This is an observational, cross-sectional study, using a self-administered online survey.

We surveyed a sample of adults residing in the United States via the Ipsos KnowledgePanel®, an online panel representative of the entire US population. Ipsos recruits panel members using probability-based sampling methods; potential panel members are provided with internet access and hardware if needed. Both English and Spanish speakers are recruited to the KnowledgePanel®.

For the present survey, we recruited KnowledgePanel® members who identified as White, Black, and/or Hispanic. We oversampled Black, Hispanic, and Spanish-proficient Hispanic adults to enhance racial and ethnic diversity and ensure adequate representation of non-English speaking adults in the sub-sample of respondents with symptoms consistent with long COVID. In addition, we oversampled Spanish-proficient Hispanic adults to confirm the findings of a prior pilot survey that suggested there may be disparities in long COVID awareness among Spanish speakers. Ipsos sent email invitations to panel members, with reminder emails to non-responders three and 7 days after the initial invitation. The survey was available in English and Spanish. Participants received an incentive valued at ~$5 for their participation. We conducted the survey between January 26 and February 5, 2023.

The primary outcome was awareness of long COVID, assessed with the question, “Have you heard of long COVID? This is also referred to as post-COVID, Long-haul COVID, Post-acute COVID-19, or Chronic COVID.” A secondary outcome was report of symptoms consistent with long COVID, defined by the presence of COVID-19 symptoms lasting longer than 1 month, among respondents with a history of COVID-19 infection. An additional secondary outcome was receipt of medical care among respondents with symptoms consistent with long COVID.

Co-variates included sociodemographic items (e.g., race, ethnicity, education, insurance status), beliefs related to long COVID, history of COVID-19 infection, COVID-19 vaccination status, and long COVID symptom duration and impact (among respondents with symptoms consistent with long COVID). Respondents who did not receive medical care for symptoms of long COVID were asked why they did not pursue care or were not able to see a provider for their long COVID symptoms (open-ended response). Those who received medical care for their long COVID symptoms were asked to provide an overall rating of the care they received with response options: excellent; very good; good; fair; and poor. Respondents who rated their care as less than excellent were asked the open-ended question, “Please tell us how your care fell short, and what should be done differently in the future.” The full set of items included in the survey for the present study are provided in Supplementary file S1. The survey items were pre-tested in a small (n ~ 300) study with a local sample. Demographic information (e.g., age, education level, household income) was provided by members when they joined the panel.

We constructed five categories to reflect participants’ reported race, ethnicity, and language in a single variable: White, non-Hispanic; Black, non-Hispanic; Hispanic, survey completed in English; Hispanic, survey completed in Spanish; other or more than one race, non-Hispanic.

For the question “Have you had COVID-19,” we combined ‘Yes, had it once’ and ‘Yes, had it more than once’ into a single category. We dichotomized insurance status into ‘no insurance’ and ‘any insurance’. To examine care seeking related to long COVID, we categorized participants as having symptoms consistent with long COVID if they responded that they had COVID-19 at least once and they had experienced COVID-19 symptoms that lasted for at least 1 month. We further categorized participants who met these criteria as (a) having sought and received medical care or (b) having not received medical care at the time of the survey (not sought or sought care but not received). We combined the latter two groups (not sought care and sought care but not received) because our goal was to predict receipt of care and describe the experience of care among those who received it. The group who sought care but had not received it was small (n = 20).

Spanish language responses to the open-ended items (e.g., Why did not you pursue care for your long COVID symptoms?) were professionally translated into English for analysis.

We used means, frequencies, and percentages to summarize participant characteristics. We computed chi-square statistics to examine the associations between participant characteristics and awareness of long COVID. To create a parsimonious model predicting long COVID awareness, we used a backward stepwise approach starting with standard demographic variables (race/ethnicity/language, age, gender, and education), and all additional variables that we hypothesized to be associated with long COVID awareness that also showed some evidence of possible association in the bivariate analyses (i.e., resulted in p < 0.20). The final model retained race/ethnicity/language, age, and gender independent of statistical significance. We used a similar approach to identify factors associated with receipt of medical care among respondents who reported symptoms consistent with long COVID. In addition to retaining standard demographics independent of statistical significance, this model also retained insurance status (no insurance vs. any insurance) given its relevance to receipt of care. For participants who received care, we computed descriptive statistics for their perceptions of care (taken seriously and satisfaction), whether they were diagnosed with long COVID and whether they were referred elsewhere for further care. Statistical analyses were performed with SAS version 9.4 (SAS Institute Inc., Cary, NC, United States) and SPSS version 28.0.1.0 (IBM SPSS Statistics, Armonk, NY, United States).

