Ten months of temporal variation in the clinical journey of hospitalised patients with COVID-19: An observational cohort

  1. ISARIC Clinical Characterisation Group
  2. Matthew D Hall  Is a corresponding author
  3. Joaquín Baruch
  4. Gail Carson
  5. Barbara Wanjiru Citarella
  6. Andrew Dagens
  7. Emmanuelle A Dankwa
  8. Christl A Donnelly
  9. Jake Dunning
  10. Martina Escher
  11. Christiana Kartsonaki
  12. Laura Merson
  13. Mark Pritchard
  14. Jia Wei
  15. Peter W Horby
  16. Amanda Rojek
  17. Piero L Olliaro
  1. Big Data Institute, Nuffield Department of Medicine, University of Oxford, United Kingdom
  2. ISARIC Global Support Centre, Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, United Kingdom
  3. Department of Statistics, University of Oxford, United Kingdom
  4. MRC Centre for Global Infectious Disease Analysis, Abdul Latif Jameel Institute for Disease and Emergency Analytics and Department of Infectious Disease Epidemiology, Imperial College London, United Kingdom
  5. MRC Population Health Research Unit, Clinical Trials Service Unit and Epidemiological Studies Unit, Nuffield Department of Population Health, University of Oxford, United Kingdom
  6. Infectious Diseases Data Observatory, Centre for Tropical Medicine and Global Health, University of Oxford, United Kingdom
  7. Royal Melbourne Hospital, Melbourne, Australia Centre for Integrated Critical Care, University of Melbourne, Australia

Abstract

Background:

There is potentially considerable variation in the nature and duration of the care provided to hospitalised patients during an infectious disease epidemic or pandemic. Improvements in care and clinician confidence may shorten the time spent as an inpatient, or the need for admission to an intensive care unit (ICU) or high dependency unit (HDU). On the other hand, limited resources at times of high demand may lead to rationing. Nevertheless, these variables may be used as static proxies for disease severity, as outcome measures for trials, and to inform planning and logistics.

Methods:

We investigate these time trends in an extremely large international cohort of 142,540 patients hospitalised with COVID-19. Investigated are: time from symptom onset to hospital admission, probability of ICU/HDU admission, time from hospital admission to ICU/HDU admission, hospital case fatality ratio (hCFR) and total length of hospital stay.

Results:

Time from onset to admission showed a rapid decline during the first months of the pandemic followed by peaks during August/September and December 2020. ICU/HDU admission was more frequent from June to August. The hCFR was lowest from June to August. Raw numbers for overall hospital stay showed little variation, but there is clear decline in time to discharge for ICU/HDU survivors.

Conclusions:

Our results establish that variables of these kinds have limitations when used as outcome measures in a rapidly evolving situation.

Funding:

This work was supported by the UK Foreign, Commonwealth and Development Office and Wellcome [215091/Z/18/Z] and the Bill & Melinda Gates Foundation [OPP1209135]. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Editor's evaluation

This large multicenter study tracked the clinical journeys of COVID-19 hospitalized patients over 2020, and found variations in clinical outcomes over time. This paper will be of interest to the large class of clinicians, public health workers and healthy policy makers who want to know the variation in the nature and duration of the care provided to hospitalised patients during an infectious disease epidemic. The study highlights the importance of maintaining the capacity of registration of infectious disease like COVID-19, during a pandemic and after. While the cohort recruited patients from multiple countries, the vast majority of patients came from the UK, so the results are most applicable to this country.

https://doi.org/10.7554/eLife.70970.sa0

Introduction

During an epidemic or pandemic of a novel infectious disease, variations in the duration of each stage of a hospitalised patient’s progress from symptom onset, to hospital admission, and hence to outcome are critical for an effective response. Clinicians use these data as a proxy for disease severity, and to provide prognostic information to patients and their families. Policy makers use these data to inform system wide planning for staffing, infrastructure, to predict requirements for consumables (such as personal protective equipment), and to assess performance of the hospital system. And for clinical research, these measures are used as trial outcomes to determine the efficacy of novel treatments.

Often, the extent to which patient journeys vary during an epidemic is not understood. There are changes in clinical practice (World Health Organisation, 2021) – clinical understanding of the natural history of diseases improves with time (Docherty et al., 2021), and so too does confidence in safe discharge criteria or in alternative models of care (Rojek and Horby, 2016), such as remote monitoring (Nunan et al., 2020; Bell et al., 2021). Moreover, the introduction of effective treatments (Rochwerg et al., 2020) and standardisation of care may rapidly reduce the severity or time course of illness (Dennis et al., 2021). However, decisions about whether to admit or escalate care are also dependent on logistic factors such as the availability of resources (e.g. ventilators, intensive care beds, staff) that may be rationed during the peak of a pandemic, but abundant at other phases of an outbreak (Tyrrell et al., 2021; National Institute for Health and Care Excellence, 2021; Pagel et al., 2020). There may also be changes in policy to admit patients for indications that are not clinical – such as to facilitate effective quarantine (Wuhan Novel Coronavirus, 2021) or supervise provision of treatments in clinical trials. We hypothesise that there is significant variation in the patient journey over a pandemic period, and that this variability may limit the way these data can be responsibly used.

In this paper, we assess temporal changes in hospital admission, length of stay, and escalation of care for hospitalised patients with SARS-CoV-2 infection included in the International Severe Acute Respiratory and emerging Infection Consortium (ISARIC) WHO Clinical Characterisation Protocol International cohort (ISARIC Clinical Characterisation Group, 2020). This is to our knowledge the largest, prospective international cohort including standardised clinical data, and, as of the time of writing, includes data collected from 26 January 2020 to 20 September 2021 on 708,085 people hospitalised with COVID-19 in 1669 sites in 64 countries.

We use this dataset to determine whether these variables did indeed change over the course of the SARS-CoV-2 pandemic during 2020, and where there are changes, explore if there are predictable influences that account for this.

Methods

As previously described (ISARIC Clinical Characterisation Group, 2021), eligible for recruitment were patients with confirmed or suspected COVID-19 infection admitted to an ISARIC partner site and submitted to the ISARIC-hosted REDCap system. Additional data contributed to ISARIC via other mechanisms have not been included due to differences in data structure. The datasets used in analyses in this paper are drawn from a population of all patients with a symptom onset date, or hospital admission date, recorded between March and December 2020 inclusive. Follow-up could be conducted until 8 March 2021, at which point the dataset was closed. Some patients had a recorded hospital admission date before their symptom onset date. In most cases, this would represent a nosocomial infection, but sometimes it could instead be that the patient was coincidentally admitted for a separate medical condition during their incubation period. In either case their admission date would not represent the start of their hospital treatment for COVID-19, and so we recorded the latter variable, hereafter ‘COVID-19 admission date’, as the later of the symptom onset date and hospital admission date. Patients were followed until they left the study site, due to death, hospital discharge, or another reason (such as transfer to another facility). Patients lost to follow-up before any of these outcome events were included, unless the time to that event was the variable of interest in a particular analysis. (For example, in question 5, below, time to death and discharge were used as dependent variables, and patients lost to follow-up, for whom this time was not recorded, were excluded.)

We used the complete dataset described above to explore temporal variation in six variables or collections of variables. Not all patients had recorded information for the variables of interest in each question, so, in each case, a subset was analysed. The questions and data subsets were as follows:

Question 1

Variation in the time from symptom onset to hospital admission. Patients were excluded if this variable was not available, or if they were admitted prior to symptom onset.

Question 2

Variation in the proportion of patients being admitted to an ICU or HDU. Patients were excluded if this variable was not available.

Question 3

Variation in the time from COVID-19 admission to ICU/HDU admission. Patients were excluded if they were never admitted to an ICU or HDU, or if this variable was otherwise not available.

Question 4

Variation in the overall case fatality rate. Patients were excluded if the final outcome of their hospital stay was either not recorded or recorded as something other than “death” or “discharge” (for example, transfer to another facility).

Question 5

Variation in the time from COVID-19 admission to death or discharge. (We describe either as an “outcome”.) Exclusions were as in question 4, as well as patients who had a recorded outcome but no recorded outcome date.

Question 6

Variation in the status of patients (admitted, ICU/HDU admitted, dead, discharged, or unknown outcome) on a given day after admission. Excluded here were patients whose ICU/HDU status on the day of admission was unknown.

Further filtering was done to (a) remove any nonsensical values (such as recorded time of hospital admission after hospital exit), (b) remove patients admitted to hospital in 2021 for all questions except 1 (such patients were included when exploring the latter due to right censoring concerns if they were omitted), and (c) when considering hospital admission (question 1), ICU/HDU admission (questions 2 and 3) or final outcome (questions 4–6), exclude patients for whom the time to that event was in the top 2.5 % of recorded values (as this range included extreme outliers that may have been the result of incorrect data entry). We designate these three limits by ladmission, lICU and Ioutcome respectively.

The values of lICU and Ioutcome also define a time period over which the events of ICU/HDU admission and final outcome (respectively) can be defined for questions 2 and 4. Thus, in question 4, the actual variable of interest is death or discharge within Ioutcome days. For question 2, the variable is ICU/HDU admission within lICU days of observation. As a result, we additionally excluded patients with incomplete follow-up who were observed for less than lICU days from COVID-19 admission without experiencing ICU/HDU admission, as such an event may have occurred within the time limit without it being observed.

For all analyses with a single outcome variable, we plotted its mean value against the epidemiological week of symptom onset (question 1) or COVID-19 admission (others), both overall and with respect to various variables of interest (e.g. age group).

For the exploration of patient status by day after COVID-19 admission (question 6), the progress of a patient along the course of their hospital stay was visualised by means of Sankey diagrams. Five states were considered: ward occupancy, ICU/HDU occupancy, the final outcomes of death or discharge, and unknown outcome. We recorded the state of each patient on the day of admission, on every subsequent day, and their final outcome. Where a patient’s exact location (ward or ICU/HDU) in the hospital was not recorded on a given day, their last known location was used. For the figures in this article, we present only the data on day of COVID-19 admission (A), 3 days later (A + 3), 7 days later (A + 7), and final outcome (O + 1). An interactive version of this diagram is currently under development and will be made available to the research community as soon as possible.

The four most frequent symptoms at admission were cough, fatigue, fever, and shortness of breath. We introduced a new variable to the dataset counting the number of these present at admission for each patient. Missing data was disregarded here, so this represents a lower bound.

Statistical analysis

Multivariable linear regression was used to investigate factors associated with time from onset of symptoms to hospital admission (question 1), time from COVID-19 admission to ICU/HDU admission (question 3), and time from COVID-19 admission to death or discharge (question 5). In all cases, the dependent variable was log-transformed and a pseudocount of 1 added in order to prevent taking logarithms of zero. Multivariable logistic regression was used to investigate factors associated with ICU/HDU admission (question 2) and fatal outcome (question 5), and to adjust for these factors as potential confounders for our primary outcome variables; for a full list of these variables, see Supplementary file 1. For all regression analyses, we analysed the presence of comorbidities as covariables. As there was a considerable amount of missing data for each of these, we introduced an ‘unknown’ class to the regression models for these variables rather than exclude patients without values for them entirely. After this modification, every regression analysis was performed as a complete case analysis.

The significance of every dependent variable in every model, including interaction terms used in regressions for hCFR (question 4) and time to outcome (question 5) was assessed using the Wald test.

Software

All analyses were performed in R 4.0.3 (R Development Core Team, 2013), with packages including the tidyverse (Wickham et al., 2019), and ggalluvial (Brunson, 2020). Code for processing the data and performing the regression analyses is available, copy archived at swh:1:rev:ce42035d6cf80852089d95264215f7bb487cb998 (Hall, 2021).

Results

Patient characteristics

Our complete dataset consisted of 142,540 patients (60,977 female, 81,325 male, 238 unknown sex), of median age 70 [IQR 56–82], admitted at 620 sites in 47 countries. Table 1 shows a summary of baseline characteristics, and more detail, including country of origin and cross tabulation by month of admission, can be found in Supplementary file 2.

Table 1
Baseline characteristics of the included patients.

*Some patients admitted in early 2021 are included in order to fully represent patients with symptom onset in December 2020.

VariableValueCount%
Month of admissionMarch27,10819.4
April42,26730.3
May12,3118.82
June5,3423.83
July2,8112.01
August2,2181.59
September5,2653.77
October13,8229.91
November15,15510.9
December13,2059.47
SexFemale59,71942.8
Male79,55057
Unknown2350.168
Age group0–192,6971.93
20–399,3026.67
40–5930,39921.8
60–6922,81516.4
70–7929,90121.4
80+41,57129.8
Unknown2,8192.02
Symptom onset post-admissionNo118,87485.2
Yes11,6958.38
Unknown8,9356.4

Table 2 shows the prevalence of symptoms at admission and comorbidities. A total of 1030 individuals (0.7%) were pregnant women.

Table 2
Prevalence of symptoms at hospital admission and comorbidities.

The final column gives the number of times the condition is recorded as present over the number of times its presence or absence is recorded (i.e. the data is non-missing). Designated “common” symptoms are indicated with a (C); the number and percentages of patients presenting with combinations of these are separately presented.

