Comparative Diagnostic Utility of SARS-CoV-2 Rapid Antigen and Molecular Testing in a Community Setting

Abstract

Accessible and reliable diagnostic tests to detect severe respiratory syndrome coronavirus 2 (SARS-CoV-2) infection has remained a public health priority since the onset of the coronavirus disease 2019 (COVID-19) pandemic. Considered the gold standard for diagnosis of COVID-19, the real-time reverse transcription polymerase chain reaction (rRT-PCR) assay is a highly sensitive, laboratory-based, nucleic acid amplification assay that detects SARS-CoV-2 infection in both symptomatic and asymptomatic individuals [1]. In 2020, pandemic mitigation guidance from the Centers for Disease Control and Prevention (CDC) included recommendations for rRT-PCR testing following the onset of COVID-like illness symptoms, exposure to persons who recently tested positive for SARS-CoV-2, or before and after high-risk activities, such as travel or indoor gatherings [2]. Subsequently, rRT-PCR testing programs were scaled-up globally to meet the unprecedented demand for diagnostic testing. However, despite its high sensitivity, sustained and frequent use of rRT-PCR testing poses feasibility challenges due to cost and requirements for laboratory space, reagents, and trained personnel [3].

Antigen-detection rapid diagnostic tests (Ag-RDTs) are a point-of-care, self-testing option with advantages including timeliness of results, relative affordability, and convenience compared to rRT-PCR assays. Since the Food and Drug Administration issued its first emergency use authorization for an Ag-RDT in December 2020, at-home diagnostics to detect COVID-19 have become increasingly available and widely used [4]. Compared to rRT-PCR, initial Ag-RDT performance analyses against wild-type SARS-CoV-2 or the Alpha and Delta variants yielded sensitivities and specificities ranging from 50% to 80% and 97% to 100%, respectively [3, 5–8]. However, it is important to reassess Ag-RDT accuracy as new variants of concern (VOCs) with considerable mutations compared to ancestral SARS-CoV-2 emerge and predominate, such as Omicron [9]. Additionally, longitudinal characterization of Ag-RDT performance over the course of an illness episode in highly vaccinated, community-based settings remain incompletely understood. This study aims to evaluate Ag-RDT performance with rRT-PCR and identify characteristics associated with reduced diagnostic accuracy in the context of SARS-CoV-2 Omicron predominance.

METHODS

Study Design

The Husky Coronavirus Testing Study provided voluntary SARS-CoV-2 testing to students, faculty, and staff at the University of Washington, a large public university in Seattle, WA. The research study design, data collection, and laboratory methods have been previously described [10, 11]. Eligible participants were aged ≥13 years, had a valid university identification number, lived within a 100-mile radius of the Seattle or 2 nearby satellite campuses, and provided informed consent in English. Participants self-reported baseline demographic, social, and behavioral information through an electronic questionnaire, including sex, race, ethnicity, on-campus visit frequency, and household characteristics. Additional electronic questionnaires were administered regularly to ascertain updated eligibility and other demographic information. Electronic questionnaires and data management were conducted through REDCap [12, 13].

During the 2021–2022 academic year, participants received a daily attestation survey via email or text message and were invited to self-test following report of out-of-state travel, exposure to a known SARS-CoV-2 case, and new or worsening COVID-19 symptoms (Supplementary Table 1). Additionally, members of campus groups experiencing an outbreak were invited to test, and walk-in testing was available at any time. Self-collected anterior nasal swabs for rRT-PCR testing were supervised when conducted at staffed on-campus testing sites, and unsupervised for samples returned via on-campus drop boxes or by rapid courier. Nasal swab specimens were tested for SARS-CoV-2 using a laboratory-developed rRT-PCR test at the Brotman Baty Institute at the University of Washington and results were provided to participants through a secure electronic portal [14]. rRT-PCR–positive specimens with a high quantity of SARS-CoV-2 RNA underwent genomic sequencing, as previously described [10]. Beginning late February 2022, questions were added to the daily attestation survey to collect self-reported SARS-CoV-2 Ag-RDT testing dates and results. Ag-RDTs were acquired through pharmacies, government-supplied programs, or the university (Supplementary Figure 1). University-provided Flowflex Ag-RDTs were free and available to participants at rRT-PCR testing sites. Participants could complete Ag-RDTs at any time and report their results from the past 7 days through electronic daily attestations.

