Diagnostic Value of Patient-Reported and Clinically Tested Olfactory Dysfunction in a Population Screened for COVID-19

Key Points

Question  What are the sensitivities and specificities of patient-reported olfactory dysfunction and gustatory dysfunction and of a newly developed simple and fast clinical test for assessing olfactory dysfunction (Clinical Olfactory Dysfunction Assessment) for diagnosing coronavirus disease 2019 (COVID-19) in outpatients with mild or no symptoms?

Findings  In this diagnostic study including 809 participants evaluated prior to reverse transcriptase–polymerase chain reaction testing for severe acute respiratory syndrome coronavirus 2, 35% of 58 participants with confirmed COVID-19 reported olfactory dysfunction, gustatory dysfunction, or both; this compared with only 4% of 751 without COVID-19. Olfactory anamnesis and clinical testing results were complementary, yielding similar, strong diagnostic values for COVID-19.

Meaning  Findings from this study indicate that, in a screening setting of outpatients with no or mild to moderate symptoms of COVID-19, olfactory dysfunction and gustatory dysfunction should be systematically assessed by anamnesis and by clinical testing because their results appear to have strong diagnostic value for COVID-19.

Importance  Recent studies have suggested that olfactory dysfunction and gustatory dysfunction are associated with coronavirus disease 2019 (COVID-19). However, olfaction has been evaluated solely on reported symptoms, after COVID-19 diagnosis, and in both mild and severe COVID-19 cases, but rarely has it been assessed in prospectively unselected populations.

Objective  To evaluate the diagnostic value of a semiobjective olfactory test developed to assess patient-reported chemosensory dysfunction prior to testing for the presence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in patients attending a COVID-19 screening facility.

Design, Setting, and Participants  This prospective diagnostic study with participants and observers blinded to COVID-19 status was conducted in a COVID-19 screening center of a tertiary university hospital in France from March 23 to April 22, 2020. Participants were 854 consecutively included health care workers or outpatients with symptoms or with close contact with an index case. Exclusion criteria were prior chemosensory dysfunction, testing inability, or contraindications (n = 45).

Main Outcomes and Measures  Participants were interviewed to ascertain their symptoms and then underwent Clinical Olfactory Dysfunction Assessment (CODA), an ad hoc test developed for a simple and fast evaluation of olfactory function. This assessment followed a standardized procedure in which participants identified and rated the intensity of 3 scents (lavender, lemongrass, and mint) to achieve a summed score ranging from 0 to 6. The COVID-19 status was assessed using reverse transcriptase–polymerase chain reaction to detect the presence of SARS-CoV-2 in samples collected via nasopharyngeal swab (reference standard) to calculate the diagnostic values of patient-reported chemosensory dysfunction and CODA.

Results  Of 809 participants, the female to male sex ratio was 2.8, and the mean (SD) age was 41.8 (13.0) years (range, 18-94 years). All participants, if symptomatic, had mild disease at the time of testing, and 58 (7.2%) tested positive for SARS-CoV-2. Chemosensory dysfunction was reported by 20 of 58 participants (34.5%) with confirmed COVID-19 vs 29 of 751 participants (3.9%) who tested negative for COVID-19 (absolute difference, 30.6% [95% CI, 18.3%-42.9%]). Olfactory dysfunction, either self-reported or clinically ascertained (CODA score ≤3), yielded similar sensitivity (0.31 [95% CI, 0.20-0.45] vs 0.34 [95% CI, 0.22-0.48]) and specificity (0.97 [95% CI, 0.96-0.98) vs 0.98 [95% CI, 0.96-0.99]) for COVID-19 diagnosis. Concordance was high between reported and clinically tested olfactory dysfunction, with a Gwet AC1 of 0.95 (95% CI, 0.93-0.97). Of 19 participants, 15 (78.9%) with both reported olfactory dysfunction and a CODA score of 3 or lower were confirmed to have COVID-19. The CODA score also revealed 5 of 19 participants (26.3%) with confirmed COVID-19 who had previously unperceived olfactory dysfunction.