Name% presentN (present)/n (data recorded)
Symptoms at admissionCough (C)66.6(87218/131002)
Shortness of breath (C)64.4(89611/139244)
Fever (C)63.4(84665/133494)
Fatigue (C)44.7(52837/118184)
Confusion24.9(31167/125123)
Vomiting19.9(24577/123625)
Myalgia18.8(20921/111419)
Diarrhoea18.2(22375/123121)
Headache12(13424/112069)
Abdominal pain11.1(13294/120175)
Ageusia8.8(6758/76396)
Wheezing7.7(8846/115511)
Anosmia6.8(5281/77751)
Runny nose3.4(3704/108623)
Ulcers2.2(2291/105394)
Bleeding1.8(2093/119266)
Rash1.5(1713/113636)
Seizures1.5(1801/120755)
Lymphadenopathy0.7(774/112245)
Conjunctivitis0.5(553/113083)
Ear pain0.5(484/94873)
Number of recorded ‘common’ symptoms (C)07.6(10836/142540)
120.5(29257/142540)
226.4(37681/142540)
329(41359/142540)
416.4(23407/142540)
ComorbiditiesHypertension47.6(50174/105433)
Chronic cardiac disease29.7(38175/128374)
Diabetes16.8(20037/119155)
Chronic pulmonary disease16.5(22040/133662)
Chronic kidney disease15.7(20894/133256)
Obesity14.4(16624/115463)
Asthma13.2(17656/133341)
Dementia12.9(16404/127239)
Smoking12.8(7299/57164)
Chronic neurological disorder11.5(15248/132789)
Rheumatological disorder11.2(13814/123453)
Malignant neoplasm9.3(12343/132537)
Chronic haemotologic disease4.1(5117/123739)
Liver disease3.5(4443/128733)
Malnutrition2.6(3094/119518)
HIV/AIDS0.4(515/119235)

A basic summary of the various components of the patient journey that we investigated can be found in Table 3.

Table 3
Summary of the components of the inpatient journey and their variation over the course of 2020.

All time periods are in days. Patients are categorised by month of symptom onset for onset to admission, and by month of COVID admission in all other cases. Patients with COVID admission in 2021, who are included in the analysis of time from onset to admission if their onset date was in 2020, are not listed here as they are excluded from any analysis where the outcome variable is not time from onset to admission. “Outcome” is either death or discharge, and the ‘admission to outcome’ column gives the total length of hospital stay. For all durations, the top 2.5 % of values are excluded as potentially mis-entered.

MonthOnset to hospital admissionProportion entering ICU/HDUCOVID-19 admission to ICU/HDUhCFRCOVID-19 admission to deathCOVID-19 admission to dischargeCOVID-19 admission to outcome
MeanSDMeanSDMeanSDMeanSDMeanSD
March6.825.150.251.782.340.33118.4110.99.7110.89.29
April4.274.530.161.632.360.339.17.9610.39.149.918.78
May4.094.580.171.452.450.3108.41119.6210.99.28
June4.44.510.31.042.120.27108.6510.58.8310.58.78
July4.774.220.351.12.290.21118.689.538.379.888.46
August5.494.60.361.412.490.22128.828.887.889.448.16
September6.35.010.241.382.210.22149.819.258.6410.29.08
October5.724.890.191.692.590.26128.699.788.7510.58.81
November5.174.750.181.482.490.26128.389.18.089.788.24
December4.424.210.221.512.480.291189.528.16108.15

Time from symptom onset to hospital admission (Question 1)

A total of 11,695 patients (8.2%) were recorded as having symptom onset following hospital admission, while for 8,935 (6.3%) patients this information was missing. After excluding these, we analysed length of illness before hospitalisation for those patients for whom it was recorded (n = 127,915, 89.7%). The 97.5 % quantile of time to admission (ladmission) was 24 days, and patients with recorded durations longer than this were excluded as described above. The median time from symptom onset to admission was 5 days (IQR 1–8). This variable showed a marked decline during March, from a median of 9 days (IQR 5–14) for patients with onset in the week beginning March 1–3 (IQR 0–7) in that beginning April 5. Little further variation occurred until late July, when a gradual increase started, which then peaked at a median of 6 (IQR 2–9) for the weeks in late August and early September before a decline to a low of a median 4 (IQR 1–7) days in November; this was followed by another slight increase in December. Times from onset of symptoms to admission were shortest in the oldest and youngest age groups (Figure 1b). Patients with a fatal outcome had, generally, shorter time from onset of symptoms until admission compared to survivors (Figure 1c).

Time from reported symptom onset to hospital admission, by week of reported symptom onset.

(A) Blue cells represent binned patients, with darker colours corresponding to more individuals. The black line represents the mean. (B)-(D) Mean time to admission plotted by patient characteristics: (B) age group, (C) final outcome, (D) number of the four most common symptoms (cough, fatigue, fever, and shortness of breath) present upon admission.

Figure 1—source data 1

Number of individuals for each combination of week of symptom onset and count of days from symptom onset to admission.

https://cdn.elifesciences.org/articles/70970/elife-70970-fig1-data1-v2.zip
Figure 1—source data 2

Mean number of days from symptom onset to admission by week of symptom onset.

https://cdn.elifesciences.org/articles/70970/elife-70970-fig1-data2-v2.zip
Figure 1—source data 3

Mean number of days from symptom onset to admission by week of symptom onset, by age group.

https://cdn.elifesciences.org/articles/70970/elife-70970-fig1-data3-v2.zip
Figure 1—source data 4

Mean number of days from symptom onset to admission by week of symptom onset, by final outcome (death or discharge).

https://cdn.elifesciences.org/articles/70970/elife-70970-fig1-data4-v2.zip
Figure 1—source data 5

Mean number of days from symptom onset to admission by week of symptom onset, by number of common symptoms recorded at admission.

https://cdn.elifesciences.org/articles/70970/elife-70970-fig1-data5-v2.zip

The four most frequent symptoms at admission were cough, fever, shortness of breath, and fatigue; we class these as ‘common’ (see Methods). The number of these that were present increased with time to hospital admission (Figure 1d), with the shortest durations of all occurring amongst those presenting with none of them. Amongst the 4636 patients in this analysis presenting with none, the most common other symptoms were confusion (51.6%), vomiting (31.7%), abdominal pain (26%), and diarrhoea (18.9%). Within this group, confusion was the single most common presenting symptom documented in patients over 60, while in younger age groups the most prevalent symptoms were gastrointestinal (see Supplementary file 3).

We further explored this question using a multivariable linear regression analysis (Supplementary file 4). When compared to April, times from symptom onset to admission were shorter in May and June, and longer in March and from August onwards (all p < 0.01; see Supplementary file 4 for confidence intervals). There was very strong evidence of an overall effect of month of onset on time from onset to admission (Wald test p < 0.001). Patients aged 40–59 showed the longest times to admission when compared to any other age group (all p < 0.001; see Supplementary file 4 for CIs). Time from symptom onset to admission was also positively associated with the number of ‘common’ symptoms (25.1 % increase per symptom, 95% CI 24.5%–25.7%), male sex (3.3 % increase, 95% CI 2.2%–4.4%) and discharge as the final outcome (14.3 % increase, 95% CI 12.9%–15.7%).

ICU/HDU admission and time to ICU/HDU admission (questions 2 and 3)

Of the 139,504 patients with COVID-19 admission in 2020, 136,849 (98.1%) had recorded data on whether they ever admitted to an ICU/HDU or not; of these, 28,171 (20.6%) had been admitted at least once. Where time to ICU/HDU was recorded, the 97.5 % quantile of this duration (lICU) was 13 days. We excluded patients for whom this variable was greater than that or unknown, along with those whose outcome was unknown and who had follow-up for less than 13 days with no ICU/HDU admission, for a total of 122,368 patients. The outcome variable in this section is thus ICU/HDU admission within 13 days of COVID-19 admission. The proportion of individuals experiencing this showed a marked decline over March followed by a renewed peak, and then subsequent decline, in June through August (Figure 2A). The oldest age group (80+) had by far the smallest proportion of ICU/HDU admissions over the whole timeline (5.2%, compared with, for example, 33.5 % in the age-group 60–69). In a multivariable logistic regression model (Supplementary file 5), the following patterns were observed: there were higher odds of ICU/HDU admission during all months except May and November, when compared to April (all p < 0.05, see Supplementary file 5 for CIs); those aged 80+ had lower odds of ICU/HDU admission (OR 0.12 for admission when compared to the 40–59 age group, 95% CI 0.11–0.13). Males were more likely to be admitted to ICU/HDU (OR 1.57, 95% CI 1.45–1.63). Patients who died had greatly increased odds of having been previously admitted (OR 6.1, 95% CI 5.8–6.41). Compared to those with symptom onset less than a week before hospital admission, patients with admitted prior to onset had lower odds of being admitted to ICU/HDU (OR 0.32, 95% CI 0.28–0.36), whereas those with longer times to hospital admission had increased odds (OR 1.43 for 7–13 days, 95% CI 1.37–1.5, 1.31 for 14 or more days, 95% CI 1.22–1.4). An overall effect of month of COVID admission on odds of ICU/HDU admission was highly significant (Wald test p < 0.0001). Comorbidities associated with higher odds of admission were hypertension (OR 1.27, 95% CI 1.21–1.34) and obesity (OR 1.78, 95% CI 1.69–1.88), whereas a wide variety of serious or chronic medical conditions were associated with lower odds (see Supplementary file 5), as was smoking (OR 0.79, 95% CI 0.72–0.87). The most extreme fitted odds ratio for a comorbidity with a positive association was 1.78 for obesity, while that for an inverse association was 0.19 for dementia.

Patients entering ICU/HDU within 13 days of COVID-19 admission (A) and time from COVID-19 admission to ICU/HDU admission (B) over time.

Each line is the proportion (A) or mean value (B) amongst all patients (black, dotted) or patients in each age group (coloured).

Figure 2—source data 1

Proportion of individuals entering ICU by week of COVID-19 admission, according to age group and overall.

https://cdn.elifesciences.org/articles/70970/elife-70970-fig2-data1-v2.zip
Figure 2—source data 2

Mean time in days from COVID-19 admission to ICU admission by week of COVID-19 admission, according to age group and overall.

https://cdn.elifesciences.org/articles/70970/elife-70970-fig2-data2-v2.zip

Of the 28,171 patients with recorded ICU/HDU admission, 27,167 (96.4%) had non-missing data for time from COVID-19 admission to first ICU/HDU admission. The 97.5 % quantile rule again excluded patients whose value of this variable was greater than 13 days. The median time to ICU/HDU was 1 day (IQR 0–2). Raw time trends in this variable were modest (Figure 2B). Multivariable linear regression (Supplementary file 6) nevertheless did show evidence for an overall association with month of COVID-19 admission (Wald test p < 0.001), with, when compared to April, evidence for longer times to ICU/HDU in March, October, November and December (all p < 0.05; see Supplementary file 6). Time to ICU/HDU also showed a general increase with age (Wald test for overall association p < 0.001). There was no evidence of an association with final outcome (death or discharge) or with sex. Compared to patients admitted to hospital within a week of symptom onset, those admitted prior to onset had a 67.6 % increase in time to ICU/HDU (95% CI 55.9–80.2%) while those with longer times to admission had shorter times to ICU/HDU (8.6 % decrease for 7–13 days, 95% CI 6.6–10.6%, 14.9 % decrease for 14 or more days, 95% CI 11.9–17.8%). Comorbidities associated with longer time to ICU/HDU were asthma (6.5 % increase, 95% CI 3.3–9.7%), chronic haematological disease (20.4 % increase, 95% CI 12.6–28.8%), and chronic kidney disease (9.9 % increase, 95% CI 6–14.1%). In contrast, obesity (7.7 % decrease, 95% CI 5.2–10%), diabetes (3.7 % decrease, 95% CI 0.9–6.3%) and smoking (5.2 % decrease, 95% CI 0.4–9.8%) were associated with shorter time to ICU. There was also evidence of a longer time to ICU/HDU amongst pregnant patients (compared to non-pregnant females) (15.6%, 95% CI 1.7–31.4%).

Case fatality rate and time from COVID-19 admission to outcome (questions 4 and 5)

We next analysed the final outcome of death or discharge, and the total time from hospital admission to one of those outcomes, in a set of 116,537 patients admitted during 2020 with one of those outcomes recorded (83.5 % of the total admitted during 2020). The 97.5 % quantile of time to outcome (loutcome) was 45 days, and once more patients with recorded durations in hospital longer than this were excluded; as a result, in practice the outcome here is death or discharge within 45 days. (As patient data was collected until 8 March 2021 and all patients here were admitted in 2020, patients admitted at the end of the period of interest had the same chance of complete follow-up as any other.) The raw hCFR was 0.3. The median time to death was 8 days (IQR 4–15) and to discharge 7 (IQR 4–14). (Amongst patients with no recorded outcome, excluded here, the median follow-up time was 9 days with an IQR of 2–22; the median follow-up for all patients regardless of outcome, recorded or not, was also 9 days with an IQR of 5–16). Over the entire 10-month period of interest (Figure 3—figure supplement 1), peak hCFR was 0.35 in the week beginning 8 March. There was a decline over the spring to a low of 0.17 in the week beginning 12 July, but this trend subsequently reversed and reached 0.32 by mid-December. At the same time, the mean time from admission to outcome in this whole population showed very little change following a dramatic decline during March, from 16 days in the week beginning 1 March to 10 at the start of April (Figure 3—figure supplement 2). These overall patterns, however, mask substantial variation by on ICU/HDU admission and, in the latter case, outcome (Figure 3). The trend in hCFR is largely driven by patients who were not admitted to an ICU or HDU. The most consistent decline in time to outcome was observed in ICU/HDU admissions who survived (a decline in the mean of 7.6 days between the first and last weeks studied, Figure 3B, bottom left) while survivors with no ICU/HDU admission showed, as with the overall trend, little change after March (bottom right). Variation in time to death appeared very modest amongst patients with an ICU/HDU admission (top left), while there was a distinct peak around August and September in those without (top right). When age is also considered (Figure 3—figure supplements 3 and 4), a notable additional pattern is the clear correlation of time to discharge and age in surviving non-ICU/HDU patients, which is much less obvious, if present at all, in patients with an ICU/HDU admission.

