Acceptability and Feasibility of Saliva-delivered PCR Coronavirus 2019 Tests for Young Children

ECE sites (N = 26) were recruited as part of the Rapid Acceleration of Diagnostic Testing – Underserved Populations Back to Early Care and Education Safely with Sustainability via Active Garden Education project (NCT05178290). ECE sites (N = 26) were located in Maricopa County, Arizona, where 31.1% of the population self-identifies as Hispanic or Latino, and 21% of total residents 5 years and older report using Spanish at home instead of English.²²  In 2020, Maricopa residents speaking Spanish at home were more than twice as likely to live below the federal poverty line (20.0%) than were those speaking only English at home (9.9%).²²  The research team recruited sites using ECE site contact information acquired through the Arizona Department of Education’s Child and Adult Care Food Program Site Directory.²³  ECE sites were selected based on their eligibility to participate in either Child and Adult Care Food Program or the National School Lunch Program to reach an intended low-income population. Proximity of sites to the Phoenix-metro area (≤1 hour travel time from the city center) was also considered in the selection process. Sites were then invited to participate via phone calls conducted by the research team offering federally approved quantitative polymerase chain reaction (molecular) diagnostics saliva testing.²¹  ECE sites were enrolled and scheduled for a 1-time testing event.

All participants provided written (teachers, staff, and parents) or verbal (children) consent. Parents were recruited from participating ECE sites to complete on-site saliva tests with their children; although both parents and children were encouraged to participate and generally did, either or both could participate or refuse to provide samples. In this study, “teacher” is defined as a person working within a licensed ECE site who is directly involved with teaching and caring for the children. Teachers distributed informational flyers and encouraged parents to participate. A total of n = 70 teachers, n = 227 parents, and n = 227 children enrolled in the study. Six children had recorded ages >71 months (ie, 6 years old or older), and 4 children had no recorded age; data from these children and their parents were not included in the analyses reported here, yielding samples of n = 217 parents and n = 217 children. Adults received a $20 gift card in compensation for their time, and children received stickers and a book for their participation. All protocols were approved by the Arizona State University Institutional Review Board (protocol ID: 00003761).

Demographic measures were drawn from the PhenX Toolkit to provide harmonized data across all Rapid Acceleration of Diagnostic Testing participating projects.²⁴  Demographic variables included parent and child age, annual family income, parent educational attainment, adult and child sex (female, male, intersex, none of these describes me, or prefer not to answer), ethnicity (Mexican, Mexican American, or Chicano, Puerto Rican, Cuban, another Hispanic, Latino, or Spanish origin, Salvadoran, Dominican, Colombian, or prefer not to answer) and race (American Indian or Alaska Native, Black or African American, Asian, Native Hawaiian or other Pacific Islander, white, some other race, or prefer not to answer). Participants were also asked whether they spoke a different language at home (Spanish, Vietnamese, Mandarin, Cantonese, Tagalog, Hawaiian, Ilokano, Navajo, or other).²⁵ 

Participants completed a measure created for this study to assess COVID-19 saliva test acceptability and feasibility. Adults were asked about satisfaction with COVID-19 saliva testing experience, the level of difficulty the participant and their child experienced producing the sample, satisfaction with the amount of time it took to complete the COVID-19 saliva test, satisfaction with the incentive amount, and whether the participant would recommend the program to a friend. Responses to questions were recorded using 5-point Likert scales to measure agreement with each item.

Children were asked about the level of difficulty they experienced, how much fun they had while producing a sample, and whether they would consider completing the test again (on a different day). These questions used a visual 5-point response scale with “smiley face” emojis ranging from very unhappy (depicted as a red frowning face) to very happy (depicted as a green smiling face) to indicate level of agreement (Supplemental Fig 1). All 5-point scales were converted to dichotomous indicators with a value of 1 through 3 representing a negative outcome (eg, not easy; too long) and a value of 4 or 5 representing a positive outcome (eg, easy; just right). The amount of time it took for the child to produce a sample was recorded by on-site research staff after obtaining permission from the parent and child.

Participants completed an in-person, interviewer-administered survey in either English or Spanish between December 2021 and June 2022 at ECE sites. Testing events were scheduled at ECE sites to coincide with pick-up and drop-off times. Bilingual (Spanish and English) community health workers offered culturally-grounded welcomes to participants, guided parents to the testing site, enrolled participants in the secure online results reporting system, supervised collection and storage of saliva specimens, and referred participants to the research team to complete a post COVID-19 testing survey. Research team staff, using a stopwatch timer, timed how long each child salivated into a small saliva collection tube.

Surveys were administered through REDCap HIPAA with a tablet computer connected to a secure mobile broadband network after teachers or parents and their children had completed the COVID-19 saliva test. Surveys were also conducted with participants who were unable to produce a saliva sample.

We first examined univariate and bivariate descriptive statistics to summarize scores on acceptability and feasibility indicators by group (participant’s preferred language: English versus Spanish; child’s age group) and for the entire sample. We also examined bivariate associations between parent-reported testing success (child was able to produce a saliva sample versus not) and indicators of acceptability and feasibility of the child testing procedure. Next, we examined intraclass correlation coefficients (ICCs) to assess the degree of clustering (nonindependence) of outcome scores at the ECE site level.