Data Analysis

Ag-RDT performance was assessed among participants having undergone ≥1 Ag-RDT within 7 days of rRT-PCR, which was used as the reference standard. We defined SARS-CoV-2 infection as a laboratory-confirmed rRT-PCR–positive result with a cycle threshold (Ct) <40, as previously described [10, 11]. For concordance analyses, each Ag-RDT was matched to the closest rRT-PCR by test date within 7 days. Positive and negative concordance were defined as an Ag-RDT result matching a positive or negative rRT-PCR result, respectively. Sensitivity and specificity estimates were calculated among Ag-RDTs performed within 1 day of rRT-PCR, and obtained using intercept-only logistic regression models fitted on the relevant subset of the data using generalized estimating equations (GEE) under an independence working correlation structure [15]. While such estimates agree exactly with empirical sensitivity and specificity estimates, the use of GEEs and robust standard errors facilitated the construction of 95% Wald confidence intervals (CIs) accounting for potential intraparticipant correlation from repeated sampling. For each stratification factor, comparisons of sensitivity and specificity across strata were performed using inverse variance-weighted multivariate Wald tests of the null hypothesis that all nonintercept coefficients are zero in GEE-fitted logistic regression models including only stratum indicators.

All symptom and vaccine status data were self-reported. Symptom status was defined relative to rRT-PCR testing and for the illness episode overall. Participants who reported symptom onset 7 days prior to or on the date of rRT-PCR testing were considered symptomatic. Individuals who reported symptoms presence in the 7 days after rRT-PCR testing were considered asymptomatic at rRT-PCR but symptomatic for the illness episode. A participant was considered fully vaccinated 2 weeks after completing the primary COVID-19 vaccine series and boosted 2 weeks after receiving a booster dose for fully vaccinated persons. Participants who received less than a full primary series or reported no prior COVID-19 vaccination at the time of rRT-PCR testing were categorized as unvaccinated, and those who provided invalid vaccination dates or did not report any information were classified as having an unknown vaccination status. Data cleaning and statistical analyses were performed in R and SAS.

RESULTS

Participant Characteristics

A total of 5575 participants who reported an Ag-RDT result within 7 days of rRT-PCR testing from 23 February to 14 December 2022, were included in this analysis (Table 1). Median age was 29 years (range, 18 to 82 years), over half of participants were female (67%), and most were white (60%) or Asian (26%). The sample of study participants included 54.3% students, including 3.6% fraternity and sorority community members, and 44.8% staff and faculty. Most participants were vaccinated against COVID-19 at the time of their first rRT-PCR test during the analysis period, including 78% (n = 4332) who received a monovalent booster and 14% (n = 800) who received the primary series only. Only 0.7% of participants were unvaccinated and vaccination status was unknown for 7.3%. University policies were updated making indoor masking optional starting 28 March 2022, in alignment with CDC guidance on 25 February 2022 [16, 17]. Despite relaxed COVID-19 mandates, many participants continued to sometimes or always adhere to nonpharmaceutical interventions throughout the study, with only 0.8% and 4.5% reporting never using a mask or social distancing during the analysis period, respectively.