Figure 3 with 4 supplements see all
Temporal trends in outcome and time to outcome.

(A) Case fatality ratio in patients experiencing death or discharge within 45 days of COVID-19 admission, by recorded ICU/HDU admission. (B) Mean time from COVID-19 admission to the outcome of death or discharge, further faceted by ICU/HDU admission. Error bars represent 95 % confidence intervals. Numbers along the x-axis indicate the numbers of patients involved in each category.

Figure 3—source data 1

Estimate and 95 % confidence interval for hCFR by week of COVID-19 admission, according to ICU/HDU admission status.

https://cdn.elifesciences.org/articles/70970/elife-70970-fig3-data1-v2.zip
Figure 3—source data 2

Estimate and 95 % confidence interval for time from COVID-19 admission to outcome by week of COVID-19 admission, according to ICU/HDU admission status and outcome.

https://cdn.elifesciences.org/articles/70970/elife-70970-fig3-data2-v2.zip
Figure 3—source data 3

Estimated overall hCFR by week of COVID-19 admission.

https://cdn.elifesciences.org/articles/70970/elife-70970-fig3-data3-v2.zip
Figure 3—source data 4

Estimated mean time from COVID-19 admission to outcome by week of COVID-19 admission.

https://cdn.elifesciences.org/articles/70970/elife-70970-fig3-data4-v2.zip
Figure 3—source data 5

Estimate and 95 % confidence interval for hCFR by week of COVID-19 admission, according to ICU/HDU admission status and age group.

https://cdn.elifesciences.org/articles/70970/elife-70970-fig3-data5-v2.zip
Figure 3—source data 6

Estimate and 95 % confidence interval for time from COVID-19 admission to outcome by week of COVID-19 admission, according to ICU/HDU admission status and outcome, and age group.

https://cdn.elifesciences.org/articles/70970/elife-70970-fig3-data6-v2.zip

The results of the three multivariable regression analyses can be seen in Table 4; some variables (country and the ‘unknown’ class for comorbidities) are excluded for brevity, but the full version is provided as Supplementary file 7. Note that all variables are adjusted for all others, which is also the case for all the other regressions presented in this paper. There was strong evidence of an association of month of COVID-19 admission with all three variables (Wald test p < 0.001 in all cases).

Table 4
Combined results of a logistic regression analysis identifying predictors of death as an outcome, and two linear regression analyses identifying correlates of time to death and time to discharge.

All analyses are multivariable. For brevity, the country variable, as well as the ‘unknown’ class for each comorbidity (representing patients with missing data for that condition) are omitted here; see Supplementary file 7 for a version with them included. The p-values of Wald tests for the inclusion of each variable in each regression are included as a separate column; these were calculated including the ‘unknown’ class for comorbidities.

Odds ratio (death v discharge)Time to death (% change, days)Time to discharge (% change, days)
Estimate95 % confidence intervalWald test p-valueEstimate95 % confidence intervalWald test p-valueEstimate95 % confidence intervalWald test p-value
Month of COVID admission (ref: April)< 0.001< 0.001< 0.001
March1.1(1.1, 1.2)14.7(12.2, 17.3)3.7(2.1, 5.2)
May0.7(0.7, 0.8)15.5(12.1, 19.1)1.3(–0.6, 3.3)
June0.5(0.5, 0.6)20.8(14.2, 27.7)–2.8(−5.6,–0.02)
July0.3(0.3, 0.4)28.1(14.9, 42.8)–8.5(−12.2,–4.7)
August0.4(0.3, 0.5)47.2(29.5, 67.3)–10.8(−15.0,–6.4)
September0.6(0.5, 0.6)40.7(32.6, 49.3)–3.2(−5.8,–0.5)
October0.6(0.6, 0.7)33.2(28.9, 37.7)–1.4(–3.2, 0.4)
November0.6(0.6, 0.7)26.7(22.8, 30.8)–3.5(−5.2,–1.8)
December0.8(0.8, 0.9)26.3(22.3, 30.4)2.2(0.2, 4.2)
Age group (ref: 40–59)< 0.001< 0.001< 0.001
10–190.3(0.2, 0.4)–0.5(–24.3, 30.6)–33.6(−35.7,–31.4)
20–390.3(0.2, 0.3)–1.7(–15.9, 14.8)–16.9(−18.5,–15.3)
60–692.9(2.7, 3.2)–2.1(–6.8, 3.0)17.4(15.6, 19.2)
70–796.1(5.7, 6.6)1.9(–2.6, 6.6)36.4(34.3, 38.6)
80+10.9(10.1, 11.8)4.3(–0.3, 9.0)52.9(50.4, 55.4)
ICU/HDU admission7.6(6.8, 8.4)< 0.00168.3(59.1, 78.0)< 0.001140.1(133.0, 147.3)< 0.001
Sex (ref: female)< 0.001< 0.0010.0059
Male1.3(1.3, 1.4)2.9(1.4, 4.5)1.2(0.2, 2.1)
Pregnant (ref: female, no)0.0970.069< 0.001
Female, yes0.6(0.4, 1.0)–5.4(–27.4, 23.3)–18.6(−22.3,–14.6)
Days from symptom onset to hospital admission (ref: 0–6)< 0.001< 0.001< 0.001
Symptom onset post-admission1.3(1.2, 1.3)20.7(17.9, 23.5)45.7(42.9, 48.6)
7–130.7(0.7, 0.8)–1.7(–3.5, 0.1)–3.6(−4.6,–2.5)
14+0.7(0.7, 0.8)–0.2(–2.9, 2.5)–6.1(−7.6,–4.5)
Comorbidities
Asthma0.9(0.9, 1.0)< 0.0012.5(0.2, 4.9)0.11–1.6(−2.9,–0.3)0.048
Chronic cardiac disease1.2(1.2, 1.3)< 0.001–2.8(−4.4,–1.3)< 0.0011.3(0.1, 2.6)0.075
Chronic haemotologic disease1.2(1.1, 1.3)< 0.0010.2(–3.1, 3.6)0.625.8(3.1, 8.6)< 0.001
Chronic kidney disease1.4(1.4, 1.5)< 0.001–2.4(−4.1,–0.6)0.0125.8(4.2, 7.4)< 0.001
Chronic neurological disorder1.4(1.3, 1.4)< 0.001–0.7(–2.7, 1.4)0.6115.3(13.4, 17.2)< 0.001
Chronic pulmonary disease1.4(1.3, 1.4)< 0.001–2.2(−3.9,–0.5)0.0343.9(2.4, 5.3)< 0.001
Dementia1.5(1.4, 1.6)< 0.001–1.2(–3.1, 0.8)0.439(7.0, 11.0)< 0.001
Diabetes1.1(1.1, 1.2)< 0.001–4.2(−6.0,–2.3)< 0.0011.8(0.4, 3.2)< 0.001
HIV/AIDS1.2(1.0, 1.6)0.0054(–8.2, 17.8)0.582.2(–5.1, 10.1)0.0025
Hypertension1(0.9, 1.0)< 0.0010.2(–1.5, 1.9)0.142(0.8, 3.2)< 0.001
Liver disease1.4(1.3, 1.6)< 0.0015.2(1.1, 9.5)< 0.00112(9.0, 15.1)< 0.001
Malignant neoplasm1.4(1.4, 1.5)< 0.0013.5(1.2, 5.8)0.00592.2(0.3, 4.1)< 0.001
Malnutrition1.3(1.2, 1.5)< 0.0010.6(–3.6, 5.0)0.8111.5(7.5, 15.7)< 0.001
Obesity1.1(1.1, 1.2)< 0.001–5.5(−7.7,–3.2)< 0.0016.3(4.8, 7.9)< 0.001
Rheumatological disorder1(0.9, 1.0)0.0452.3(0.1, 4.7)0.0520.6(–1.0, 2.2)0.44
Smoking1.1(1.1, 1.2)< 0.0014.2(0.5, 8.0)0.12.2(0.04, 4.4)0.034
Interaction: ICU/HDU admission _ month of admission (ref: April)< 0.001< 0.001< 0.001
March1(0.9, 1.1)–14.3(−18.1,–10.4)0.6(–3.1, 4.5)
May1.1(0.9, 1.3)–12(−17.6,–6.1)-9(−13.9,–3.9)
June1.2(1.0, 1.5)–9.1(−16.8,–0.8)–11.3(−17.1,–5.1)
July1.4(1.1, 1.9)–14.8(−25.8,–2.1)–18.8(−25.2,–11.8)
August1.1(0.8, 1.6)–27.9(−38.4,–15.6)–15.6(−23.2,–7.3)
September1.1(0.9, 1.4)–16.6(−24.1,–8.4)–6.9(−13.0,–0.3)
October1.1(1.0, 1.3)–13.7(−18.9,–8.2)–11.6(−15.9,–7.0)
November1.1(1.0, 1.3)–6.6(−12.1,–0.7)–15.9(−20.0,–11.6)
December0.8(0.7, 0.9)–11.6(−17.0,–5.8)–27.9(−31.4,–24.2)
Interaction: ICU/HDU admission _ age group (ref: 40–59)< 0.001< 0.001< 0.00
10–190.8(0.4, 1.4)–42.8(−59.3,–19.6)1.7(–6.0, 10.2)
20–391.8(1.4, 2.4)–11.8(–25.9, 4.9)0.8(–3.6, 5.4)
60–690.7(0.7, 0.8)2.5(–3.7, 9.0)–7.2(−10.4,–3.9)
70–790.6(0.5, 0.7)–11.8(−16.8,–6.6)–21.8(−24.9,–18.5)
80+0.4(0.3, 0.5)–28.4(−32.9,–23.6)–40.1(−43.6,–36.4)
Observations102,14731,25070,897

Amongst non-ICU/HDU patients, the month with the greatest odds of death was March (OR 1.12 compared to April, 95% CI 1.07–1.17), while that with the smallest was July (OR 0.35, 95% CI 0.29–0.43). April (the reference category) was the month with the shortest time to death, while August had the longest (54.2 % increase, 95% CI 32.9–78.8%). Variation in time to discharge was more modest; the month with the largest value of this variable was March (3.7 % increase, 95% CI 1.9–5.5%), and that with the shortest was August (12.3 % decrease, 95% CI 7.2–17.1%).

ICU/HDU admission was associated with a 7.56-fold higher odds of death (95% CI 6.81–8.4), a 81.8 % increase in time to death (95% CI 70.3–94.2%) and a 171.6 % increase in time to discharge (95% CI 162.3–181.3%). As a result of the patterns observed in Figure 3, we also fitted interaction terms of month of COVID-19 admission with IDU/HCU admission. Their inclusion was consistently statistically significant (Wald test p < 0.001 in all cases), although for odds of death this ceases to be true when December is removed (p = 0.23). Hence, the overall increased odds of death amongst ICU/HDU patients was significantly mitigated in December (combined OR 6.07 vs non-ICU/HDU admissions in December, 95% CI 5.23–7.06). There was no evidence that ICU/HDU patients admitted in March, May, or August had a longer time to death than April, but the estimates for all other months were significantly greater, with the peak in November (21.5 % increase, 95% CI 15.1–28.2%). The longest times to discharge in these patients were in March (4.2 % increase vs April, 95% CI 0.8–7.8%) and the shortest in December (28.1 % decrease, 95% CI 24–32%).

Increasing age was associated with monotonic increases in odds of death and time to discharge, with and without ICU/HDU admission. Time to death showed little evidence of variation by age in non-ICU/HDU patients except for marginal evidence for an increase in the oldest age group (5.7 % increase vs 40–59, 95% CI 0.3–11.2%). In ICU/HDU patients, however, where an interaction term was again fitted, the shortest times to death were recorded in both the youngest (49.2% decrease, 95% CI 35.8–59.9%) and oldest (27.6 % decrease, 95% CI 24.2–30.9%) groups; longest times to death were in middle-aged adults (40-69). Male sex was associated with higher odds of death (OR 1.33, 95% CI 1.29–1.38), and small increases in time to both death (3.4 % increase, 95% CI 1.7–5.2%) and discharge (1.6 % increase, 95% CI 0.5–2.7%). Symptom onset following admission was also associated with higher odds of death (OR 1.28, 95% CI 1.21–1.35) and large increases in time to death (24.4 % increase, 21.1–27.8%) and discharge (53.2 % increase, 95% CI 49.7–56.7%). Patients admitted more than a week from symptom onset had lower odds of death, and shorter stays in hospital, regardless of outcome (see Table 4). Where associations with comorbidities were detected, the majority were in the direction of poorer outcomes (increased hCFR, decreased time to death, and increased time to discharge), with a few exceptions. Most notably, asthma was associated with lower odds of death (OR 0.93, 95% CI 0.88–0.97), longer times to death (2.9 % increase, CI 0.2–5.6%) and shorter times to discharge (1.9 % decrease, 95% CI 0.3–3.4%).