Conclusions and Relevance  In this prospective diagnostic study of outpatients with asymptomatic or mild to moderate COVID-19, systematically assessed anamnesis and clinical testing with the newly developed CODA were complementary and specific for chemosensory dysfunction. Olfactory dysfunction was suggestive of COVID-19, particularly when clinical testing confirmed anamnesis. However, normal olfaction was most common among patients with COVID-19.

The present outbreak of coronavirus disease 2019 (COVID-19) and the high rate of morbidity and mortality in aged and comorbid populations justified the first lockdown by the French authorities from March 17 to May 11, 2020. Incidence of COVID-19 hospitalizations decreased nationally starting March 31, 2020. Health authorities have now imposed regulations of reverse transcriptase–polymerase chain reaction (RT-PCR) testing of individuals who were exposed to the virus or who have COVID-19 symptoms, with isolation of new clusters. Identifying COVID-19 cases in health care settings is also paramount to reduce nosocomial transmission.^1,2

Although commonly reported clinical features associated with COVID-19 include fever, cough, myalgia, and fatigue,^3-5 there are increasing reports of other symptoms and apparent asymptomatic cases.^2,6-8 The American Academy of Otolaryngology–Head and Neck Surgery, the British Rhinological Society, and the French Society of ENT promptly alerted the general public and medical community to increasing cases of acute olfactory dysfunctions (ODs) and gustatory dysfunctions (GDs) in patients with COVID-19,^9-11 with disparate prevalence ranging from 33.9% to 88.0%.^12-23 By contrast, studies from East Asia have reported lower prevalence ranging from 6.3% to 15.7%.^24,25

In all of these studies, COVID-19 status was systematically established prior to smell and taste assessment, potentially inducing selection and measurement biases. Most studies also relied on patient-reported symptoms,^{12,14,15,19-25} which are prone to underdiagnosis if chemosensory loss is mild or to overdiagnosis with chemosensory simulation. Only a few studies have evaluated ODs or GDs with semiobjective tests, but these tests are cumbersome in a pandemic setting, difficult to perform by nonspecialists, and have not been evaluated to provide a threshold of OD suggestive of COVID-19.^13,16-18 To address these caveats, we developed the simple and fast semiobjective olfactory test Clinical Olfactory Dysfunction Assessment (CODA) for use during the COVID-19 pandemic.

The objective of the present study was to assess the diagnostic value of CODA in assessing patient-reported ODs and GDs prior to testing for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in patients attending a COVID-19 screening facility.

This study, a prospective single-center diagnostic study registered at ClinicalTrials.gov,²⁶ followed the Standards for Reporting of Diagnostic Accuracy (STARD) reporting guideline. This study, which received no funding, was approved by the institutional review board at the University Hospital of Montpellier, France. All participants provided written informed consent in a manner that was consistent with French regulation (Loi Jardé). No one received compensation or was offered any incentive for participating in this study.

Participants were health care workers and adult outpatients attending a COVID-19 screening facility of a tertiary referral center (University Hospital of Montpellier, France), all consecutively recruited between March 23 and April 22, 2020. Patients were referred to this facility if they had symptoms compatible with COVID-19 or if they had been closely exposed to a confirmed COVID-19 index case (whether or not the referred patient was symptomatic) and were included after providing informed consent. Patients requiring hospitalization were admitted via an independent emergency ward and were not tested in the screening center. Persons with a known history of smell and taste dysfunction, testing inability (cognitive impairment, nonnative French speakers), or olfactory testing contraindication (pregnancy, breastfeeding) were excluded.

Patient-reported chemosensory dysfunction was recorded. Data collection of symptoms was performed in a binary manner (yes or no) and did not specifically target olfactory and gustatory impairment to avoid measurement bias. Olfactory dysfunction was defined by pure olfactory loss (characterized by hyposmia and anosmia). For gustatory dysfunction, taste-GD was defined by the pure gustatory loss of sweet, salty, bitter, sour, and umami tastes; and flavor-GD corresponded to the olfactory component of taste (eg, coffee or strawberry flavors).²⁷