We used conventional content analysis to separately analyze responses to each open-ended item (25). At least two authors reviewed a subset of the responses and generated an initial set of codes. Four authors (KF, KM, LG, HA) reviewed and discussed the initial coding scheme; modifications and clarifying distinctions were made based on that discussion. Two authors (LG, HA) then independently coded all responses in Excel. These two authors met to compare codes and to identify disagreements and questions. Questions and disagreements which were not readily resolved were brought to the senior authors (KF and/or KM, both experienced qualitative researchers) for further discussion and resolution. We summarized the themes identified in reasons given for not seeking and receiving care, and themes in responses describing how care fell short for those who received care but rated the care as less than excellent. Coding of responses to open-ended items was performed in Excel.

There were 2,840 completed surveys among the 5,097 KnowledgePanel® members invited to respond, for an overall completion rate of 55.7%. The analytic sample for this study consists of 2,828 respondents who either responded to the question assessing long COVID awareness (n = 2,827) and/or responded affirmatively to experiencing symptoms consistent with long COVID (n = 272). Of these, the mean age was 50.4 years (range 18–94), approximately half were female (n = 1,493; 52.8%), 40.2% identified as Hispanic, 29.8% as Black, and 26.7% as White. Nearly one-fifth (n = 508; 18.0%) completed the survey in Spanish. Additional demographic characteristics are provided in Table 1.

Overall, 62.5% of respondents had heard of long COVID. Characteristics associated with having heard of long COVID on bivariate analysis are shown in Table 1. In multivariate analysis (Table 2), several factors were associated with increased odds of having heard of long COVID. For example, the odds of having heard of long COVID were more than twice as high in respondents with a master’s degree or higher (OR 2.21; 95% CI 1.49, 3.26) than in those without a high school diploma or GED, and those with an annual income of greater than $75,000 had a two-fold higher odds of having heard of long COVID (OR 2.08; 95% CI 1.38, 3.12) than those earning less than $10,000 annually. Additional factors associated with having heard of long COVID on multivariate analysis included having health insurance (OR 1.55; 95% CI 1.15, 2.08), a history of prior COVID-19 infection (OR 1.37; 95% C.I. 1.15, 1.64), and receipt of at least one dose of a COVID-19 vaccine (OR 1.34; 95% C.I. 1.06, 1.69). After multivariate adjustment, members of racial and ethnic minority groups remained significantly less likely to have heard of long COVID. Compared to White respondents, odds of having heard of long COVID were more than 25% lower among respondents who identify as Black (OR 0.64; 95% CI 0.51, 0.81), and odds of having heard of long COVID were 40% lower among those who identify as Hispanic and completed the survey in English (OR 0.59; 95% CI 0.46, 0.76). Respondents who identify as Hispanic and completed the survey in Spanish were least likely to have heard of long COVID (OR 0.31, 95% CI 0.23, 0.41) compared to White respondents who completed the survey in English.

Of the 1,365 respondents who reported a history of COVID-19 infection, 272 (19.9%) reported symptoms consistent with long COVID (any symptom lasting longer than 1 month). Of those with symptoms consistent with long COVID, 25.4% (n = 68) had not heard of it. Nearly half (n = 132; 48.9%) reported a symptom duration of 1–3 months, and most were either not limited at all (n = 88; 32.5%) or limited a little (n = 150; 55.4%) by their long COVID symptoms (Table 3). Approximately one-quarter (n = 72; 26.8%) reported having been seen by a healthcare provider for their symptoms of long COVID. The most common reasons for not having been seen by a healthcare provider for symptoms of long COVID included that it did not occur to the respondent to seek care (n = 121), their symptoms were mild (n = 15), they expected their symptoms would get better with time (n = 15), they did not think anything could be done (n = 7), they could not afford an appointment (n = 7), there was a long wait for an appointment (n = 5), and they were too busy with competing health issues or had not prioritized seeking care (n = 5).

Factors associated with having seen a healthcare provider for symptoms of long COVID on bivariate analysis are shown in Table 3. In multivariate analysis, factors associated with having been seen by a healthcare provider for long COVID included older age (OR 1.03; 95% CI 1.01, 1.05), symptom duration longer than 12 months (OR 2.62; 95% CI 1.24, 5.57), and being limited “a lot” by long COVID symptoms (OR 3.44; 95% CI 1.28, 9.28) (Table 4).

Of the respondents with symptoms consistent with long COVID who had been seen by a healthcare provider for these symptoms (n = 72), approximately one-half (47.2%) reported being diagnosed with long COVID, and 14 (19.4%) were referred for additional medical care elsewhere. Most (56/72; 77.8%) rated the care they received as less than excellent. Of those who provided a reason why care fell short (n = 30), the most common reason (n = 19) was dissatisfaction with the lack of knowledge (“Her answer was ‘not a lot is known about long COVID.’ What do you do with that answer?”) and/or treatments (“I feel like I still have symptoms and there are no medications to heal it”) for long COVID. Other reasons included feeling like the provider did not listen, take them seriously, or was dismissive (“the cough is still there to this day and the physician ignored or never addressed my concerns”; n = 7), and wanting more testing or evaluation of their symptoms (n = 4). Most respondents (n = 61; 84.7%) somewhat or strongly agreed that the provider they saw for symptoms of long COVID took them seriously, but some (n = 11; 15.3%) disagreed.