To further illustrate these findings, Figure 4 displays time trends in model predictions for hCFR, time to death and time to discharge for typical patients of both sexes in every age group, both for those with disease serious enough to trigger ICU/HDU admission and those without. These patients were assumed to be admitted in the UK after less than a week of symptoms. Each was assigned the median number of comorbidities for their combination of sex and age group in the real dataset, and the exact comorbidities chosen were also the most common in that demographic group; see inset table, Figure 4. (For example, males in the 60–69 age group had a median of two comorbidities recorded, and the two most common were hypertension and chronic cardiac disease.)

Regression model predictions for hospital CFR (A), predicted time to death in fatal cases (B) and predicted time to discharge in non-fatal cases (C) in a set of hypothetical typical patients.

Lines are plotted by month of COVID-19 admission (y-axis), age group (facets, left to right), sex (red: female, blue: male), and ICU admission (solid lines: at least once, dotted lines: never). The inset table (D) lists the comorbidities assigned to the individuals in each combination of sex and age group.

Figure 4—source data 1

Predicted hCFR, time to death and time to discharge for all hypothetical patients.

https://cdn.elifesciences.org/articles/70970/elife-70970-fig4-data1-v2.zip

Status by days since admission

Figure 5 displays Sankey diagrams reflecting the location of patients within hospital (ward or ICU) or their final status (dead, discharged, or unknown) on the day of COVID-19 admission (A), 3 days later (A + 3), 7 days later (A + 7) or, to represent the final status only, 1 day after the last day in hospital (O + 1). The plot is facetted by age group and month of COVID-19 admission. For simplicity, only four months (April, June, August and October) appear in the main figure, but see Figure 5—figure supplement 1 for all months, featuring a total of 129,044 patients (90.5%).

Figure 5 with 1 supplement see all
Sankey diagrams depicting the progress through the inpatient journey for patients with COVID-19 admission in April, June, August and October 2020, and subdivided by age.

Bars are presented for the day of admission (A), 3 and 7 days later (A + 3 and A + 7), and the day after final outcome (O + 1).

Figure 5—source data 1

Number of patients occupying a ward bed, occupying an ICU/HDU bed, dead, discharged and with unknown outcome on the day of admission (A), 3 and 7 days later (A + 3 and A + 7), and the day after final outcome (O + 1), by age group and month of COVID-19 admission.

https://cdn.elifesciences.org/articles/70970/elife-70970-fig5-data1-v2.zip

Discussion

To the best of our knowledge and at the time of publication, this is the largest international cohort of COVID-19 patients in the world. Considerable temporal variations in the events preceding and during hospitalisation for patients with confirmed COVID-19 were observed during the period March to December 2020. We specifically looked at length of illness before admission, probability of ICU/HDU admission, time to ICU/HDU admission for those so admitted, case fatality rate, and duration of admission overall.

These results highlight key findings with practical implications for case management, resource allocation, performance benchmarking, and reporting of outcomes in research, and point to the fact that patients’ journeys vary over time and must be interpreted with the background of transmission intensity, policy, and practice where cases occur. Therefore, static ‘snapshots’ of the situation at any one time may lead to misguided practice and management if not regularly monitored and approaches adapted accordingly.

In a recent preprint (Kirwan et al., 2021), analysed temporal variation on time from hospital admission to death, discharge or ICU/HDU admission amongst a smaller cohort of UK patients. We confirm many of the trends that they identified, including the lower hCFRs over the summer and the increased odds of ICU/HDU admission in middle-aged age groups. They did not, however, detect the increase in the proportion of patients with an ICU/HDU admission during the summer, or the decline in time to discharge amongst non-ICU/HDU patients over the entire time period. As there were many fewer hospitals included in that study than in ours (31 vs 620) this may be suggestive of variation in available ICU/HDU capacity and usage amongst participating sites in the two studies.

Prior to admission

Across all age groups, the length of illness before seeking hospital care was longest in July and August when case numbers were lower, and shortest at the extremities of the age distribution and for females. The latter variations may, at least partially, be explained by differences in health-care-seeking behaviour by different demographics, and by differences in clinical progression of disease for different groups. Along similar lines, the fact that patients who died had consistently shorter duration of illness before hospital admission may reflects the fact that more serious cases evolve more rapidly and those affected seek care earlier. In this scenario, patients admitted after experiencing symptoms for longer than 1 week would be less likely to die because they were less serious cases and the individuals acted accordingly. The peak in time to admission during late summer and early autumn in the Northern Hemisphere may reflect delayed presentation following return from holiday, particularly given the high proportion of UK patients in this dataset and known viral importations to the UK from continental Europe around that time (Hodcroft et al., 2021).

At the same time, when considering the four most frequent symptoms at admission (fever, short of breath, cough, or fatigue), more symptoms were associated with a longer period between onset of symptoms and admission – and this was consistently so across the entire period under observation. This could be ascribed at least in part to variations in individual behaviour; some patients may present to hospital with a single symptom while others may wait a longer period until several have emerged. In addition, these phenomena could also partially be attributed to how case definitions are applied by physicians, or to the patient’s own perceptions, or to those of their families. Some presentations are likely to be more alarming to the latter two groups than others; for example, individuals with none of the four symptoms described above were admitted fastest of all and, amongst these, confusion was the most prevalent other symptom.

During hospitalisation

Treating variables such as final outcome, ICU/HDU admission, or length of stay, as variables that remain static throughout an evolving epidemic is problematic, as demonstrated by our analyses. To give three examples: first, the case fatality rate showed an overall decline from 0.35 for cases admitted in March to 0.21 in July, followed by a renewed increase to 0.29 in December (Figure 3—figure supplement 1). Second, the data underlying the alluvial plots (Figure 5) allow us to determine that the proportion of patients discharged within a week of admission rose from 0.24 in March to a peak of 0.34 in September. Third, the proportion of still-admitted patients occupying an ICU/HDU bed showed considerable variation: for example, at day three this went from 0.19 in March to 0.13 in April, then rose to a peak of 0.38 in August before declining again, reaching a low of 0.15 in November. Variations in clinical care, the influence of treatments, and changes in available bed capacity are all likely to account for many of these differences. In older patients, the availability of social care space is another important variable.

Patients older than 80 had odds of being admitted to ICU/HDU over eight times smaller than those in the 40–59 category, which may reflect prognosis and the expected benefits of ICU/HDU admission, as well as patient preferences. Many serious chronic conditions were also associated with decreased odds, independently of age, likely for similar reasons. These decreased odds are also reflective of the temporality of the data. March and April represent our data’s highest volume, which might reflect hospital capacity and the necessity for ICU/HDU prioritisation. For the patients who were admitted to ICU/HDU, there was no clear trend in the time from hospital admission to transfer to ICU/HDU after March. Length of illness before admission to the hospital and young age were associated with a shorter time from hospital admission to ICU/HDU (for example, a 9.2 % decrease for those waiting 7–13 days from onset compared to those waiting less than a week, and 32 % decrease amongst under-20s compared to the 40–59 age group), while a smaller proportion of older patients are escalated to ICU/HDU (OR 0.51 for ICU/HDU admission in the 70–79 age group) and after a longer time spent in the ward (a 4.2 % increase in the same age group).

Outcome

As mentioned above, in patients with an outcome of death or discharge, hCFRs decreased from 0.35 in March to 0.21 by mid-2020 to increase again to 0.29 in December, mostly following the waves in the pandemic and therefore the number of admissions. System capacity may be an important predictor of patient outcome and may supersede other factors such as increasing case management skills and the influence of new therapies. This also warns against using outcome data that are not adequately controlled to assess efficacy and safety of treatments or other interventions, as effects may rather reflect capacity of a system to provide high-quality care.

We found that shorter time to death is associated with female sex, lack of ICU/HDU admission, and, amongst ICU/HDU patients, the extremes of age. Shorter time to discharge is also associated with female sex and lack of ICU/HDU admission, and this variable increases monotonically with age.

The finding of an association of asthma with reduced disease severity in COVID-19 is not unique to this study (Alberca et al., 2021; Matsumoto and Saito, 2020), but is also not a universal finding (Choi et al., 2021a; Choi et al., 2021b). A number of possible mechanisms for a protective effect have been suggested, including reduced ACE2 expression in the airway (Jackson et al., 2020), eosinophilia (Ferastraoaru et al., 2021), or simply the existing use of beneficial corticosteroids in this population (Halpin et al., 2020).

Cautionary notes in interpreting these findings. First, the dataset analysed is made of patients on the more severe end of the spectrum of disease compared to cases occurring in the community. Second, about half of these patients were hospitalised in just 2 months (March-April), were predominantly from the UK, and about half were over 70 years old. These demographics explain the high raw hCFR and the large proportion of patients presenting with age-related comorbidities – nearly half have hypertension, one-in-five chronic cardiac disease, and one-in-six diabetes. The regression results here should, however, be quite generalisable to hospitalised populations worldwide as country was accounted for as a predictor. Third, there are inherent limitations of observational data, however large the dataset; in particular, we cannot attribute a cause to many of the phenomena described here. It is most notably not entirely possible to unpick biological effects from clinical decisions. As one example, the association of ICU/HDU admission with male sex is may be due not just to increased disease severity amongst males, but also clinician knowledge of the potential for more severe disease. Similarly, we see a lower rate of ICU/HDU admissions amongst individuals whose symptoms started following admission. On the one hand, it is likely that the population of patients with symptoms emerging in hospital had on average less severe disease, as mild community-acquired infections are less likely to present to hospital. On the other, as those patients will receive clinical care starting at the moment of diagnosis, the need for ICU/HDU is likely reduced even in more serious cases. Fourth, some variables are based on patient self-report which can be inexact; for example, it can be clearly seen in Figure 1a that multiples of seven reported days from symptom onset to admission are overrepresented, suggesting reports in units of weeks. Fifth, some variables are not available to us; for example, resuscitation status and suitability for intensive care admission was not collected in our cohort, and without those variables the reasons for death or lack of ICU/HDU admission cannot be entirely unpicked. Similarly, we are not aware of what resource or bed capacity constraints may have affected individual sites at different times. Sixth, there may be a selection bias with respect to calendar time as a result of case volumes. Recruitment was performed by sites, upon identification of a patient with COVID-19 symptoms, according to their capacity, which was determined by the availability of staff to invite informed consent (as applicable) and complete data forms. Capacity will be subject to both geographical and temporal variations, and it is likely that both the proportion of patients recruited and the proportion completing follow-up would be reduced at times of high pressure on the site and the national healthcare system. However, enrolment was prospective, and as such staff would be blind to our outcomes of interest. In addition, while it is possible that individual sites chose patients to recruit (or cease following up) based on clinical characteristics, it is unclear why the basis for these decisions would show a consistent direction of bias amongst diverse locations. Lastly, one should refrain from overinterpreting data: some of the changes observed reflect adjustments in practice and logistics, or combined pressure on health systems, more than actual effects of interventions.

Implications of findings

Often, in high-income countries, patient outcomes are seen through the lens of individualised treatment provided at the clinician patient interface. This paper demonstrates that outbreak epidemiology has an important influence on patient outcomes – the patient journey from likelihood of admission, through to disposition and length of stay in hospital, and overall outcome, change over the course of a pandemic. There are various explanations for variability – systems may at times be overwhelmed and unable to provide the usual quality of care to their patients; patient behaviour may change depending on perceptions of the status of the outbreak and the performance of the healthcare system at a given time; clinician familiarity with management of patients may vary; and changes in transmissibility and virulence are expected to occur.

The observed variability should inform on the limitations of using observational data during a long-lasting pandemic for management purposes in practice, and also question the use of some variables, such as length of stay in hospital or in ICU, as clinical trial outcomes. This demonstrates the importance of controlling for patient outcome data when designing clinical trials; for example, using our data, assessing a new treatment during the months of March to July will have shown a decrease in hCFR from 33%, to 21 % that may have been falsely attributed to a treatment effect without a concurrent randomised control.

At the same time, these findings also highlight the need for preparedness and resilience; the crucial importance of pre-positioned observational data collection systems that are adhered upon by a representative number of sites and are maintained for as long as the pandemic lasts; and the need for such capacity to be kept in-between epidemics.

Data availability

The data that underpin this analysis are highly detailed clinical data on individuals hospitalised with COVID-19. Due to the sensitive nature of these data and the associated privacy concerns, they are available via a governed data access mechanism following review of a data access committee. Data can be requested via the IDDO COVID-19 Data Sharing Platform (http://www.iddo.org/covid-19). The Data Access Application, Terms of Access and details of the Data Access Committee are available on the website. Briefly, the requirements for access are a request from a qualified researcher working with a legal entity who have a health and/or research remit; a scientifically valid reason for data access which adheres to appropriate ethical principles. The full terms are at https://www.iddo.org/document/covid-19-data-access-guidelines. A small subset of sites who contributed data to this analysis have not agreed to pooled data sharing as above. In the case of requiring access to these data, please contact the corresponding author in the first instance who will look to facilitate access. We have provided the R code used to process data and run the regression analysis at https://github.com/ISARICDataPlatform/InpatientJourneyDataProcessing, (copy archived at swh:1:rev:ce42035d6cf80852089d95264215f7bb487cb998). Source data for all figures has also been provided.

