RESULTS
The data extraction process yielded a dataset of 132,446 unique patients who had a diagnosis of COVID-19 either on admission or during their stay. Of these 13,401 (10.1%) were admitted to critical care as part of a hospital stay. Of those admitted to critical care, 7,993 (59.6%) had advanced respiratory support and 2,200 (16.4%) had a tracheostomy procedure recorded.
Tracheostomy use in critical care patients changed markedly over time. Use of tracheostomy and associated mortality rates for month of discharge are presented in Figure 1, tracheostomy use for each region over time by month of discharge are presented in Figure 2. By discharge date, tracheostomy rates were very low for patients discharged in March (2.6%), peaked for discharges in June (36.7%) and declined thereafter to 4.1% in October. This trend was seen across all regions. Mortality rates in those with a tracheostomy revealed a mirror image trend, with the lowest death rates seen at times of highest tracheostomy use. When plotted by admission month the pattern is similar, but with an earlier and smaller peak in the proportion of patients with a tracheostomy (11.9% in February, 20.5% in April). However, there was a similar decline in the proportionate use of tracheostomy in late summer 2020 (3.8% in August).
The profile of those admitted to hospital, those admitted to critical care and those who had advanced respiratory support and a tracheostomy is summarised in Table 1 . Those admitted to critical care were more likely to be aged 40-69 years and less likely to be aged 70 years and over than the general hospitalised population. They were also more likely to be male and from a non-White ethnic background. The deprivation profile of those admitted to critical care reflected the wider population. Obese patients were over-represented in those admitted to critical care and patients with dementia, cardiovascular disease, renal disease and cancer were under-represented. The profile of those who were recorded as having a tracheostomy was similar to the wider critical care population, although there was a smaller percentage of people aged 70 years and over with a tracheostomy.
Factors associated with having a tracheostomy for those admitted to critical care were explored using multilevel logistic regression and the results are presented in Table 2 . Compared to the 18-39 years age group, tracheostomy was significantly more common in the 40-79 years age group and significantly less common in the 80 years and over age group. Tracheostomy was significantly more common in males, in Asian and Black ethnic groups and in patients with cerebrovascular disease. Tracheostomy was less common in patients with peripheral vascular disease, chronic heart failure, acute myocardial infarction, connective tissue/rheumatic disease, moderate/severe liver disease, renal disease and cancer.
In patients admitted to critical care, outcomes for those with and without a tracheostomy are presented in Table 3 with the adjusted association of tracheostomy with each outcome. Tracheostomy was significantly associated with reduced odds of in-hospital mortality and increased odds of length of stay greater than the median after adjusting for covariates.
The time from hospital admission to critical care admission was the same for both those who survived to discharge and those who died during their stay: median 1 day (IQR 0 to 3). Of those with a tracheostomy, 120 (5.5%) had tracheostomy malfunction recorded during their stay. Of those with a malfunction recorded, 36 died in hospital (30.0%) and in the 2080 without a malfunction 437 (21.0%) died in hospital.
Of the 2,200 patients with a tracheostomy, data on time from critical care admission to tracheostomy were available for 1,777 (80.8%) patients. Data on the timing of tracheostomy from critical care admission are presented in Table 4 for those who died and those who survived to discharge. Patients who underwent a tracheostomy at ≤ 14 days from critical care admission and survived to discharge had a shorter hospital and critical care stay both overall and post-tracheostomy. Undergoing a tracheostomy ≤ 14 days from critical care admission was associated with significantly shorter time from tracheostomy to critical care discharge (β = -0.100 (95% CI -0.170 to -0.031) and hospital discharge (β = -0.061 (95% CI -0.115 to -0.007).