Development of a rapid test to screen for smell loss: Validation in a large COVID-19 testing sample

ABSTRACT

Aims: Commercial smell tests are too costly and time-consuming for population surveillance in health emergencies like COVID-19 where smell loss is a concern. To address this need, Parma et al. (2021) developed SCENTinel, a brief test of three olfactory functions detection, intensity, and identification. We examined the psychometric properties of SCENTinel, hypothesizing that SCENTinel components would be positively intercorrelated and that SCENTinel would be associated with self-reported smell loss and SARS-CoV-2+ results. Method(s) We performed a cross-sectional study of a sample of adults presenting for outpatient SARS-CoV-2 testing at Northwestern Medicine locations. Staff handed out SCENTinel test cards to interested patients who voluntarily self-administered the test at home, completing questions on demographics and COVID-19 symptomatology. After smelling each of three resealable odor patches, participants selected the location of the odor (detection), rated the intensity of the odor, and identified the odor via a 4-alternative-forced-choice task (identification). Data were matched with medical record demographics and SARS-CoV- 2(PCR) results. Result(s) Participants completed 2413 SCENTinel tests;after matching participants to medical record data containing concurrent SARS-CoV-2(PCR) results, 1557 tests (64%) were analyzed. Mean age was 49 (SD = 16), 62% were female, 76% were white. 4-5% tested positive for SARS-CoV-2 infection (ndelta = 50;ndelta + omicron = 74). The SCENTinel components showed moderate-to-high correlations with the overall score, r = 0.35-0.84. Identification and detection were moderately correlated (r = 0.42);intensity was slightly correlated with both (rs = 0.13, 0.08), averaging to Cronbach's alpha 0.44 for SCENTinel-overall. Self-reported smell loss was only related to intensity (r = - 0.11). SARS-CoV-2delta+ was slightly related to SCENTinel-overall (r = - 0.09), entirely driven by intensity (r = - 0.27). Mean intensity scores were much lower for the SARSCoV- 2delta+ group (Cohen's d = - 0.76;Fig. 1). Regressions showed that self-reported smell loss and SCENTinel-overall were both uniquely predictive of SARS-CoV-2delta+. As a screener of SARS-CoV- 2delta+, SCENTinel-overall was highly specific (89%), but not sensitive (28%), with kappa of 0.74 (CI 0.70-0.77). Conclusion(s) SCENTinel components are intercorrelated, but effects were small to moderate. Intensity formed a pattern of associations distinct from detection and identification. SCENTinel-overall was highly specific and predictive of SARS-CoV-2delta+, even controlling for self-reported smell loss. The low sensitivity of SCENTinel could reflect asymptomatic infection. Future directions include optimizing scoring and validation with goldstandard olfaction tests.