Fig. 1

Timeline of the 59 COVID-19 cases treated in the intensive care unit (ICU). a For each case are plotted admission and discharge dates, periods of ventilation, and vital status at discharge. b Cumulative numbers of admissions and discharges according to time, with discharged patients’ lengths of stay (LOS). c Evolution of the estimates of ICU average LOS issued from the two methods of estimation according to the date chosen for estimation. The expected estimate is shown together with the corresponding 95% confidence interval. CPE, method including censored cases; DPE, method considering only stays for which the patient was already discharged from ICU at the date of estimation. Whenever some patients of the cohort remain treated in the ICU at the date of follow-up, c indicates that DPE yields a biased underestimation of ALOS: discharges observed early are more likely to concern patients with a short LOS or conversely, the discharges occurring at the end of the process are more likely to concern patients with a long LOS. b illustrates the latter pattern: nine out of the 10 first occurring discharges concern LOS < 15 days, while eight out of the 10 last occurring discharges concerned LOS > 30 days