Laryngeal symptoms after COVID-19 peaked one to two years later, not in the acute phase

Design

retrospective propensity-matched cohort study in a global electronic health record network

Population

adults with COVID-19 and uninfected controls, no prior laryngeal disorder or major comorbidity; network of over 180 million records

Primary outcome

new-onset chronic cough, dysphagia, voice disorders, vocal fold paralysis and laryngeal spasm up to 5 years

Effect

peak odds ratios 7.12 for chronic cough, 2.71 dysphagia, 3.25 voice disorders, 2.17 vocal fold paralysis, 2.79 laryngeal spasm; absolute risk differences 0.003% to 0.36%

A retrospective cohort drawn from a global electronic health record network of more than 180 million records compared adults with COVID-19 against uninfected controls, excluding anyone with pre-existing laryngeal disorders or major comorbidity, and followed both groups for up to five years after propensity score matching.

All five outcomes were commoner after infection: chronic cough (peak odds ratio 7.12), dysphagia (2.71), voice disorders (3.25), vocal fold paralysis (2.17) and laryngeal spasm (2.79). Incidence peaked at one to two years after infection for most, and at two to three years for laryngeal spasm. Hospitalisation and mechanical ventilation raised the rate of dysphagia (hazard ratios 2.63 and 5.26), voice disorders (1.15 and 4.45) and vocal fold paralysis (2.09 and 9.35) - unsurprising after intubation - but were associated with *lower* rates of chronic cough. Vaccinated patients had slightly higher recorded rates of chronic cough and voice disorders.

Read the absolute numbers alongside the odds ratios, because they change the story. The risk difference for chronic cough was 0.33 percentage points, for dysphagia 0.36, for voice disorders 0.12, for vocal fold paralysis 0.01 and for laryngeal spasm 0.003. An odds ratio of 7 on a very small baseline is still a very small absolute risk. The findings on vaccination and on reduced cough after ventilation both point towards detection bias - people in contact with health services get more diagnoses recorded - which is the standard hazard of this kind of database work. What survives is a timing observation worth carrying into clinic: a patient presenting with new dysphonia or chronic cough a year or two after COVID-19 is describing a recognised pattern, not an implausible one.

• Ask about COVID-19 infection in the preceding two years when assessing new chronic cough, dysphonia or dysphagia
• Do not let the history stop the work-up - a post-COVID label must never substitute for laryngoscopy in new dysphonia
• Expect vocal fold paralysis and dysphagia mainly in patients who were intubated, where the mechanism is mechanical
• Put the absolute risk to a worried patient: these are increases of well under half a percentage point
• Treat the vaccination association as detection bias rather than as a finding about vaccines

Why it matters

It gives a timeline for symptoms that patients are attributing to COVID-19 and clinicians are dismissing.

Don't overread it

A coded electronic health record cohort - these are diagnoses recorded, not diagnoses confirmed, and the effects are tiny in absolute terms.

The statistics, in plain English

This is the clearest example you will see of a large odds ratio on a tiny baseline: the peak odds ratio for laryngeal spasm was 2.79, and the absolute risk difference was 0.003 percentage points - three extra cases per hundred thousand. Propensity matching balances recorded variables only, and the two findings that make no biological sense - less chronic cough after mechanical ventilation, more after vaccination - both point the same way, towards differences in who gets seen and coded rather than who gets ill.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.