To test the statistical significance of the bivariate associations examined in our descriptive analyses, we estimated multivariable logistic and linear regression models and cluster-robust standard errors to account for observed site-level clustering (for dichotomous outcome measures, ICCs = 0.015–0.050; mean = 0.031, median = 0.028; for time required for the child to complete the saliva test, ICC = 0.308). Separate logistic regression models were used to examine the association between each of our dichotomous outcome indicators and participant’s preferred language, adjusting for adult participant’s age. Two parallel sets of logistic regression models were estimated for child testing feasibility measures with (1) child age (in months) as the focal predictor; and (2) parent-reported testing success for the child as the focal predictor. Linear regression models were estimated to examine associations between each of our focal predictors and time required for the child to complete the saliva test procedure, adjusting for parent’s age. Regression coefficients were estimated in R 4.1.3²⁶  using the glm and lm functions (for logistic regression and linear regression, respectively), and robust standard errors (“CR2” estimates with Satterthwaite degrees of freedom)²⁷  were estimated in R using the clubSandwich package (version 0.5.5.).²⁸ 

Models of parent-reported child outcomes (ease of testing, time required to complete testing) and child-reported outcomes drew on reports from parents. Models of adult outcomes drew on reports from parents and teachers. Because of item-level differences in response rates, analytic samples for regression models used data from varying numbers of participants (ns = 138–206) representing all 26 ECE sites, except for models of time required for the child complete the testing, which used data from a smaller subset of participants (n = 60) from 20 sites, and models of associations of successful testing completion with child acceptability and feasibility indicators (25 sites).

Demographic characteristics are presented in Table 1. On average, adult participants were approximately 36 years old (teachers: mean = 36.6 ± 14.7 years; parents: mean = 35.5 ± 9.1 years). Nearly 90% of teacher and parent participants were female, and slightly more than half identified as Hispanic or Latino. Among 56 teachers who reported their racial identity, white (n = 31; 55.4%), Black or African American (n = 5, 8.9%), and some other race (n = 13, 23.2%) were the most frequently endorsed options. Among 191 parents, Black or African American, White, and some other race were also the most frequently endorsed responses, but the proportion identifying as African American (n = 42, 22.0%) was higher than that among teachers. A large majority (n = 236, 82.2%) of participants preferred completing the study surveys in English, but about half (n = 122, 49.2%) reported speaking a language other than English at home. Overall, teachers tended to have slightly higher levels of education than parents; approximately 70% of teachers reported having some posthigh school education, including some college or technical school (n = 31, 47.0%), or a bachelor’s degree or higher (n = 15, 22.7%). Approximately 57% parents reported at least some posthigh school education, including some college or technical school (n = 67, 30.9%), or bachelor’s degree or higher (n = 57, 26.3%). Among child participants, there were slightly more girls (n = 114, 54.0%) than boys (n = 97, 46.0%) and, on average, child participants were 52.3 ± 8.1 months old (modal age group: 4 years old, n = 105, 48.4%).

Among adults, rates of acceptability for completing the test for themselves were high (76.3% to 95.5%) on nearly all dimensions. Overall, rates of acceptability were lower among children (58.0% to 73.2%) than among adults.

Patterns of proportions, means, and medians suggested that the testing procedure was more acceptable and feasible among 4- and 5-year-old children than 3-year-old children, and on most measures, more acceptable and feasible among 5-year-olds than 4-year-olds (Table 2). There was a clear age-group gradient in parents’ perception of how easily their children completed the testing with 25% of 3-year-olds’ parents, 45% of 4-year-olds’ parents, and 75% of 5-year-olds’ parents reporting that testing was easy for their children. Similarly, older children took less time to complete the test than younger children. Notably, 4- and 5-year-olds were nearly equally likely to rate the procedure as being fun to do (78.3% and 76.0%, respectively).

Patterns of proportions indicated that parents of children who completed the test by successfully providing a sample were much more likely to report that the test was easy for their children (64.2%) than were parents whose children did not provide a sample (7.7%; Table 3). Similarly, children who completed the test were more likely to report that the testing was easy, fun, and something they would like to do again than children who attempted, but did not complete the test.

Patterns of proportions suggested that, in general, English- and Spanish-preferring adult participants did not differ appreciably in their acceptability ratings (Table 4). However, English-preferring participants were somewhat more likely to report being satisfied with the testing experience than Spanish-preferring participants (87.6% and 75.3%, respectively). Conversely, Spanish-preferring parents were more likely to report that the testing procedure was easy for their children (50.0%) than were English-preferring parents (41.1%).

Children of Spanish-preferring parents were slightly more likely to report that the testing was easy and fun, as well as somewhat more likely to want to participate in the testing (65.7%) again on another occasion, than were children of English-preferring parents (55.3%).

