Diagnosed and subjectively perceived long-term effects of COVID-19 infection on olfactory function assessed by supervised machine learning

Abstract Background Loss of olfactory function appears to be a typical COVID-19 symptom, at least in early variants of SARS-CoV2. The time that has elapsed since the emergence of COVID-19 now allows us to assess the long-term prognosis of its olfactory impact. Methods Participants (n = 722 of whom n = 464 reported having had COVID-19 dating back with a mode of 174 days) were approached and tested in a museum as a relatively unbiased environment. Olfactory function was diagnosed by assessing odor threshold and odor identification performance. Subjects also rated their actual olfactory function on a 100-mm visual analog scale and provided analogous retrospective estimates of their smelling ability before the COVID-19 infection and immediately after it. Results Diagnosed olfactory function did not differ in former COVID-19 patients from controls. Olfactory diagnoses included 20% decreased olfactory function in former patients and 18.7% in controls, which was not significant. Of former patients, 145 (31.2%) retrospectively reported temporarily reduced olfactory function. Only nine 9 patients (2.2%) reported ongoing reduction of their olfactory function. Overall, former patients rated their current olfactory function significantly better than controls. Supervised machine learning was able to detect past COVID-19 infection from self-assessment of current olfactory function, but not from diagnosed current olfactory function, better than by guessing, although accuracy was intermediate, not suggesting utility as a clinical test. Conclusions While retrospectively about one-third of former CVID-19 patients recalled olfactory symptoms associated with prior infection, the present results suggest a positive long-term prognosis for COVID-19-associated olfactory loss.