SARS-CoV-2 Results and Clinical Characteristics

A total of 12 674 rRT-PCR and 17 572 Ag-RDT results from 5575 individuals were included in this analysis. A positive result was reported for 8% (n = 1350) of the 17 572 Ag-RDTs performed within 7 days of rRT-PCR testing. Of the 12 674 rRT-PCR samples, 995 (8%) were SARS-CoV-2 positive. The results of this analysis should be interpreted in the context of the predominant SARS-CoV-2 lineages circulating during the study period. Genomic sequencing of 584 (59%) rRT-PCR–positive samples from 515 participants identified Omicron BA.2 (n = 277, 47%), BA.5 (n = 134, 23%), and BA.2.12.1 (n = 114, 20%) as the predominant lineages. Among 12 674 rRT-PCRs, 43% (n = 5440) of participants were symptomatic at testing whereas 5% (n = 694) became symptomatic in the following 7 days, resulting in 6134 (48%) symptomatic illness episodes. Of the 6134 rRT-PCRs where participants had symptomatic illness episodes, 14% (n = 882) were rRT-PCR positive, representing 89% of the 995 positive results. Among these 882 symptomatic SARS-CoV-2-positive individuals, the most reported symptoms were sore throat (72%), cough (59%), and rhinorrhea/congestion (55%; Supplementary Table 1). The most predictive self-reported symptoms of rRT-PCR-positivity (number of positive results out of all tests where a symptom was reported within 3 days) were loss of taste or smell (26 of 62, 42%), chills (208 of 604, 34%), sweats (137 of 422, 33%), and feeling feverish (282 of 923, 31%). Similarly, the most predictive symptoms of Ag-RDT positivity were loss of smell or taste (38 of 113, 34%), sweats (176 of 715, 25%), chills (246 of 1018, 24%), and rash (14 of 60, 23%).

Ag-RDT Performance

Among 7704 Ag-RDTs performed within 1 day of 860 rRT-PCR–positive and 6844 rRT-PCR–negative tests from 3918 individuals, estimated overall sensitivity was 53.0% (95% CI, 49.6%–56.4%) and specificity was 98.8% (95% CI, 98.5%–99.0%; Figure 1). Adjusted for potential intraparticipant correlation, estimated positive predictive value (PPV) was 84.8% (95% CI, 81.4%–87.6%) and negative predictive value (NPV) was 94.4% (95% CI, 93.7%–95.0%) for 7.9% SARS-CoV-2 positivity in the study overall. Based on the overall sensitivity (53.0%) and specificity (98.8%) estimates, probability curves were constructed to estimate PPV and NPV for prevalence ranging from 2% to 14% (Figure 2). We also evaluated the probability of detecting SARS-CoV-2 infection with multiple Ag-RDT tests from days −1 to +1 and +7 of a positive rRT-PCR (Supplementary Table 2). The probability of a positive result for at least 1 out of all Ag-RDTs performed within 1 day was 62.5% (95% CI, 58.8%–66.1%) versus 66.0% (95% CI, 62.3%–69.3%) for all Ag-RDTs performed through the 7 days after a positive rRT-PCR.

Figure 1.

Sensitivity and specificity of SARS-CoV-2 Ag-RDT performed within 1 day of rRT-PCR testing, n = 7704 testing instances from 3918 individuals, adjusted for intraparticipant correlation. *Symptomatic if reported symptoms at least once with onset in the 7 days prior to or on rRT-PCR test date. †Symptomatic if reported symptoms at least once with onset between the 7 days prior to and 7 days following rRT-PCR test date. ‡First test date of each rRT-PCR to Ag-RDT match. §COVID-like illness indicated if participant reported acute onset of (1) at least 1 symptom of cough, loss of taste or smell, difficulty breathing or chest pain, or (2) at least 2 symptoms of fever, chills, muscle or body aches, headache, sore throat, nausea or vomiting, diarrhea, fatigue, or runny nose within ± 3 days of rRT-PCR. ¶Supervised if in-person sample collection at a study testing site, and unsupervised if the sample was returned via dropbox or rapid mail courier. Abbreviations: Ag-RDT, antigen-detection rapid diagnostic test; CI, confidence interval; Ct, cycle threshold; COVID, coronavirus disease; rRT-PCR, real-time reverse transcription polymerase chain reaction; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.

Open in new tabDownload slide

Figure 2.

Estimated PPV and NPV probabilities by SARS-CoV-2 prevalence at the overall antigen-detection rapid diagnostic test sensitivity of 53.0% and specificity of 98.8%. Abbreviations: PPV, positive predictive value; NPV, negative predictive value; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.