To easily test olfaction, we designed the semiobjective CODA, which we adapted from the quick 3-item versions of the University of Pennsylvania Smell Identification Test²⁸ and the Sniffin’ Sticks test.^29,30 These 2 tests are relatively costly, not readily available, or not disposable and thus are not adapted to the mass screening and hygiene requirements during the COVID-19 outbreak. We applied CODA identically to all participants in the study by following a standardized procedure. For the assessment, participants used both nostrils to sniff 3 different scents located on the extremity of paper strips previously dipped in 5 μL of essential oil. The scented strips were presented separately, one strip after the other, with 10 seconds between presentations. Participants were asked to identify each scent (scoring 0 if not recognized or 1 if recognized) in an open choice, without clues or a list, and to rate the intensity of the scent on a 4-item Likert scale (0, no sensation; 1, weak; 2, moderate; and 3, strong). The identification score was the sum of each correctly recognized scent, and the intensity score was the mean of the intensity rating that was given to each scent; thus, each score ranged from 0 to 3. These 2 subscores were then summed to obtain the CODA total score, which ranged from 0 to 6. This score could have decimal values that were rounded to the closest integer to simplify calculations. We combined qualitative and quantitative components of olfaction in the CODA total score to limit misclassifications because scents can be perceived but not correctly named due to variability in olfactory awareness and education. We chose lavender, lemongrass, and mint because these scents are widely known in our area. Ammonia (13% solution) was used as a positive control to validate the reliability of the CODA score and to detect malingerers because the perception of ammonia is conveyed by the trigeminal nerve.³¹ Participants who reported experiencing no sensation from the ammonia were excluded from the study.

The RT-PCR test for SARS-CoV-2 performed using a sample collected from a nasopharyngeal swab was the reference standard for diagnosis of COVID-19. In a recent meta-analysis, this test showed sensitivity of nearly 90%.³² Antibody detection has shown limitations in the diagnosis of COVID-19 in the early phase of infection, with most seroconversions occurring by the end of the second week of symptoms.^33,34 Thus, we used serologic tests solely for COVID-19 post hoc diagnosis in participants with OD but whose RT-PCR test results were negative for SARS-CoV-2. Lung computed tomography scans, which have shown high sensitivity,³² were not performed because they are not recommended in outpatients.

The extraction of SARS-CoV-2 RNA from the nasopharyngeal swab (Sigma Virocult) was conducted using the QIAamp Viral RNA Mini Kit on the QIAsymphony platform (Qiagen), following the manufacturers’ instructions. The SARS-CoV-2 RNA was assessed using RT-PCR targeting the RNA-dependent RNA polymerase as previously described.³⁵

Patients attending the COVID-19 screening center were asked about their exposure risks to SARS-CoV-2, tobacco use, active pollen allergies, and COVID-19–like manifestations, such as fever, cough, myalgia, fatigue, OD, flavor-GD, and taste-GD, prior to the administration of CODA. All clinical assessments were conducted by the same trained physicians (including C.V.). Nasopharyngeal swabs were performed by experienced health professionals after CODA administration. Thus, investigators and participants were blinded to COVID-19 status. The whole procedure took less than 10 minutes, including 2 minutes for administering CODA.

In cases of patient-reported or clinically tested OD with negative results for SARS-CoV-2 using the RT-PCR test, we performed a SARS-CoV-2 serologic test measuring IgG antibodies directed against the nucleocapsid protein with an enzyme-linked immunosorbent assay CE-IVD–marked ELISA kit (ID screen SARS-CoV-2-N IgG indirect ELISA; ID.Vet).³⁶ Blood samples were collected more than 15 days after either the initial reported symptom or collection of the swab sample when the participant was asymptomatic.

The characteristics of the participants were described using means with SDs or medians with interquartile ranges for continuous variables and using numbers with percentages for categorical variables. Patient-reported OD alone; OD or GD and both combined; and CODA were separately evaluated as index tests vs the presence of SARS-CoV-2 on nasopharyngeal swab samples using RT-PCR as the reference test, with calculations of sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (PLR), and negative likelihood ratio (NLR). Various thresholds of the CODA total score were analyzed using a receiver operating characteristic curve and calculation of the Youden index.³⁷ Both analyses assess diagnostic marker effectiveness and enable the selection of an optimal threshold value (cut point). Diagnostic values of the index tests were compared using the McNemar test or the Fisher exact test. Concordance between patient-reported OD and the CODA score was calculated using the Gwet AC1 coefficient, which provides a more stable interrater reliability coefficient than Cohen κ because the former is less affected by prevalence and remains valid in the event of a substantial imbalance in the marginal totals of the table.³⁸