References

  1. Software
    1. R Development Core Team
    (2013) R: A Language and Environment for Statistical Computing
    R Foundation for Statistical Computing, Vienna, Austria.

Decision letter

  1. Talía Malagón
    Reviewing Editor; McGill University, Canada
  2. Miles P Davenport
    Senior Editor; University of New South Wales, Australia

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Decision letter after peer review:

Thank you for submitting your article "Ten months of temporal variation in the clinical journey of hospitalised patients with COVID-19: an observational cohort" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Miles Davenport as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1. Provide further rationale for the separation of gastrointestinal and common symptoms for analyses.

2. There are several indications that the cohort may not be entirely representative of all hospitalised patients with COVID-19 in the study settings, including a surprisingly high case fatality rate and few participants recruited at some study sites. Please provide more discussion regarding the representativeness of the cohort, eligibility criteria, and study participation rates.

3. Please provide date of end of follow-up and an estimate of the average/total follow-up time.

4. Indicate whether all participants were followed-up over equivalent periods of time to assess study outcomes, in order to justify the use of logistic regression rather than survival analysis.

5. Provide more information regarding what variables are being included in the model, and whether Table 4 represents results from a model including all variables or multiple separate models. This influences how readers interpret the variable coefficients, as it is important to know what variables have been adjusted for and whether the ratios should be interpreted as crude or adjusted associations, and whether interactions are being accounted for in the estimated effects.

6. While less essential, the reviewers made several other suggestions/queries that should be considered by the authors in their revision.

Reviewer #1:

In this study, COVID-19 hospitalized patient trajectories were tracked over 2020 in order to assess whether patient outcomes changed over time. The study of these time trends can be useful for assessing how health systems respond to pandemics, and may help in better planning for future outbreaks and pandemics.

Some of the major strengths of this study include its very large sample size, which allowed a detailed examination of clinical pathways of patients, and the ability to demonstrate clear time trends by age group, as well as interesting interactions between age and ICU admission on the risk of death. The study was able to look at multiple outcomes of interest, including death, discharge, ICU admission, and time to these events. Detailed analyses are provided in either the main text or the appendix.

As the authors point out, it is likely that temporal changes in patient outcomes were influenced not only by changes in clinical management and treatments over the course of the pandemic, but potentially by surges in the numbers of admissions which may have taxed the system capacity and quality of care at specific points in time. The authors rightly point out that system capacity should be adequately controlled for when assessing the effects of interventions against COVID-19. However, in their own analysis they do not control for any variable that would be an indicator of system capacity, and do not provide any suggestions for how other studies could control for this. Therefore, the analysis in this study by itself does not demonstrate that system capacity had an impact on patient outcomes. The reasons underlying the time trends they observe therefore remain unclear.

While the cohort recruited patients from multiple countries, the vast majority (83%) of patients came from the UK. It is therefore likely that most time trends reflect the evolution of the pandemic and health-seeking patterns in the UK. It is not clear to what extent the time trends are generalizable to other countries who may have had epidemic peaks at different times and that may have different holidays (which are deeply influential on infection patterns and health-seeking behaviors).

It is unclear to what extent selection bias may have influenced the results and observed time trends. In terms of selection into the cohort, there is not much information on how study sites were selected, when they started contributing cases to the cohort, if there were any additional selection criteria applied to participants beyond being a confirmed or suspected COVID-19 case, and what proportion of theoretically eligible participants were recruited into the study. The low number of participants recruited at some study sites suggests that the recruited participants only represent a fraction of all potentially eligible COVID-19 patients at each study site. This may have led to a selection bias if those who were not recruited differed in some systematic way from those included. If recruitment probability varied over time, then this might have also influenced the time trends observed. In terms of selection out of the cohort, there is little information on how many patients were lost to follow-up, had missing outcome data, or were transferred to other facilities. It is possible these patients may have systematic differences from those with complete data, which may lead to selection bias.

Comments for the authors:

1. Please provide date of end of follow-up and an estimate of the average/total follow-up time (STROBE item 14C).

2. Please provide more information regarding participant recruitment, such as numbers potentially eligible and recruited into the study (STROBE item 5,6,13). Were all eligible COVID-19 cases included, or were only some included? Are there more eligibility criteria than those mentioned? The discussion suggests it is the latter (which the authors refer to as "enrolment bias", would recommend rephrasing this as selection bias). If site case loads influenced recruitment then this could potentially lead to selection bias if certain types of patients are favored for study inclusion and this varies over time.

3. Please report in the main text the number (%) of patients lost to follow-up or with missing data for each analysis (STROBE item 13,14).

4. It is unusual to have an outcome of odds or probability of an event where the time frame over which the patient can experience the outcome is not specified. Presumably the outcomes for question 2 and 4 are ICU admission or death over the entire study period. This is not a fair comparison, because patients recruited in later months may not have been followed-up as long as patients recruited earlier, and the follow-up of patients may have varied over the duration of the study. It is generally more rigorous to specify a time frame over which the outcome is assessed and only analyze patients with that follow-up time (ex. death within 3 months of admission) or to use analysis methods which account for differential follow-up time of participants, such as Kaplan-Meier or survival analyses. I would suggest modifying the logistic regressions to have the outcome over a specified time frame or to use survival analyses instead (or to explain why these methods were not used).

5. P39: "Patients lost to follow-up before any of these outcomes were included unless the time to that event was the outcome of interest" It is unclear to me what this means. Does this mean that patients lost to follow-up are included as non-events in the denominator?? It seems to me a better way to treat these patients would be to exclude them or better yet to include them as right-censored observations in survival analyses, as suggested above.

6. Table 4:

– There is an excessive number of decimal places in this table (pseudo-precision) that are not clinically meaningful. I would suggest rounding all values to the nearest decimal (odds ratios) or to the nearest percentage point (percentages). Reducing the number of decimal points will make this table much simpler to read and understandable.

– Please consider dropping the p-values for each individual value. The confidence intervals are much more informative as they indicate the precision around each estimate, convey the same information, and put the emphasis on the estimation of the parameter rather than its statistical significance. The small p-values in most cases are a reflection of the large sample size.

– Conversely, the addition of p-values for the Wald type III test for the overall effect of each variable to the table would be informative.

– The above suggestions (decreasing decimal places and dropping p-values) should create more space in the table. Please consider using this space to add descriptive data on the probability of experiencing death and the number days to death or discharge for each category. This descriptive information helps to contextualize the odds ratios and relative time increases, and may be more clinically useful for some readers than the ratio measures.

– It is not clear whether the results from this table present univariate or multivariable regression model results (presumably the latter). If the analyses are multivariable, then all the variables that were included in the model should be described in a footnote, as this influences the interpretation of the results. For example, it is important to know whether the monthly estimates are adjusted for sex and age, as the age and sex of patients varied over time.

– The pregnancy variable is conflated with the effect of sex because presumably all the men where categorized as 0 for this variable. This variable should be treated as an interaction variable with sex (sex + sex*pregnancy) in order to separate the effect of sex from the effect of pregnancy.

7. There is no Table 3.

8. The authors mention system capacity during epidemic surges as a potential contributor to time trends, but did not include any indicator of system capacity in their analysis, so this interpretation remains speculative. Have the authors considered included such indicators as a variable in their multivariable model (ex. number of COVID-19 infections or hospitalizations reported during the day/week of admission in each country)?

Reviewer #2:

This paper reports here the multi-national collaborative effort on establishment of cohort of 142,540 patients hospitalised with COVID-19. The authors confirmed the previous findings of the trend of hospitalization, ICU admission and case fatality ratio. The author not only shared the large cohort data, including UK and other countries, they also did Multivariable logistic or linear regression to investigate factors associated with the main outcomes, including time from symptom onset to hospital admission, probability of ICU/HDU admission, time from hospital admission to ICU/HDU admission, case fatality ratio (CFR) and total length of hospital stay.

Strengths:

This is a large database from ISARIC Clinical Characterisation Group. Based on the large data, the author answered six important questions, especially time to hospital admission, ICU admission, length of stay and case fatality ratio. The findings are valuable, such as health systems may at times be overwhelmed; system capacity may be an important predictor of patient outcome and may supersede other factors such as increasing case management skills and the influence of new therapies.

Weaknesses:

Although the paper does have strengths in principle, the weaknesses of the paper are that these strengths are not directly demonstrated. In particular:

1. In this study, why gastrointestinal symptoms (abdominal pain, diarrhoea, and vomiting) are emphasized?

2. In this study, patients with a recorded hospital admission date before their symptom onset date, were taken to be a nosocomial infection. But patients who had got infection after exposure to virus may present symptoms later after admission. That is ,on the day of admission, the patient may be in the incubation period.

3. Similarly, "patients with nosocomial infections had lower odds of being admitted to ICU/HDU (OR 0.68, 95% CI 0.62-255 0.74)". The underlying reason may be not nosocomial infection, but that these patients had been admitted earlier than those with adverse outcomes.

4. Line 205: "times" should be "time". Similarly, in the lines 209 and 218.

5. For admission to ICU as an example, the comparison between different months means little, as there are many confounding factors for admission to ICU. One and half years have passed, it is time to look forward to improve treatment.

6. Why less elderly patients were admitted to ICU?

7. Why a wide variety of serious or chronic medical conditions were associated with lower odds of ICU admission. Is it because of lower admission of elderly patients?

8. The raw case fatality rate (CFR) was extremely high in this paper. The readers would be interested in the representativeness of this cohort.

9. In this study, the authors confirmed what is known that Age, male and comorbidities are risk for higher mortality. The readers would like to understand why asthma was associated with lower mortality.

10. Discussion: "patients admitted after experiencing symptoms for longer than 1 week were less likely to die." I am afraid that the explanation lacks scientific evidence.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Ten months of temporal variation in the clinical journey of hospitalised patients with COVID-19: an observational cohort" for further consideration by eLife. Your revised article has been evaluated by Miles Davenport as the Senior Editor, and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

1. The authors mention that the time periods over which the probability of ICU admission and death/discharge were measured are implicit in the cutoff times for time to event analyses (13 days and 45 days). This should be made explicit in the manuscript by indicating that the outcomes were ICU admission within 13 days for question 2 and death/discharge within 45 days for question 4.

2. While the authors specify that all analyses are multivariable in the text, this should also be mentioned in the table legends and/or footnotes in order for tables to be self-sufficient.

3. Table 1: because 'nosocomial infection' is no longer defined in the methods, this row should be relabeled as post-admission symptom onset infections

4. There is an error in Table 2, the rows for symptoms at admission are mistakenly labeled as comorbidities and vice-versa.

5. The authors refer to an non-existent supplementary figure 2 in the results. Presumably they mean Figure 2. This should be fixed.

6. Figure 4 is missing a C inset.

7. On line 446 the authors refer to Figure 3 when they should be referring to Figure 4.

8. Table 3 appears to be lacking the standard deviation for time from symptom onset to hospital admission.

https://doi.org/10.7554/eLife.70970.sa1

Author response

Essential revisions:

1. Provide further rationale for the separation of gastrointestinal and common symptoms for analyses.

After consideration, we have removed the gastrointestinal symptoms variable from the regressions. They were a legacy of earlier iterations of this work, and we agree with the reviewer’s implicit query about what they really add. Its removal makes little appreciable difference to the results. The common symptoms score is only used for question 1, and the main observation regarding it is the initially counter-intuitive finding that patients with fewer symptoms are admitted more quickly, which we explain in the discussion (L984-996).

2. There are several indications that the cohort may not be entirely representative of all hospitalised patients with COVID-19 in the study settings, including a surprisingly high case fatality rate and few participants recruited at some study sites. Please provide more discussion regarding the representativeness of the cohort, eligibility criteria, and study participation rates.

Our reported CFR is in line with that from Docherty et al. (http://doi.org/10.1016/S2213-2600(21)00175-2), and, with UK patients from early in the epidemic forming a very large proportion of our total, as well as the elderly profile of this patient population, this is not surprising. We acknowledge that our recruitment will be influenced by the capacity of each site to enrol patients at any given time and to complete their follow-up. However, there are no obvious reason why either enrolment of completion of follow-up should be biased with respect to our outcomes. Patients were prospectively enrolled, at which point all outcomes would be unknown, and final outcomes would be unknown at the point where follow-up is abandoned. We have expanded this part of the limitations paragraph (L1094-1103).

On a related point, we have specified throughout that the case fatality rate is the hospital CFR (hCFR).

3. Please provide date of end of follow-up and an estimate of the average/total follow-up time.

The last date at which it was possible that any patients in this set to receive follow-up was 8 March, some time after any patients with a recorded outcome had died or been discharged. We have added this to the text (L85-86, 517-521). We have also added median and IQR follow-up times for both all patients and patients with censored outcomes (L615-617).

4. Indicate whether all participants were followed-up over equivalent periods of time to assess study outcomes, in order to justify the use of logistic regression rather than survival analysis.