Open in new tabDownload slide

Ag-RDT sensitivity varied by symptom status at rRT-PCR and was higher for symptomatic (53.9%; 95% CI, 50.3%–57.4%) versus asymptomatic (44.0%; 95% CI, 32.3%–56.4%) individuals (P > .05). COVID-like illness was defined as acute onset of at least 1 symptom of cough, loss of taste or smell, difficulty breathing or chest pain, or at least 2 symptoms of fever, chills, muscle or aches, headache, sore throat, nausea or vomiting, diarrhea, fatigue, or runny nose within 3 days of RT-PCR testing [18]. Ag-RDTs were 56.3% (95% CI, 52.4%–60.2%) sensitive among participants who met the COVID-like illness definition and 43.9% (95% CI, 37.5%–50.5%) for those who did not (P < .01). Among symptomatic participants, sensitivities were significantly different by days from symptom onset to first test date of each Ag-RDT to RT-PCR match (P < .001). Sensitivity was 41.2% (95% CI, 35.3%–47.4%) on the day of symptom onset and increased to 70.1% (95% CI, 58.1%–79.9%) for tests used 4 to 7 days after. Conversely, estimated specificity was high for both, but slightly higher among asymptomatic (99.6%; 95% CI, 99.4%–99.8%) than symptomatic (97.8%; 95% CI, 97.3%–98.3%) persons (P < .001).

Estimated Ag-RDT sensitivity and specificity were comparable regardless of COVID-19 vaccination status and supervised versus unsupervised rRT-PCR sample collection (P > .05; Figure 1). Estimated sensitivity was highest when the Ag-RDT was performed 1 day after a positive RT-PCR (69.0%; 95% CI, 59.9%–76.9%) versus the same day (62.0%; 95% CI, 57.2%–66.6%) or 1 day before (38.3%; 95% CI, 33.5%–43.5%; P < .001). In contrast, estimated specificity was highest for Ag-RDTs performed 1 day before a negative rRT-PCR (99.5%; 95% CI, 99.1%–99.7%), compared to 1 day after (95.7%; 95% CI, 94.3%–96.8%) or the same day (99.3%; 95% CI, 98.9%–99.5%; P < .001). The association between Orf1b Ct, analyzed categorically, and sensitivity was assessed among individuals who performed an Ag-RDT within 1 day of an rRT-PCR–positive sample. Lower Orf1b Ct (ie, higher semiquantitative viral loads) were associated with notably higher estimated sensitivity: 82.7% (95% CI, 72.0%–89.8%) for Ct ≤20 compared to 36.5% (95% CI, 30.4%–43.0%) for Ct between 30 and 35 (P < .001; Figure 1). Mean Ct values were lower for rRT-PCR–positive tests with a concordant Ag-RDT-positive result within 1 day and were lowest among those performed 3 days after symptom onset (24.3, [SD 6.3] cycles; Figure 3).

Figure 3.

Distribution of Orf1b Ct as a surrogate marker for inverse of viral load by days from symptom onset to rRT-PCR among rRT-PCR positives with an Ag-RDT performed within 1 day, stratified by Ag-RDT result. Abbreviations: CT, cycle threshold; rRT-PCR, real-time reverse transcription polymerase chain reaction; Ag-RDT, antigen-detection rapid diagnostic test.

Open in new tabDownload slide

Among 17 572 Ag-RDTs matched to the closest rRT-PCR test date within 7 days, negative concordance (95.6%; range, 89.9%–99.0%) was higher than positive concordance (52.9%; range, 36.9%–73.0%; Figure 4A and Supplementary Table 3). Positive concordance was highest when the Ag-RDT was performed 1 to 7 days after rRT-PCR (69.2%; range, 62.9%–73.0%), compared to 1 to 7 days before (40.2%; range, 36.9%–67.7%) or the same day (59.3%). Positive concordance was low for asymptomatic individuals (39.3%; range, 22.0%–66.7%) and among symptomatic individuals when Ag-RDT was performed 1 to 7 days prior to rRT-PCR (39.9%; range, 36.5%–68.0%; Figure 4B and 4C, and Supplementary Table 4).