According to preliminary analysis results of the first 90 participants, the prevalence of confirmed COVID-19 cases was 13% (12 participants). To obtain sensitivity and specificity estimates with an accuracy of 10%, we had to include 724 patients in the study. The number of included patients was subsequently increased because the incidence of new COVID-19 cases decreased further with time. All tests were 2-sided, and the statistical significance threshold was set at 5%. Analyses were performed using SAS Enterprise Guide, version 7.3 (SAS Institute Inc).

Between March 23 and April 22, 2020, 854 patients were referred to the COVID-19 screening center. Forty-five patients were not included because of prior OD or GD, testing inability, olfactory testing contraindication, or breach in the protocol (Figure 1). Thus, 809 participants (94.7%) were enrolled, with 755 health care workers among them. In total, 463 participants were referred for symptoms compatible with COVID-19, and 340 participants were asymptomatic. Six participants were referred to the screening center for unknown reasons.

Demographic characteristics and clinical presentations of the analyzed population are given in Table 1. The female to male sex ratio was 2.8. The mean (SD) age was 41.8 (13.0) years (range, 18-94 years). All symptomatic participants had mild disease at the time of testing. Of 809 participants, 38 (4.7%) reported OD, 32 (4.0%) reported flavor-GD, and 13 (1.6%) reported taste-GD. Clinical assessment and nasopharyngeal swabbing were conducted after a median (interquartile range) of 5.0 days (2.0-10.0) after the initial symptom, when present. In total, 58 participants (7.2%) had positive RT-PCR results for SARS-CoV-2.

Of 58 confirmed COVID-19 cases, 14 (24.1%) were asymptomatic. The most common symptoms were an inconsistent combination of cough, fever, headache, myalgia, and OD or GD or both. Olfactory dysfunction or GD or both were reported by 20 of 58 participants (34.5%) with confirmed COVID-19 vs 29 of 751 participants (3.9%) with negative RT-PCR results for SARS-CoV-2, with an absolute difference of 30.6% (95% CI, 18.3%-42.9%). Olfactory dysfunction was reported as the initial symptom in 10 RT-PCR–confirmed COVID-19 cases (17.2%) and flavor-GD in 7 RT-PCR–confirmed COVID-19 cases (12.1%). Olfactory dysfunction with flavor-GD was the sole symptom in 3 participants, with 2 of these 3 having positive RT-PCR results for SARS-CoV-2.

The diagnostic values of symptoms associated with COVID-19 are given in Table 2. The patient-reported OD sensitivity was 0.31 (95% CI, 0.20-0.45), specificity was 0.97 (95% CI, 0.96-0.98), PPV was 0.47 (95% CI, 0.31-0.64), NPV was 0.95 (95% CI, 0.93-0.96), PLR was 11.65, and NLR was 0.71. Compared with the sensitivity of OD, the sensitivity of flavor-GD (0.26; 95% CI, 0.15-0.39) and of taste-GD (0.07; 95% CI, 0.02-0.17) were lower, but these dysfunctions also showed excellent specificity (flavor-GD, 0.98 [95% CI, 0.96-0.99]; taste-GD, 0.99 [95% CI, 0.98-0.99]). The specificity of chemosensory symptoms was higher than any other reported symptom. Combining OD or GD and both into a single group did not decisively modify sensitivity (0.34; 95% CI, 0.22-0.48) or specificity (0.96; 95% CI, 0.95-0.97) vs OD alone.