As shown in Table 5, child age was significantly related to parent-reported child outcomes (ease of child testing, time required for child to complete testing). Parents of older children were more likely to report that the test was easy or very easy than parents of younger children. Although bivariable results for child-reported outcomes suggested that older children were more likely than younger children to report the testing was easy, fun, and something they would like to do again on another occasion, patterns were mixed across measures, and none of these associations was statistically significant after adjusting for parent age and ECE site-level clustering (P > .06).

The child’s successful production of a saliva sample was strongly associated with all parent- and child-reported child acceptability and feasibility indicators, even after accounting for parent age and site-level clustering (odds ratios = 3.25–28.73, P < .015; Table 6).

Results of these analyses are summarized in Table 7. There was some indication from bivariable relationships that Spanish-preferring adults were less likely to be satisfied with the COVID-19 testing than English-preferring adults, and children of Spanish-preferring parents completed testing faster and were more likely to report wanting to repeat the testing on another day than children of English-preferring parents; however, these associations were not statistically significant (all P > .13). Likewise, results for other outcomes did not support the hypothesis that acceptability or feasibility would differ depending on the adult participant’s preferred language.

This study is the first to investigate the acceptability and feasibility of using COVID-19 saliva tests for young children, their parents, and teachers at ECE sites serving low-income families, adding new information to existing calls for research to build evidence about school-based COVID-19 testing.²⁹  Overall, participants’ acceptability and feasibility ratings were generally high, consistent with reports from school-based studies.³⁰  Parents’ perceptions of how easy the testing was for their children had an ordinal association with child age, with parents of older children reporting that the test was easier for their children than parents of younger children. As expected, a child’s ability to produce a saliva sample, as reported by parents, was strongly related to both parent- and child-reported measures of acceptability and feasibility for both parents and children. Language preference was not associated with acceptability and feasibility ratings.

We expected to find that children who were a bit more developmentally advanced would have an easier time with the testing experience, and this was supported by our parent reports. Although young children were able to produce plenty of saliva, the concept of spitting into a small tube may have represented a novel skill, a possibility suggested by parents’ reports. In particular, parents of 4- and 5-year-olds reported that their children had little trouble with the test. Patterns of bivariable relationships of children’s responses with child age were mixed and not statistically significant, perhaps owing to small within-age group sample sizes. Also, the use of a smiley face visual analog scale to express their feelings about the testing experience may not have been intuitive for younger children, which could have resulted in higher levels of measurement error in this group, thereby attenuating or masking underlying associations.

As noted in this investigation and by others, people who are Hispanic or Latino, or predominantly speak Spanish, may feel less comfortable participating in studies such as this.⁵  However, this study did not reveal any differences according to Spanish language preference; this finding was supported in both patterns of bivariable statistics and in multivariable models. Strategies including the deployment of bilingual and bicultural staff to community sites to interact with participants likely reduced potential cultural and linguistic barriers that possibly could have undermined acceptability of the COVID-19 testing as observed in another study.³¹  Although not measured in this study, another study reported participation by this population may additionally be reduced by concerns about immigration status.³²  This study relied on bicultural and bilingual community health workers and research team staff for conducting the actual testing itself, as well as for administering surveys, successfully following best practices for recruiting and retaining hard-to-reach populations.^33,34 

Strengths of the study include careful attention to culture and language to optimize accurate data collection, interviewer administration of surveys, meeting participants at a familiar place (the ECE site), relying on a Food and Drug Administration-approved, noninvasive and highly reliable COVID-19 test, and clearly articulated analyses to account for potential clustering effects in a sizable sample of sites. This study also successfully oversampled underserved young children; 3 of every 5 parents identified as Hispanic or Latino, and 1 of 5 as Black or African American, providing important insight and information to guide future research and promotion efforts for historically underrepresented population groups. Limitations of the study include a smaller sample of younger, 3-year-old children and relying on parents’ reports of the children’s ability to produce a sample. Participants were volunteers and may not be a true representation of all families enrolled at the sites. Privacy considerations prohibited linking test results with survey data. Although the study employed a community-engaged approach designed to conduct research with underserved populations, prioritizing Hispanic families with young children, unforeseen barriers may have introduced biases that influenced the results.

In conclusion, this study demonstrated that saliva-based COVID-19 testing with young children, parents, and teachers at ECE sites is generally well received and may serve as an additional layer of protection to ensure safe return of children to ECE in the wake of the pandemic. Many (∼70%) US families rely on some kind of nonparental daytime care site for their preschool-aged (3–5 years) children, most of which are structured ECE or preschool.³⁵  In response to this need, our research provides important information to help young children return safely to ECE. Additional research should consider strategies to keep young children in school and reduce isolation and quarantine time as vaccination uptake may lag behind availability. Alternative strategies may be needed for younger children (0–3 years) to participate in ECE site-based, COVID-19 screening.

We thank the many community partners, teachers, families, and students and trainees who aided in the completion of this study; Erin Campbell, MS, provided editorial review and submission of this manuscript; Ms Campbell did not receive compensation for her contributions, apart from her employment at a collaborating institution supporting the publication of this study.

COVID-19

coronavirus disease 2019

ECE

early care and education

ICCs

intraclass correlation coefficients

SARS-CoV-2

severe acute respiratory syndrome coronavirus 2