Figure 4

. A–C, SARS-CoV-2 Ag-RDT concordance with rRT-PCR by difference in days between tests stratified by rRT-PCR test result and symptom status, n = 17 572 rRT-PCR to Ag-RDT comparisons from 5575 individuals. Each Ag-RDT was matched to the closest rRT-PCR by test date within 7 days; 95% confidence intervals were adjusted for potential intraparticipant correlation using GEE methods. Symptomatic includes individuals who reported any symptoms within 7 days of rRT-PCR testing. Abbreviations: Ag-RDT, antigen-detection rapid diagnostic test; GEE, generalized estimating equation; rRT-PCR, real-time reverse transcription polymerase chain reaction; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.

Open in new tabDownload slide

Serial Testing

Serial testing was examined longitudinally over a 15-day period among 756 Ag-RDTs with ≥1 discordant result within 1 day of 177 rRT-PCR–positive tests, for which estimated overall sensitivity was 51.2% (95% CI, 47.7%–54.7%; Figure 5 and Figure 6). Sensitivity was significantly higher with Ag-RDT use 1 to 7 days after rRT-PCR (61.9%; 95% CI, 55.7%–67.7%) compared to 1 to 7 days before (38.2%; 95% CI, 32.2%–44.6%) and same day testing (60.0%; 95% CI, 49.1%–70.0%; P < .001). Likewise, sensitivity was significantly different by time from Ag-RDT to symptom onset (P < .001). Ag-RDTs were least sensitive when used 1 to 7 days before (12.5%; 95% CI, 6.5%–22.7%) and the same day as symptom onset (26.1%; 95% CI, 18.0%–36.3%). Sensitivity increased substantially when Ag-RDTs were used 1 to 3 days (62.3%; 95% CI, 53.3%–70.4%) and 4 to 7 days (82.6%; 95% CI, 75.7%–87.9%) after symptom onset, but were only 49.6% (95% CI, 39.6%–59.7%) sensitive thereafter. Among those who serially tested following an initial Ag-RDT–negative result, subsequent Ag-RDTs were 68.5% (95% CI, 62.0%–74.3%) sensitive when performed at least 2 days later but only 34.3% (95% CI, 24.1%–46.2%) sensitive before 2 days (P < .001). Similarly, sensitivity was significantly higher with repeat testing at least 4 days after an initial Ag-RDT-negative result (75.8%; 95% CI, 68.4%–81.9%) versus before 4 days (43.4%; 95% CI, 36.0%–51.1%; P < .001).

Figure 5.

SARS-CoV-2 Ag-RDT serial testing among participants with at least 1 discordant Ag-RDT within 1 day of rRT-PCR–positive test, n = 177 rRT-PCR–positive tests matched to 756 Ag-RDT results. Abbreviations: Ag-RDT, antigen-detection rapid diagnostic test; rRT-PCR, real-time reverse transcription polymerase chain reaction; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.

Open in new tabDownload slide

Figure 6.

Sensitivity of serially performed SARS-CoV-2 Ag-RDTs compared to rRT-PCR testing, n = 177 rRT-PCR–positive tests matched to 756 Ag-RDT results, adjusted for intraparticipant correlation. Abbreviations: Ag-RDT, antigen-detection rapid diagnostic test; CI, confidence interval; rRT-PCR, real-time reverse transcription polymerase chain reaction; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.

Open in new tabDownload slide

DISCUSSION

This prospective longitudinal study assessed characteristics associated with Ag-RDT performance in a highly vaccinated university population when the SARS-CoV-2 Omicron variant lineages predominated on campus. Estimated Ag-RDT sensitivity and specificity were 53% and 99%, respectively, compared to rRT-PCR. Our findings suggest substantial differences in Ag-RDT performance by clinical characteristics and testing patterns. Sensitivity was notably higher for symptomatic (54%) versus asymptomatic (44%) testing and those with lower Ct values (ie, higher SARS-CoV-2 loads; >80% sensitivity for Ct ≤20). Ag-RDT performance differed by testing order, where sensitivity was significantly higher for Ag-RDTs performed 1 day after rRT-PCR (69%) compared to 1 day before (38.3%).