Diagnostic performances for each cut point of the CODA total score are given in Table 2, and the number (percentages) of participants included within each set of ranges for the CODA total score are given in Table 3. The area under the receiver operating characteristic curve was 0.68 (95% CI, 0.60-0.76) (Figure 2), and the highest Youden index of 0.34 was reached for CODA scores of 4 or lower. A CODA score of 4 or lower showed a sensitivity of 0.47 (95% CI, 0.33-0.60), specificity of 0.87 (95% CI, 0.85-0.90), PPV of 0.22 (95% CI, 0.15-0.31), NPV of 0.95 (95% CI, 0.94-0.97), PLR of 3.72, and NLR of 0.61. A CODA score of 3 or lower showed a sensitivity of 0.34 (95% CI, 0.22-0.48), specificity of 0.98 (95% CI, 0.96-0.99), PPV of 0.53 (95% CI, 0.36-0.69), NPV of 0.95 (95% CI, 0.93-0.96), PLR of 14.39, and NLR of 0.67. Because patient-reported OD appeared to be a highly specific symptom, we focused on the CODA specificity to enable diagnostic confirmation, comparison, and concordance analyses. Thus, we chose the cutoff value of 3 instead of 4 because it shared a similar Youden index (0.32 vs 0.34) but showed far better specificity, PPV, and PLR. Indeed, the increase of false-positive cases (18 with CODA scores ≤3 vs 94 with CODA scores ≤4) was disproportionate compared with the increase of true-positive cases (20 with CODA scores ≤3 vs 27 with CODA scores ≤4). Of the confirmed COVID-19 cases, 11 participants (19.0%) presented with a CODA score of 0, 20 participants (34.5%) with CODA scores of 3 or lower, and 38 participants (65.5%) with CODA scores higher than 3.

Concordance was high between patient-reported OD and CODA total score at score cutoffs of 3 (AC1, 0.95; 95% CI, 0.93-0.97) and 4 (AC1, 0.84; 95% CI, 0.80-0.87). Concerning OD or GD plus both combined into a single group, concordance with the CODA score was slightly lower, with an AC1 of 0.94 (95% CI, 0.92-0.95) for CODA scores of 3 or lower and an AC1 of 0.83 (95% CI, 0.80-0.86) for CODA scores of 4 or lower.

Of participants having both patient-reported OD and CODA total scores of 3 or less, 15 of 19 (78.9%) were RT-PCR–confirmed COVID-19 cases vs 8 of 38 confirmed COVID-19 cases (21.1%) in participants with a discordant olfactory evaluation (Figure 1). Among participants reporting OD but having CODA total scores higher than 3, 16 of 19 (84.2%) had negative RT-PCR results for SARS-CoV-2. Inversely, among normosmic participants with CODA total scores of 3 or lower, 5 of 19 (26.3%) had positive RT-PCR results for SARS-CoV-2.

We collected blood samples from participants after a mean (SD) of 48.0 (11.6) days after the initial symptom and 35.7 (10.9) days after collection of nasopharyngeal swab. Of 19 participants, 18 (94.7%) with both patient-reported OD and CODA total scores of 3 or lower had COVID-19, 15 confirmed by RT-PCR analysis and 3 confirmed by serologic test results (Figure 1). No seroconversion occurred in 26 participants with RT-PCR results negative for SARS-CoV-2 and showing discrepancies between patient-reported olfactory function and CODA total score.

We prospectively evaluated 809 patients attending an outpatient screening facility to determine the diagnostic value of patient-reported OD and a clinical OD test. There were 58 (7.2%) confirmed COVID-19 cases as assessed by RT-PCR, of which 31.0% had patient-reported OD and 34.5% had clinically tested OD (CODA scores ≤3). Of participants with negative RT-PCR results, 97.3% did not report OD and 97.6% had CODA total scores higher than 3. Of 38 participants who reported OD, 47.4% tested positive for SARS-CoV-2; of 38 participants with CODA total scores of 3 or lower, 52.6% were COVID-19–confirmed cases.

Of patients with confirmed COVID-19, 34.5% reported chemosensory dysfunction, similar to other studies.^12,16,39 However, OD and GD prevalence varies greatly between reports, ranging from 6.3% to 88.0%,^12-25 as underscored in a recent meta-analysis.⁴⁰ Most of those studies were performed on a convenience sample of patients already diagnosed as having COVID-19, thus preventing the calculation of the diagnostic value of chemosensory dysfunction and potentially being exposed to selection and measurement biases. The only exception was for the study by Bénézit et al,³⁹ who reported sensitivity and specificity of 0.44 and 0.92, respectively, for patient-reported OD and 0.42 and 0.95, respectively, for combined OD and GD, in line with our results but using exclusively self-reported data.

