone numbers provided by a commercial vendor. A sample of households is selected via random sampling of the phone numbers associated with them, stratified on neighborhood as defined by the United Hospital Fund’s neighborhood designations. Selected households are contacted and one eligible adult per household is randomly selected and invited to complete a 25-minute computer-assisted telephone interview. Those interviews are offered in a variety of languages including English, Spanish, Russian, Chinese, Bengali, and Haitian Creole. Individuals \(<\)18 years old and those who do not reside in NYC, those in households without telephone service (land or cell phone), those living in group quarters (e.g., college dorms or nursing facilities), and institutionalized adults are not eligible to participate in the study [16].
Response rates (the proportion who participated among those selected, which includes those contacted and deemed eligible as well as those who could not be contacted) and cooperation rates (the proportion who participated among those who were successfully contacted and deemed eligible) were 8.4% and 82.8% respectively for 2018, 7.2% and 79.6% respectively for 2019, and 7.4% and 74.4% respectively for 2020 [16, 17]. CHS participants provided informed consent prior to completing the survey and study procedures were reviewed and approved by the NYC DOHMH institutional review board [16].
Measures
The outcome of interest was unmet healthcare, determined by an affirmative response to the question: “Was there a time in the past 12 months when you needed medical care [healthcare] but did not get it? Medical care includes doctor’s visits, tests, procedures, prescription medication, and/or hospitalizations” [18]. The exposure of primary interest was the first year of the COVID-19 pandemic (2020) versus the two years before the (2018–2019). We examined race/ethnicity and having health insurance as possible effect modifiers of the association between the COVID-19 pandemic and unmet healthcare need. Race/ethnicity was determined based on response to two survey questions: (1) “Are you Hispanic or Latino?” and (2) “Would you describe yourself as...” For the latter question, response options included White/North African/Mid-Eastern, Black, Asian/Pacific Islander, and Other. Responses to these two questions were combined into a 5-category race/ethnicity variable: White (reference), Black, Asian, Hispanic, and Other. Health insurance coverage was defined by a “Yes” response to the question: “Do you have any kind of health insurance coverage, including private health insurance or government plans such as Medicare or Medicaid?”
We also describe the population in terms of birth sex (female versus male), age (18–24 [reference], 25–44, 45–64, and 65 + years), education level (less than high school, high school graduate, at least some college or technical school [reference]), household income (< 200% Federal Poverty Level [FPL], 200–399% FPL, and ≥ 400% FPL [reference]), and not having a personal doctor/healthcare provider. Lastly, we looked at self-rated health (SRH) based on response to the question, “Would you say that in general your health is excellent, very good, good, fair, or poor?” Response options were dichotomized into an indicator for poor SRH (fair and poor) versus good SPH (excellent, very good, and good) [18].
Statistical Analyses
CHS data from 2018, 2019, and 2020 were merged and three-year combination weight and strata variables were obtained from the NYC DOHMH and added for analysis of the complex sample data. The population was described overall and by unmet healthcare need during the past 12 months. The statistical significance of differences in unmet healthcare was assessed using the Rao Scott chi-square test. Simple and multivariable logistic regression models evaluated crude and adjusted associations between the COVID-19 pandemic (2020 versus 2018–2019 survey years) and covariates with unmet healthcare. The population characteristics of New Yorkers were unlikely to change significantly between 2018–2019 and 2020, thus the multivariable model only adjusted for the hypothesized effect modifiers. However, in case demographic shifts occurred during the first year of the pandemic, perhaps due to an increase in residents moving away from NYC in 2020, we conducted a sensitivity analysis and ran a multivariable model adjusted for participant characteristics.
Race/ethnicity and health insurance coverage were evaluated as potential effect measure modifiers. We assessed three-way interaction among the variables by adding the 3-way interaction term (survey year*health insurance*race/ethnicity) and all possible 2-way interaction terms (survey year *race/ethnicity, survey year*health insurance, and race/ethnicity*health insurance) to the multivariable model. If the 3-way interaction term was significant, we reran the multivariable model stratified on both race/ethnicity and insurance. If it was not significant, we removed the 3-way interaction term and assessed the significance of the 2-way interaction terms, running the stratified models if significant. Analyses were conducted using SAS version 9.2 (Cary, NC), adjusted for the complex sampling, and weighted to the NYC population. Statistical significance was assessed at α = 0.05.