Among symptomatic cases, sensitivity varied throughout the illness episode. We showed that Ag-RDTs do not sufficiently identify rRT-PCR–positive cases during early symptomatic illness and that sensitivity peaked at 70% when the Ag-RDT was performed 4 to 7 days after symptom onset. Sensitivity was <50% when Ag-RDTs were conducted before or on the day of symptom onset. Negative concordance declined slightly in the days following rRT-PCR among symptomatic but not asymptomatic individuals, which may suggest those with discordant Ag-RDTs received an initial rRT-PCR–negative result in the days before SARS-CoV-2 became detectable and subsequently identified by Ag-RDTs. These findings highlight the importance of serial rapid antigen testing after an initial Ag-RDT–negative result, especially among recently symptomatic individuals or those with a high pretest probability of infection (eg, known SARS-CoV-2 exposure) [19–22].

The termination of the public health emergency in the United States on 11 May 2023 had repercussions on insurance coverage for COVID-19 testing, resulting in the elimination of cost-sharing [23]. This development underscores the significance of providing guidance regarding the most effective employment of Ag-RDT and identifying situations that may necessitate confirmatory rRT-PCR testing. CDC's current self-testing guidance recommends individuals test with an at-home antigen test immediately following onset of COVID-19 symptoms and at least 5 days after exposure to someone with COVID-19 [19, 20, 24]. Serial testing is recommended for a total of 2 tests for symptomatic individuals and 3 tests for exposed individuals at 48-hour intervals following an initial negative result. Additionally, the self-testing guidance suggests benefits of testing in the absence of symptoms or known exposure to inform one's risk of transmitting SARS-CoV-2 to others [20].

Our findings are consistent with the CDC self-testing guidance; however, low sensitivity observed for asymptomatic and recently symptomatic persons poses concerns about Ag-RDT effectiveness in controlling SARS-CoV-2 transmission chains. Current recommendations for serial testing may underdetect some recently symptomatic and asymptomatic SARS-CoV-2 infections. Given that the sensitivity of Ag-RDTs peaked at 4 to 7 days after symptom onset in our study, extending the serial testing beyond 48 hours could be considered to increase detection of SARS-CoV-2 infections that initially test negative. In our study, serial testing sensitivity was significantly higher for Ag-RDTs repeated ≥2 days after an initial negative result (69%), compared to repeat testing before 2 days (34%). Additionally, there was only a 48% probability of detecting SARS-CoV-2 infection at least once out of all Ag-RDTs performed within 1 day among rRT-PCR–positive asymptomatic individuals, and the probability was comparable for all Ag-RDTs performed between days −1 and +7 (44%). Guidance regarding Ag-RDT use in the absence of symptoms or known SARS-CoV-2 exposure should be prefaced with information clarifying their reduced reliability in these groups.

Overall Ag-RDT sensitivity in our study did not meet the World Health Organization's minimum point-of-care performance criterion of 80% and was comparatively lower than prior studies conducted during Omicron emergence, which reported overall sensitivities of approximately 80% and better performance in symptomatic individuals [25, 26]. Potential reasons for lower Ag-RDT performance in our study may include variation in test brands and lower viral loads among our highly vaccinated study population. Higher estimated sensitivities of 70% and 75% reported in meta-analyses of studies conducted before 2022 may be partially due to the shedding dynamics of the Omicron lineages circulating during the study period and host factors such as COVID-19 vaccination and prior SARS-CoV-2 infection that contribute to reduced viral load in subsequent infection [27–30]. Reported patterns of pre-Omicron shedding dynamics suggest high transmissibility between 2 days before and 5 days after symptom onset, which initially led public health agencies to recommend a 10-day isolation period after the onset of symptoms [31]. However, recent studies that reported a 3-day incubation period for Omicron and serial intervals of 2 and 3 days for BA.1 and BA.2, respectively, compared to a 4-day serial interval and incubation period for Delta, show that emerging VOCs may exhibit distinct epidemiological characteristics [10, 32, 33].