Some studies have used semiobjective olfactory tests, such as the University of Pennsylvania Smell Identification Test,¹⁶ the identification portion of the Sniffin’ Sticks test,¹³ the Connecticut Chemosensory Clinical Research Center test, or other nonstandardized ad hoc tests.^17,18 All of these studies reported a high prevalence, between 62% and 98%, of ODs in patients with COVID-19. Once again, olfactory testing was performed after confirming a COVID-19 diagnosis and in higher proportions of symptomatic inpatients. Furthermore, olfaction was assessed at various times after COVID-19 diagnosis, likely explaining their disparate frequency of anosmia, ranging from 3% to 48%, thus preventing comparison between results.^13,16,17 In the present study, CODA was performed during the early phase of COVID-19, and we found 19.0% of participants with complete anosmia (CODA total score of 0).

The previously designed semiobjective tests are cumbersome and time-consuming. Although they aim to graduate olfactory dysfunction and enable the identification of minor and unperceived impairment, mild OD is not necessarily associated with COVID-19; thus, the use of those tests may lead to overestimation of SARS-CoV-2 infection. Moreover, those tests do not provide a threshold of OD severity to differentiate individuals with COVID-19 from healthy persons, and the tests appear more suitable for OD recovery assessment. Adapted from validated tests,^28-30,41 CODA enabled the establishment of a cutoff score for detecting COVID-19. The test was designed to be fast, hygienic (single use), and cost-effective, and thus it would be simple to implement in a large-scale screening setting and easily adapted to routine medical care during the current outbreak.

The CODA results appeared to be clinically relevant for detecting ODs associated with COVID-19. Indeed, such a semiobjective olfactory test is useful because discrepancies can occur between anamnesis and testing.^13,16-18 Some OD detected only by either reporting or clinical testing may potentially be explained by low olfactory awareness or education, malingering, real differences in performance of OD assessment, variability in OD duration, or possible OD recovery before olfactory testing. In the present study, the presence of both patient-reported and clinically tested OD was associated with COVID-19 in nearly all cases: 78.9% of participants reporting OD and having CODA scores of 3 or lower also had confirmed COVID-19 as assessed by RT-PCR (94.7% when also including serologic results). Moreover, CODA enabled the detection of 5 participants with a self-reported normal sense of smell but with positive RT-PCR results. Thus, CODA could be used as a simple diagnostic complement to anamnesis, enabling the confirmation of an olfactory impairment, and as a screen for unperceived olfactory impairment. However, the use of CODA may be limited in the elderly and in persons with psychiatric or cognitive disabilities—as would any psychophysical olfactory test—because the assessment depends on patients’ active participation and cognitive resources.

We found that flavor-GD and taste-GD were more frequent in COVID-19–confirmed cases than in participants without confirmed COVID-19. Although flavor-GD is related to OD, taste-GD was relatively infrequent in our COVID-19 population. Combining OD and GD did not decisively modify diagnostic values. Misinterpretation between flavors and basic tastes could explain these results. Gustatory identification tests using tastants on cotton swabs or paper strips have been described;^17,18,41 but even for basic tastes, misinterpretations can occur between sour, bitter, and salty.⁴¹ During the COVID-19 outbreak, we believe that OD is more relevant than GD to assess because OD was more frequent and easier to test.

The strengths of our study lie in its prospective design and the consecutive inclusion of patients, all attending a SARS-CoV-2 screening facility as outpatients as well as the realization of a systematic, patient-reported evaluation and a standardized clinical olfactory assessment prior to nasopharyngeal swabbing. We believe that our results are reliable because investigators and patients were blinded to COVID-19 status. Moreover, the design of our study enabled the limiting of biases in calculating the prevalence of ODs in patients with COVID-19 in an ambulatory setting. Indeed, we included people attending the screening center whether or not they were symptomatic or experiencing other differential diagnoses of OD, such as allergic and viral rhinitis. It has been previously shown that, contrary to other viruses causative of ODs, SARS-CoV-2–associated OD is not systematically associated with rhinitis symptoms.^42-44 As described elsewhere,^15,21 we found that ODs occurred as the initial sign of COVID-19 in 17.2% of cases, which suggests that they may be a warning sign of COVID-19. Hypotheses for the appearance of ODs as the initial sign include sensorineural dysfunction^14,17,24,45 or a localized inflammation of the olfactory clefts.⁴⁶