While it was not explicit in the text of the initial submission, these analyses do in fact look at ICU admission within 13 days (question 3) or death or discharge within 45 days (question 5), because we omitted those with longer time periods as having possibly mis-entered data. Patients were also able to be followed up more than 45 days into 2021, so there is no issue with censorship at the end of the study period. This has been clarified in the text (L595-613). The limit for the time to admission analysis was 24 days, but as all patients here were admitted, there is no censorship issue in this case.

On the topic of survival analysis, this was indeed our initial approach. However, as there were two outcomes (death and discharge), this approach required a competing risks model, and we found that treating those two outcomes as independent events that could occur at any time led to CFR estimates that differed considerably to those calculated using a simple stratified analysis of risk groups. We felt that this was due to invalid assumptions about how the outcomes occur in reality, and opted to pursue the current approach instead.

5. Provide more information regarding what variables are being included in the model, and whether Table 4 represents results from a model including all variables or multiple separate models. This influences how readers interpret the variable coefficients, as it is important to know what variables have been adjusted for and whether the ratios should be interpreted as crude or adjusted associations, and whether interactions are being accounted for in the estimated effects.

All the regressions presented here are multivariable, and included all the listed covariates. We have specified this explicitly in the Results section (L679-680).

6. While less essential, the reviewers made several other suggestions/queries that should be considered by the authors in their revision.

In response to points from the individual reviews, we have changed the following:

– Table 3 (previously and erroneously 4) has been revised to cut down decimal places and include a separate column giving Wald p-values for the inclusion of each variable.

– All the supplementary regression tables (S4-S7) have been adjusted in the same way, although we have kept the greater precision in estimates in the supplement.

– The regression models were also changed such that pregnancy included as an interaction term with sex

– Reviewer 1 asked for model predictions in table 3. We have calculated these for a set of hypothetical patients with typical comorbidities, but present it as a separate figure (new figure 3) instead.

– We agree that it is possible that some patients that we previously classified as having nosocomial infections did, in fact, get infected in the community and were subsequently admitted for a different medical condition before COVID-19 symptoms appeared. As such, we have changed how this is represented in the text (L87-89).

– We have included some discussion and references on the apparently protective effect of asthma (L1053-1059).

– Our interpretation of the low ICU/HDU admission rates in the elderly is that they are the result of clinical judgment and/or patient preference (L1003-1005). This would also apply to comorbidities, which had effects that did not disappear once age was controlled for; we have added words in this discussion to reflect this (L1005-1006).

– The text “Patients lost to follow-up before any of these outcomes were included unless the time to that event was the outcome of interest” merely indicated that, for example, a patient lost to follow-up before final outcome would not be included in the analysis of time to outcome, but would be included in the time to ICU analysis if that information was available. This has been clarified in the text (L95-97).

– We have added an additional discussion of the difficulty in establishing causality, using the reduced odds of ICU/HDU admission in patients admitted before symptom onset as an example. (L1055-1063)

– The statement “patients admitted after experiencing symptoms for longer than 1 week were less likely to die” is justified by our results, as the CFR for death was lower in that group (see table 3). We have, however, clarified this text somewhat, as it was intended as part of our explanation as to how the relationship between longer time to admission and better outcomes might work (L968-970).

– The number of patients included in each analysis, and the proportion this forms of the total, are included at the start of each section.

Reviewer #1:

[…] 1. Please provide date of end of follow-up and an estimate of the average/total follow-up time (STROBE item 14C).

These are now provided (see comments to editor above).

2. Please provide more information regarding participant recruitment, such as numbers potentially eligible and recruited into the study (STROBE item 5,6,13). Were all eligible COVID-19 cases included, or were only some included? Are there more eligibility criteria than those mentioned? The discussion suggests it is the latter (which the authors refer to as "enrolment bias", would recommend rephrasing this as selection bias). If site case loads influenced recruitment then this could potentially lead to selection bias if certain types of patients are favored for study inclusion and this varies over time.

The decision to enrol patients was taken by sites according to their capacity to do so at the time. We are not, unfortunately, privy to those decisions. There is likely some degree of selection bias involved at the individual site level, but it is not clear that this would behave in a consistent manner between different sites. Furthermore, all enrolments were done prospectively and staff would be blind to our outcomes of interest. See lines 955-964 for expanded discussion of these issues.

3. Please report in the main text the number (%) of patients lost to follow-up or with missing data for each analysis (STROBE item 13,14).

These are now provided (see comments to editor above).

4. It is unusual to have an outcome of odds or probability of an event where the time frame over which the patient can experience the outcome is not specified. Presumably the outcomes for question 2 and 4 are ICU admission or death over the entire study period. This is not a fair comparison, because patients recruited in later months may not have been followed-up as long as patients recruited earlier, and the follow-up of patients may have varied over the duration of the study. It is generally more rigorous to specify a time frame over which the outcome is assessed and only analyze patients with that follow-up time (ex. death within 3 months of admission) or to use analysis methods which account for differential follow-up time of participants, such as Kaplan-Meier or survival analyses. I would suggest modifying the logistic regressions to have the outcome over a specified time frame or to use survival analyses instead (or to explain why these methods were not used).

See comments to the editor above.

5. P39: "Patients lost to follow-up before any of these outcomes were included unless the time to that event was the outcome of interest" It is unclear to me what this means. Does this mean that patients lost to follow-up are included as non-events in the denominator?? It seems to me a better way to treat these patients would be to exclude them or better yet to include them as right-censored observations in survival analyses, as suggested above.

Patients lost to follow-up were indeed excluded; this text has been clarified to indicate what was intended (L95-97).

6. Table 4:

– There is an excessive number of decimal places in this table (pseudo-precision) that are not clinically meaningful. I would suggest rounding all values to the nearest decimal (odds ratios) or to the nearest percentage point (percentages). Reducing the number of decimal points will make this table much simpler to read and understandable.

We have reduced this table to one decimal place for the main text version. The supplementary regression tables retain more precision.

– Please consider dropping the p-values for each individual value. The confidence intervals are much more informative as they indicate the precision around each estimate, convey the same information, and put the emphasis on the estimation of the parameter rather than its statistical significance. The small p-values in most cases are a reflection of the large sample size.

We agree and have removed all individual coefficient p-values.

– Conversely, the addition of p-values for the Wald type III test for the overall effect of each variable to the table would be informative.

These have been added.

– The above suggestions (decreasing decimal places and dropping p-values) should create more space in the table. Please consider using this space to add descriptive data on the probability of experiencing death and the number days to death or discharge for each category. This descriptive information helps to contextualize the odds ratios and relative time increases, and may be more clinically useful for some readers than the ratio measures.

We have chosen to add this information as a new figure instead (figure 3).

– It is not clear whether the results from this table present univariate or multivariable regression model results (presumably the latter). If the analyses are multivariable, then all the variables that were included in the model should be described in a footnote, as this influences the interpretation of the results. For example, it is important to know whether the monthly estimates are adjusted for sex and age, as the age and sex of patients varied over time.

All regression results are multivariable; this has been explicitly specified (L689-690).

– The pregnancy variable is conflated with the effect of sex because presumably all the men where categorized as 0 for this variable. This variable should be treated as an interaction variable with sex (sex + sex*pregnancy) in order to separate the effect of sex from the effect of pregnancy.

This adjustment to the relevant regression models has been done.

7. There is no Table 3.

We apologise for this typo; it has been corrected.

8. The authors mention system capacity during epidemic surges as a potential contributor to time trends, but did not include any indicator of system capacity in their analysis, so this interpretation remains speculative. Have the authors considered included such indicators as a variable in their multivariable model (ex. number of COVID-19 infections or hospitalizations reported during the day/week of admission in each country)?

Unfortunately, we did not have access to information of this sort on a level comprehensive enough to use it. We have acknowledged this as a limitation (L1087-1089).

Reviewer #2:

[…] Although the paper does have strengths in principle, the weaknesses of the paper are that these strengths are not directly demonstrated. In particular:

1. In this study, why gastrointestinal symptoms (abdominal pain, diarrhoea, and vomiting) are emphasized?

We have chosen to remove these symptoms as variables in all the regression models (see comments to editor above).

2. In this study, patients with a recorded hospital admission date before their symptom onset date, were taken to be a nosocomial infection. But patients who had got infection after exposure to virus may present symptoms later after admission. That is ,on the day of admission, the patient may be in the incubation period.

We agree that a minority of these individuals were probably not genuine nosocomial infections, and have changed how we refer to them throughout.

3. Similarly, "patients with nosocomial infections had lower odds of being admitted to ICU/HDU (OR 0.68, 95% CI 0.62-255 0.74)". The underlying reason may be not nosocomial infection, but that these patients had been admitted earlier than those with adverse outcomes.

We concur that this is not necessarily a result of a difference in the severity of nosocomial infections, and that causality is difficult to establish; we now discuss this example (L1074-1082).

4. Line 205: "times" should be "time". Similarly, in the lines 209 and 218.

This text has been adjusted.

5. For admission to ICU as an example, the comparison between different months means little, as there are many confounding factors for admission to ICU. One and half years have passed, it is time to look forward to improve treatment.

We agree that using numbers calculated from some time in the past is of limited use in guiding future decisions. Indeed, this is one of our key arguments; that conditions do change during a pandemic and this needs to be accounted for in planning and clinical trials.

6. Why less elderly patients were admitted to ICU?

We hypothesise that this was due to both clinical prognosis and patient preference (L1022-1024)

7. Why a wide variety of serious or chronic medical conditions were associated with lower odds of ICU admission. Is it because of lower admission of elderly patients?

These effects persisted once age was accounted for, but we suspect the reasons were similar. We have now stated this explicitly (L1024-1025).

8. The raw case fatality rate (CFR) was extremely high in this paper. The readers would be interested in the representativeness of this cohort.

See the discussion in the comments to the editor.

9. In this study, the authors confirmed what is known that Age, male and comorbidities are risk for higher mortality. The readers would like to understand why asthma was associated with lower mortality.

We now discuss the asthma findings in the context of the literature (L1053-1059).

10. Discussion: "patients admitted after experiencing symptoms for longer than 1 week were less likely to die." I am afraid that the explanation lacks scientific evidence.

We did find lower CFRs in patients waiting for a longer period before admission; we suspect this is due to differences in patient behaviour in response to their symptoms. We do, however, accept that the intended meaning of this passage was not entirely clear, and have reworded it (L987-989).

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

1. The authors mention that the time periods over which the probability of ICU admission and death/discharge were measured are implicit in the cutoff times for time to event analyses (13 days and 45 days). This should be made explicit in the manuscript by indicating that the outcomes were ICU admission within 13 days for question 2 and death/discharge within 45 days for question 4.

2. While the authors specify that all analyses are multivariable in the text, this should also be mentioned in the table legends and/or footnotes in order for tables to be self-sufficient.

3. Table 1: because 'nosocomial infection' is no longer defined in the methods, this row should be relabeled as post-admission symptom onset infections

4. There is an error in Table 2, the rows for symptoms at admission are mistakenly labeled as comorbidities and vice-versa.

5. The authors refer to an non-existent supplementary figure 2 in the results. Presumably they mean Figure 2. This should be fixed.

6. Figure 4 is missing a C inset.

7. On line 446 the authors refer to Figure 3 when they should be referring to Figure 4.

8. Table 3 appears to be lacking the standard deviation for time from symptom onset to hospital admission.

All the requested changes have been made, with the exception that, regarding point 8, the problem was in fact simply a mis-aligned column header.