Similar to this analysis, several other studies have reported an association between lower Ct values and Ag-RDT concordance [29, 34]. However, a low viral RNA copy number or high Ct may indicate a waning or escalating viral load trajectory, and this cannot be determined from a single measure. Diminished viral load may be observed during the early phase of SARS-CoV-2 infection prior to peak infectiousness and thus, detection of cases with higher Ct values and lower viral loads is an important component of preventing transmission to susceptible persons [29]. Several individual-level factors may contribute to variations in infectious virus shedding, such as heterogeneity in the neutralizing antibody response and viral genome load dynamics [35]. As more individuals acquire hybrid immunity from COVID-19 vaccination and 1 or more SARS-CoV-2 infections, more effective control of viral replication by the host immune system may explain the increased frequency of asymptomatic or mildly symptomatic infections [29]. However, the potential for asymptomatic or paucisymptomatic transmission of SARS-CoV-2 to vulnerable populations in congregate settings, such as skilled nursing facilities, remains a public health risk. Thus, it may still be reasonable to recommend molecular rather than rapid antigen testing among certain groups for the protection of vulnerable persons. This may include continuing PCR-based testing for individuals in high-risk occupational settings, including among health care workers who may have been exposed to SARS-CoV-2 but exhibit minimal or no symptoms, and persons at risk of severe illness.

The study has limitations. This university-based study population was disproportionately vaccinated and aged 18 to 24 years, but underrepresentative of men and children compared to the general US population. The sample of participants differed from the full study cohort in certain sociodemographic characteristics; most notably, fewer students (54% vs 67% in the full cohort) and Greek sorority and fraternity members (4% vs 24% in the full cohort) were represented [11]. Participants who voluntarily self-reported test results were neither representative of the university overall nor the full cohort. While participants and individuals from the full cohort who did not report Ag-RDTs were sociodemographically comparable, they may differ by other characteristics related to Ag-RDT performance such as symptom severity (Supplementary Table 5). Misclassification of unsupervised, self-reported symptoms, vaccine status, and Ag-RDT results may have occurred. However, prior studies have demonstrated Ag-RDT self-collection errors are not negatively associated with diagnostic accuracy and yield comparable sensitivities with professionally administered tests [6, 36]. Although rRT-PCR samples were self-collected in our study, our group has previously demonstrated comparable quality of self-collected rRT-PCR samples to clinician-collected samples for detection of SARS-CoV-2 [37, 38]. Lastly, the generalizability of our findings may be limited to the Omicron SARS-CoV-2 lineages circulating at the time of investigation.

In conclusion, in this longitudinal study of over 5000 individuals on a university campus, Ag-RDT performance with rRT-PCR varied by symptom status, time from symptom onset, and Orf1b Ct for detection of SARS-CoV-2 Omicron. Our findings support recommendations for repeat rapid antigen testing following an initial negative result among symptomatic individuals, until at least 4 days after illness onset, and highlight the importance of reevaluating rapid antigen diagnostic performance with the emergence of VOCs.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. We thank the study participants, the University of Washington (UW) Environmental Health and Safety COVID-19 Prevention and Response team (including, Katia Harb and Julie Skene), UW Administration and Incident Command Leadership (including Geoff Gottelib, Margaret Shepherd, Josh Gana, Pamela Schreiber, and Jack Martin), Chu Laboratory, Brotman Baty Institute, and Husky Coronavirus Testing study research assistants.

Disclaimer. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Financial support. This work was supported by the United States Senate and House of Representative Bill 748, Coronavirus Aid, Relief, and Economic Security Act (to H. Y. C. and A. W.).

References

Author notes