However, our study has some limitations. Although the use of CODA confirmed COVID-19–associated ODs with high specificity, the assessment was not designed to establish a precise diagnosis of ODs as complete semiobjective olfactory tests do, and CODA cannot be used for follow-up of patients. Furthermore, the reproducibility of CODA performance must be validated by a test-retest procedure. Finally, similar to every olfactory test, our choice of odors used in CODA limits its generalizability owing to cultural and educational factors. Thus, different versions of CODA using geographically adapted scents should be evaluated in further studies. The use of RT-PCR has been criticized for its low sensitivity in detecting SARS-CoV-2, but a recent meta-analysis has shown close to 90% sensitivity of RT-PCR for samples collected with nasopharyngeal swabs when standardized procedures are respected.³² In the present study, swabbing was routinely performed by trained health professionals who were dedicated to COVID-19 screening; thus, we believe our conclusions remain valid. Among participants with negative SARS-CoV-2 RT-PCR results, serologic tests were also negative for COVID-19 in all 26 participants with discordant olfactory evaluation but were positive for COVID-19 in 3 of 4 participants who self-reported OD and scored 3 or lower on CODA. Nevertheless, serologic tests were performed in a limited number of targeted participants and must be considered only a descriptive tool to interpret the negative RT-PCR results in our study.

In conclusion, this prospective study of patients attending a screening facility showed that OD or GD or both were commonly encountered in outpatients with COVID-19, albeit at a lower rate than that reported in studies with patients who had previously received a diagnosis of COVID-19. The presence of either patient-reported or clinically tested OD should alert clinicians to a higher probability of COVID-19, whereas the presence of both patient-reported and clinically tested OD appeared to be highly predictive in our study. Chemosensory dysfunction should be systematically assessed by anamnesis and clinical testing because they were highly specific and complementary. However, most confirmed COVID-19 cases occurred in patients without OD, and a diagnostic test with RT-PCR remains mandatory.

Accepted for Publication: November 9, 2020.

Published Online: January 7, 2021. doi:10.1001/jamaoto.2020.5074

Corresponding Author: Charles Villerabel, MD, Department of Otorhinolaryngology–Head and Neck Surgery and Maxillofacial Surgery–Hospital and University of Montpellier, Département ORL & CMF–CHU de Montpellier, Hôpital Gui de Chauliac, 80 Avenue Augustin Fliche, 34295 Montpellier, France (villerabel.charles@gmail.com).

Author Contributions: Drs Villerabel and Makinson contributed equally to this work and are considered co–first authors. Ms Jaussent and Dr Picot had full access to all data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Villerabel, Makinson, Picot, Nègre-Pagès, Crampette, Reynes, Le Moing, Venail.

Acquisition, analysis, or interpretation of data: Villerabel, Makinson, Jaussent, Nègre-Pagès, Rouvière, Favier, Morquin, Reynes, Tuaillon, Venail.

Drafting of the manuscript: Villerabel, Makinson, Jaussent, Rouvière, Crampette, Venail.

Critical revision of the manuscript for important intellectual content: Makinson, Picot, Nègre-Pagès, Favier, Morquin, Reynes, Le Moing, Tuaillon.

Statistical analysis: Jaussent, Picot, Nègre-Pagès.

Administrative, technical, or material support: Makinson, Rouvière, Favier, Tuaillon, Venail.

Supervision: Makinson, Picot, Crampette, Reynes, Le Moing, Venail.

Conflict of Interest Disclosures: None reported.

Additional Contributions: We thank the French professional network of otorhinolaryngologists Doc@Doc, which has sensitized the French scientific community to chemosensory dysfunctions as important clinical features of coronavirus disease 2019 (COVID-19). Luc Bourdiol, MD, and Guillemette Combes, MD, both from the Clinique du Parc, assisted in data collection; Mathilde Raverdeau and Maelle Dereure, both from the University Hospital of Montpellier, assisted in data management; Emilie Barde, Thomas Landragin, and Sandrine Mas, also all from the University Hospital of Montpellier, provided administrative support. No one was compensated for the stated contribution.