In addition, in responding to point 1, we realised that we also needed to make a change to the results of question 2. In order for the variable of interest to be ICU/HDU admission within 13 days, we actually needed to exclude patients with less than 13 days of follow-up, no ICU/HDU admission, and unknown outcome. Figure 2A and the relevant regression table have been changed accordingly, with minimal effect on the results.

https://doi.org/10.7554/eLife.70970.sa2

Article and author information

Author details

  1. ISARIC Clinical Characterisation Group

    Contribution
    Conceptualization, Formal analysis, Methodology, Software, Visualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
    1. Sheryl Ann Abdukahil
    2. Ryuzo Abe
    3. Laurent Abel
    4. Lara Absil
    5. Subhash Acharya
    6. Andrew Acker
    7. Shingo Adachi
    8. Elisabeth Adam
    9. Diana Adrião
    10. Saleh Al Ageel
    11. Shakeel Ahmed
    12. Kate Ainscough
    13. Ali Ait Hssain
    14. Younes Ait Tamlihat
    15. Takako Akimoto
    16. Ernita Akmal
    17. Eman Al Qasim
    18. Razi Alalqam
    19. Tala Al-dabbous
    20. Senthilkumar Alegesan
    21. Cynthia Alegre
    22. Beatrice Alex
    23. Kévin Alexandre
    24. Abdulrahman Al-Fares
    25. Huda Alfoudri
    26. Imran Ali
    27. Kazali Enagnon Alidjnou
    28. Jeffrey Aliudin
    29. Qabas Alkhafajee
    30. Clotilde Allavena
    31. Nathalie Allou
    32. João Alves
    33. João Melo Alves
    34. Rita Alves
    35. Maria Amaral
    36. Heidi Ammerlaan
    37. Phoebe Ampaw
    38. Roberto Andini
    39. Claire Andrejak
    40. Andrea Angheben
    41. François Angoulvant
    42. Séverine Ansart
    43. Massimo Antonelli
    44. Carlos Alexandre Antunesde Brito
    45. Ardiyan Apriyana
    46. Yaseen Arabi
    47. Irene Aragao
    48. Francisco Arancibia
    49. Carolline Araujo
    50. Antonio Arcadipane
    51. Patrick Archambault
    52. Lukas Arenz
    53. Jean-Benoît Arlet
    54. Christel Arnold-Day
    55. Lovkesh Arora
    56. Rakesh Arora
    57. Elise Artaud-Macari
    58. Diptesh Aryal
    59. Motohiro Asaki
    60. Angel Asensio
    61. Sheharyar Ashraf
    62. Jean Baptiste Assie
    63. Anika Atique
    64. AMUdara Lakshan Attanyake
    65. Johann Auchabie
    66. Hugues Aumaitre
    67. Adrien Auvet
    68. Laurène Azemar
    69. Cecile Azoulay
    70. Benjamin Bach
    71. Delphine Bachelet
    72. Claudine Badr
    73. Nadia Baig
    74. J Kenneth Baillie
    75. Erica Bak
    76. Agamemnon Bakakos
    77. Andriy Bal
    78. Valeria Balan
    79. Firouzé Bani-Sadr
    80. Renata Barbalho
    81. Wendy S Barclay
    82. Michaela Barnikel
    83. Helena Barrasa
    84. Audrey Barrelet
    85. Cleide Barrigoto
    86. Marie Bartoli
    87. Cheryl Bartone
    88. Joaquín Baruch
    89. Romain Basmaci
    90. Denise Battaglini
    91. Jules Bauer
    92. Diego Fernando Bautista Rincon
    93. Abigail Beane
    94. Alexandra Bedossa
    95. Sylvie Behilill
    96. Aleksandr Beljantsev
    97. David Bellemare
    98. Anna Beltrame
    99. Marine Beluze
    100. Nicolas Benech
    101. Dehbia Benkerrou
    102. Suzanne Bennett
    103. Luís Bento
    104. Jan-Erik Berdal
    105. Delphine Bergeaud
    106. Hazel Bergin
    107. José Luis Bernal Sobrino
    108. Giulia Bertoli
    109. Lorenzo Bertolino
    110. Simon Bessis
    111. Adam Betz
    112. Sybille Bevilcaqua
    113. Karine Bezulier
    114. Amar Bhatt
    115. Krishna Bhavsar
    116. Isabella Bianchi
    117. Claudia Bianco
    118. Moirangthem Bikram Singh
    119. Felwa Bin Humaid
    120. François Bissuel
    121. Patrick Biston
    122. Laurent Bitker
    123. Pablo Blanco-Schweizer
    124. Catherine Blier
    125. Frank Bloos
    126. Mathieu Blot
    127. Filomena Boccia
    128. Laetitia Bodenes
    129. Alice Bogaarts
    130. Debby Bogaert
    131. Anne-Hélène Boivin
    132. Pierre-Adrien Bolze
    133. François Bompart
    134. Diogo Borges
    135. Raphaël Borie
    136. Hans Martin Bosse
    137. Elisabeth Botelho-Nevers
    138. Lila Bouadma
    139. Olivier Bouchaud
    140. Sabelline Bouchez
    141. Dounia Bouhmani
    142. Damien Bouhour
    143. Kévin Bouiller
    144. Laurence Bouillet
    145. Camile Bouisse
    146. Anne-Sophie Boureau
    147. John Bourke
    148. Maude Bouscambert
    149. Aurore Bousquet
    150. Jason Bouziotis
    151. Bianca Boxma
    152. Marielle Boyer-Besseyre
    153. Maria Boylan
    154. Matthew Brack
    155. Axelle Braconnier
    156. Cynthia Braga
    157. Timo Brandenburger
    158. Filipa Brás Monteiro
    159. Luca Brazzi
    160. Dorothy Breen
    161. Patrick Breen
    162. Patrick Breen
    163. Kathy Brickell
    164. Alex Browne
    165. Shaunagh Browne
    166. Nicolas Brozzi
    167. Nina Buchtele
    168. Christian Buesaquillo
    169. Marielle Buisson
    170. Erlina Burhan
    171. Ingrid G Bustos
    172. André Cabie
    173. Susana Cabral
    174. Eder Caceres
    175. Cyril Cadoz
    176. Kate Calligy
    177. Jose Andres Calvache
    178. João Cam
    179. Valentine Campana
    180. Paul Campbell
    181. Cecilia Canepa
    182. Mireia Cantero
    183. Pauline Caraux-Paz
    184. Sheila Cárcel
    185. Chiara Simona Cardellino
    186. Filipa Cardoso
    187. Filipe Cardoso
    188. Nelson Cardoso
    189. Sofia Cardoso
    190. Simone Carelli
    191. Nicolas Carlier
    192. Thierry Carmoi
    193. Gayle Carney
    194. Chloe Carpenter
    195. Inês Carqueja
    196. Marie-Christine Carret
    197. François Martin Carrier
    198. Ida Carroll
    199. Gail Carson
    200. Ed Carton
    201. Maire-Laure Casanova
    202. Mariana Cascão
    203. Siobhan Casey
    204. José Casimiro
    205. Bailey Cassandra
    206. Silvia Castañeda
    207. Nidyanara Castanheira
    208. Guylaine Castor-Alexandre
    209. Henry Castrillón
    210. Ivo Castro
    211. Ana Catarino
    212. François-Xavier Catherine
    213. Paolo Cattaneo
    214. Roberta Cavalin
    215. Giulio Giovanni Cavalli
    216. Alexandros Cavayas
    217. Adrian Ceccato
    218. Minerva Cervantes-Gonzalez
    219. Anissa Chair
    220. Catherine Chakveatze
    221. Adrienne Chan
    222. Meera Chand
    223. Christelle Chantalat Auger
    224. Jean-Marc Chapplain
    225. Julie Chas
    226. Jonathan Samuel Chávez Iñiguez
    227. Anjellica Chen
    228. Yih-Sharng Chen
    229. Matthew Pellan Cheng
    230. Antoine Cheret
    231. Thibault Chiarabini
    232. Julian Chica
    233. Catherine Chirouze
    234. Davide Chiumello
    235. Hwa Jin Cho
    236. Sung-Min Cho
    237. Bernard Cholley
    238. Marie-Charlotte Chopin
    239. JosePedro Cidade
    240. Barbara Wanjiru Citarella
    241. Anna Ciullo
    242. Jennifer Clarke
    243. Sara Clohisey
    244. Necsoi Coca
    245. Cassidy Codan
    246. Caitriona Cody
    247. Alexandra Coelho
    248. Megan Coles
    249. Gwenhaël Colin
    250. Michael Collins
    251. Sebastiano Maria Colombo
    252. Pamela Combs
    253. Jennifer Connolly
    254. Marie Connor
    255. Anne Conrad
    256. Sofía Contreras
    257. Elaine Conway
    258. Graham S Cooke
    259. Mary Copland
    260. Hugues Cordel
    261. Amanda Corley
    262. Sarah Cormican
    263. Sabine Cornelis
    264. Arianne Joy Corpuz
    265. Grégory Corvaisier
    266. Camille Couffignal
    267. Sandrine Couffin-Cadiergues
    268. Roxane Courtois
    269. Stéphanie Cousse
    270. Sabine Croonen
    271. Gloria Crowl
    272. Jonathan Crump
    273. Claudina Cruz
    274. Juan Luis Cruz Berm
    275. Jaime Cruz Rojo
    276. Marc Csete
    277. Alberto Cucino
    278. Ailbhe Cullen
    279. Caroline Cullen
    280. Matthew Cummings
    281. Ger Curley
    282. Elodie Curlier
    283. Colleen Curran
    284. Paula Custodio
    285. Anada Silva Filipe
    286. Charlene Da Silveira
    287. Al-Awwab Dabaliz
    288. Andrew Dagens
    289. Darren Dahly
    290. Heidi Dalton
    291. Jo Dalton
    292. Seamus Daly
    293. Federico D'Amico
    294. Nick Daneman
    295. Corinne Daniel
    296. Emmanuelle A Dankwa
    297. Jorge Dantas
    298. Frédérick D'Aragon
    299. Markde Boer
    300. Gilliande Loughry
    301. Diegode Mendoza
    302. Etienne De Montmollin
    303. Rafael Freitas de Oliveira França
    304. Ana Isabel de Pinho Oliveira
    305. De Rosanna Rosa
    306. Thushande Silva
    307. Peterde Vries
    308. Jillian Deacon
    309. David Dean
    310. Alexa Debard
    311. Bianca De Benedictis
    312. Marie-Pierre Debray
    313. Nathalie De Castro
    314. William Dechert
    315. Lauren Deconninck
    316. Romain Decours
    317. Eve Defous
    318. Isabelle Delacroix
    319. Eric Delaveuve
    320. Karen Delavigne
    321. Nathalie M Delfos
    322. Ionna Deligiannis
    323. Andrea Dell'Amore
    324. Christelle Delmas
    325. Pierre Delobel
    326. Elisa Demonchy
    327. Emmanuelle Denis
    328. Dominique Deplanque
    329. Pieter Depuydt
    330. Mehul Desai
    331. Diane Descamps
    332. Mathilde Desvallée
    333. Santi Dewayanti
    334. Alpha Diallo
    335. Sylvain Diamantis
    336. André Dias
    337. Juan Jose Diaz
    338. Priscila Diaz
    339. Rodrigo Diaz
    340. Kévin Didier
    341. Jean-Luc Diehl
    342. Wim Dieperink
    343. Jérôme Dimet
    344. Vincent Dinot
    345. Fara Diop
    346. Alphonsine Diouf
    347. Yael Dishon
    348. Félix Djossou
    349. Annemarie B Docherty
    350. Helen Doherty
    351. Arjen M Dondorp
    352. Christl A Donnelly
    353. Maria Donnelly
    354. Chloe Donohue
    355. Sean Donohue
    356. Yoann Donohue
    357. Ciara Doran
    358. Peter Doran
    359. Céline Dorival
    360. Eric D'Ortenzio
    361. James Joshua Douglas
    362. Renee Douma
    363. Nathalie Dournon
    364. Triona Downer
    365. Joanne Downey
    366. Mark Downing
    367. Tom Drake
    368. Aoife Driscoll
    369. Claudio Duarte Fonseca
    370. Vincent Dubee
    371. François Dubos
    372. Alexandre Ducancelle
    373. Toni Duculan
    374. Susanne Dudman
    375. Abhijit Duggal
    376. Paul Dunand
    377. Jake Dunning
    378. Mathilde Duplaix
    379. Emanuele Durante-Mangoni
    380. Bertrand Dussol Lucian Durham III
    381. Juliette Duthoit
    382. Xavier Duval
    383. Anne Margarita Dyrhol-Riise
    384. Marco Echeverria-Villalobos
    385. Siobhan Egan
    386. Carla Eira
    387. Mohammed El Sanharawi
    388. Subbarao Elapavaluru
    389. Brigitte Elharrar
    390. Jacobien Ellerbroek
    391. Philippine Eloy
    392. Tarek Elshazly
    393. Isabelle Enderle
    394. Ilka Engelmann
    395. Vincent Enouf
    396. Olivier Epaulard
    397. Martina Escher
    398. Mariano Esperatti
    399. Hélène Esperou
    400. Marina Esposito-Farese
    401. João Estevão
    402. Manuel Etienne
    403. Nadia Ettalhaoui
    404. Anna Greti Everding
    405. Mirjam Evers
    406. Isabelle Fabre
    407. Marc Fabre
    408. Amna Faheem
    409. Arabella Fahy
    410. Cameron J Fairfield
    411. Pedro Faria
    412. Nataly Farshait
    413. Arie Zainul Fatoni
    414. Karine Faure
    415. Raphaël Favory
    416. Mohamed Fayed
    417. Niamh Feely
    418. Laura Feeney
    419. Jorge Fernandes
    420. Marília Fernandes
    421. Susana Fernandes
    422. François-Xavier Ferrand
    423. Eglantine Ferrand Devouge
    424. Joana Ferrão
    425. Mário Ferraz
    426. Benigno Ferreira
    427. Sílvia Ferreira
    428. Ricard Ferrer-Roca
    429. Nicolas Ferriere
    430. Céline Ficko
    431. Claudia Figueiredo-Mello
    432. Juan Fiorda
    433. Thomas Flament
    434. Clara Flateau
    435. Tom Fletcher
    436. Letizia Lucia Florio
    437. Brigid Flynn
    438. Deirdre Flynn
    439. Claire Foley
    440. Jean Foley
    441. Tatiana Fonseca
    442. Patricia Fontela
    443. Simon Forsyth
    444. Denise Foster
    445. Giuseppe Foti
    446. Erwan Fourn
    447. Robert A Fowler
    448. Dr Marianne Fraher
    449. Diego Franch-Llasat
    450. Christophe Fraser
    451. John F Fraser
    452. Marcela Vieira Freire
    453. Ana Freitas Ribeiro
    454. Caren Friedrich
    455. Ricardo Fritz
    456. Stéphanie Fry
    457. Nora Fuentes
    458. Masahiro Fukuda
    459. Valérie Gaborieau
    460. Rostane Gaci
    461. Massimo Gagliardi
    462. Jean-Charles Gagnard
    463. Amandine Gagneux-Brunon
    464. Sérgio Gaião
    465. Linda Gail Skeie
    466. Phil Gallagher
    467. Elena Gallego Curto
    468. Carrol Gamble
    469. Arthur Garan
    470. Rebekha Garcia
    471. Noelia García Barrio
    472. Esteban Garcia-Gallo
    473. Denis Garot
    474. Valérie Garrait
    475. Nathalie Gault
    476. Aisling Gavin
    477. Anatoliy Gavrylov
    478. Alexandre Gaymard
    479. Johannes Gebauer
    480. Eva Geraud
    481. Louis Gerbaud Morlaes
    482. Nuno Germano
    483. Jade Ghosn
    484. Marco Giani
    485. Carlo Giaquinto
    486. Jess Gibson
    487. Tristan Gigante
    488. Morgane Gilg
    489. Elaine Gilroy
    490. Guillermo Giordano
    491. Michelle Girvan
    492. Valérie Gissot
    493. Judy Gitahi
    494. Gezy Giwangkancana
    495. Daniel Glikman
    496. Eric Gnall
    497. Geraldine Goco
    498. François Goehringer
    499. Siri Goepel
    500. Jean-Christophe Goffard
    501. Jonathan Golob
    502. Rui Gomes
    503. Joan Gómez-Junyent
    504. Marie Gominet
    505. Alicia Gonzalez
    506. Patricia Gordon
    507. Isabelle Gorenne
    508. Laure Goubert
    509. Cécile Goujard
    510. Tiphaine Goulenok
    511. Margarite Grable
    512. Jeronimo Graf
    513. Edward Wilson Grandin
    514. Pascal Granier
    515. Giacomo Grasselli
    516. Lorenzo Grazioli
    517. Christopher A Green
    518. William Greenhalf
    519. Segolène Greffe
    520. Domenico Luca Grieco
    521. Matthew Griffee
    522. Fiona Griffiths
    523. Ioana Grigoras
    524. Albert Groenendijk
    525. Anja Grosse Lordemann
    526. Heidi Gruner
    527. Yusing Gu
    528. Fabio Guarracino
    529. Jérémie Guedj
    530. Martin Guego
    531. Dewi Guellec
    532. Anne-Marie Guerguerian
    533. Daniela Guerreiro
    534. Romain Guery
    535. Anne Guillaumot
    536. Laurent Guilleminault
    537. Maisa Guimarães de Castro
    538. Thomas Guimard
    539. Daniel Haber
    540. Hannah Habraken
    541. Ali Hachemi
    542. Nadir Hadri
    543. Olena Haidash
    544. Sheeba Hakak
    545. Adam Hall
    546. Matthew Hall
    547. Sophie Halpin
    548. Ansley Hamer
    549. Rebecca Hamidfar
    550. Terese Hammond
    551. Rashan Haniffa
    552. Hayley Hardwick
    553. Ewen M Harrison
    554. Janet Harrison
    555. Samuel Bernard Ekow Harrison
    556. Alan Hartman
    557. Muhammad Hayat
    558. Ailbhe Hayes
    559. Leanne Hays
    560. Jan Heerman
    561. Lars Heggelund
    562. Ross Hendry
    563. Martina Hennessy
    564. Aquiles Henriquez-Trujillo
    565. Maxime Hentzien
    566. Jaime Hernandez-Montfort
    567. Daniel Herr
    568. Andrew Hershey
    569. Liv Hesstvedt
    570. Astarini Hidayah
    571. Dawn Higgins
    572. Eibhilin Higgins
    573. Grainne Higgins OKeeffe
    574. Rita Hinchion
    575. Samuel Hinton
    576. Hiroaki Hiraiwa
    577. Hikombo Hitoto
    578. Antonia Ho
    579. Alexandre Hoctin
    580. Isabelle Hoffmann
    581. Oscar Hoiting
    582. Rebecca Holt
    583. Jan Cato Holter
    584. Peter Horby
    585. Juan Pablo Horcajada
    586. Koji Hoshino
    587. Kota Hoshino
    588. Ikram Houas
    589. Catherine L Hough
    590. Stuart Houltham
    591. Jimmy Ming-Yang Hsu
    592. Jean-Sébastien Hulot
    593. Samreen Ijaz
    594. Hajnal-Gabriela Illes
    595. Patrick Imbert
    596. Hugo Inácio
    597. Elias Iosifidis
    598. Sarah Isgett
    599. Palliya Guruge Pramodya Ishani Ishani
    600. Tiago Isidoro
    601. Margaux Isnard
    602. Junji Itai
    603. Daniel Ivulich
    604. Danielle Jaafar
    605. Salma Jaafoura
    606. Julien Jabot
    607. Clare Jackson
    608. Nina Jamieson
    609. Pierre Jaquet
    610. Coline Jaud-Fischer
    611. Stéphane Jaureguiberry
    612. Issrah Jawad
    613. Florence Jego
    614. Synne Jenum
    615. Ruth Jimbo-Sotomayor
    616. N Ruth
    617. Jorge García
    618. Cédric Joseph
    619. Mark Joseph
    620. Swosti Joshi
    621. Mercé Jourdain
    622. Philippe Jouvet
    623. Jennifer June
    624. Anna Jung
    625. Hanna Jung
    626. Dafsah Juzar
    627. Ouifiya Kafif
    628. Florentia Kaguelidou
    629. Sabina Kali
    630. Smaragdi Kalomoiri
    631. Darshana Hewa Kandamby
    632. Chris Kandel
    633. Ravi Kant
    634. Dyah Kanyawati
    635. Christiana Kartsonaki
    636. Daisuke Kasugai
    637. Anant Kataria
    638. Kevin Katz
    639. Simreen Kaur Johal
    640. Christy Kay
    641. Hannah Keane
    642. Seán Keating
    643. Andrea Kelly
    644. Aoife Kelly
    645. Claire Kelly
    646. Niamh Kelly
    647. Sadie Kelly
    648. Yvelynne Kelly
    649. Maeve Kelsey
    650. Ryan Kennedy
    651. Kalynn Kennon
    652. Maeve Kernan
    653. Younes Kerroumi
    654. Sharma Keshav
    655. Evelyne Kestelyn
    656. Imrana Khalid
    657. Antoine Khalil
    658. Coralie Khan
    659. Irfan Khan
    660. Michelle E Kho
    661. Saye Khoo
    662. Yuri Kida
    663. Peter Kiiza
    664. Anders Benjamin Kildal
    665. Jae Burm Kim
    666. Antoine Kimmoun
    667. Detlef Kindgen-Milles
    668. Alexander King
    669. Nobuya Kitamura
    670. Paul Klenerman
    671. Gry Kloumann Bekken
    672. Stephen Knight
    673. Robin Kobbe
    674. Chamira Kodippily
    675. Malte Kohns Vasconcelos
    676. Volkan Korten
    677. Caroline Kosgei
    678. Arsène Kpangon
    679. Karolina Krawczyk
    680. Sudhir Krishnan
    681. Oksana Kruglova
    682. Deepali Kumar
    683. Bharath Kumar Tirupakuzhi Vijayaraghavan
    684. Pavan Kumar Vecham
    685. Ethan Kurtzman
    686. Neurinda Permata Kusumastuti
    687. Demetrios Kutsogiannis
    688. Galyna Kutsyna
    689. Konstantinos Kyriakoulis
    690. Marie Lachatre
    691. Marie Lacoste
    692. John G Laffey
    693. Marie Lagrange
    694. Fabrice Laine
    695. Olivier Lairez
    696. Antonio Lalueza
    697. Marc Lambert
    698. François Lamontagne
    699. Marie Langelot-Richard
    700. Vincent Langlois
    701. Eka Yudha Lantang
    702. Marina Lanza
    703. Cédric Laouénan
    704. Samira Laribi
    705. Delphine Lariviere
    706. Stéphane Lasry
    707. Odile Launay
    708. Didier Laureillard
    709. Yoan Lavie-Badie
    710. Andrew Law
    711. Cassie Lawrence
    712. Teresa Lawrence
    713. Minh Le
    714. Clément Le Bihan
    715. Cyril Le Bris
    716. Georges Le Falher
    717. Lucie Le Fevre
    718. Quentin Le Hingrat
    719. Marion Le Maréchal
    720. Soizic Le Mestre
    721. Gwenaël Le Moal
    722. Vincent Le Moing
    723. Paul Le Turnier Hervé Le Nagard
    724. Ema Leal
    725. Marta Leal Santos
    726. James Lee
    727. Su Hwan Lee
    728. Todd C Lee
    729. Gary Leeming
    730. Bénédicte Lefebvre
    731. Laurent Lefebvre
    732. Benjamin Lefevre
    733. Sylvie Le Gac
    734. Jean-Daniel Lelievre
    735. François Lellouche
    736. Adrien Lemaignen
    737. Véronique Lemee
    738. Anthony Lemeur
    739. Gretchen Lemmink
    740. Rafael León
    741. Marc Leone
    742. Michela Leone
    743. François-Xavier Lescure
    744. Olivier Lesens
    745. Mathieu Lesouhaitier
    746. Amy Lester-Grant
    747. Bruno Levy
    748. Yves Levy
    749. Claire Levy-Marchal
    750. Erwan L'Her
    751. Gianluigi Li Bassi
    752. Janet Liang
    753. Geoffrey Liegeon
    754. Wei Shen Lim
    755. Chantre Lima
    756. Bruno Lina
    757. Andreas Lind
    758. Guillaume Lingas
    759. Sylvie Lion-Daolio
    760. Keibun Liu
    761. Marine Livrozet
    762. Patricia Lizotte
    763. Antonio Loforte
    764. Navy Lolong
    765. Diogo Lopes
    766. Dalia Lopez-Colon
    767. Anthony L Loschner
    768. Paul Loubet
    769. Bouchra Loufti
    770. Guillame Louis
    771. Silvia Lourenco
    772. Jean Christophe Lucet
    773. Carlos Lumbreras Bermejo
    774. Carlos M Luna
    775. Olguta Lungu
    776. Liem Luong
    777. Nestor Luque
    778. Dominique Luton
    779. Ruth Lyons
    780. Olavi Maasikas
    781. Oryane Mabiala
    782. Samual Mac Donald
    783. Moïse Machado
    784. Gabriel Macheda
    785. Juan Macias Sanchez
    786. Jai Madhok
    787. Hashmi Madiha
    788. Guillermo Maestro de la Calle
    789. Rafael Mahieu
    790. Sophie Mahy
    791. Ana Raquel Maia
    792. Lars S Maier
    793. Mylène Maillet
    794. Thomas Maitre
    795. Maximilian Malfertheiner
    796. Nadia Malik
    797. Paddy Mallon
    798. Fernando Maltez
    799. Denis Malvy
    800. Victoria Manda
    801. Jose M Mandei
    802. Laurent Mandelbrot
    803. Frank Manetta
    804. Julie Mankikian
    805. Edmund Manning
    806. Aldric Manuel
    807. Ceila Maria Sant'Ana Málaque
    808. Daniel Marino
    809. Flávio Marino
    810. Samuel Markowicz
    811. Ana Marques
    812. Catherine Marquis
    813. Brian Marsh
    814. Laura Marsh
    815. Megan Marshal
    816. John Marshall
    817. Celina Turchi Martelli
    818. Emily Martin
    819. Guillaume Martin-Blondel
    820. Alessandra Martinelli
    821. Ignacio Martin-Loeches
    822. Martin Martinot
    823. Ana Martins
    824. João Martins
    825. Nuno Martins
    826. Caroline Martins Rego
    827. Gennaro Martucci
    828. Olga Martynenko
    829. Eva Miranda Marwali
    830. Juan Fernado Masa Jimenez
    831. David Maslove
    832. David Maslove
    833. Sabina Mason
    834. Basri Mat Nor
    835. Moshe Matan
    836. Daniel Mathieu
    837. Mathieu Mattei
    838. Romans Matulevics
    839. Laurence Maulin
    840. Michael Maxwell
    841. Javier Maynar
    842. Thierry Mazzoni
    843. Natalie Mc Evoy
    844. Lisa Mc Sweeney
    845. Lorraine McAndrew
    846. Colin McArthur
    847. Aine McCarthy
    848. Anne McCarthy
    849. Colin McCloskey
    850. Rachael McConnochie
    851. Sherry McDermott
    852. Sarah E McDonald
    853. Aine McElroy
    854. Samuel McElwee
    855. Victoria McEneany
    856. Allison McGeer
    857. Chris McKay
    858. Johnny McKeown
    859. Kenneth A McLean
    860. Bairbre McNicholas
    861. Elaine McPartlan
    862. Edel Meaney
    863. Cécile Mear-Passard
    864. Maggie Mechlin
    865. Omar Mehkri
    866. Ferruccio Mele
    867. Luis Melo
    868. Joao Joao Mendes
    869. Ogechukwu Menkiti
    870. Kusum Menon
    871. France Mentré
    872. Alexander J Mentzer
    873. Emmanuelle Mercier
    874. Noémie Mercier
    875. Antoine Merckx
    876. Mayka Mergeay-Fabre
    877. Blake Mergler
    878. Laura Merson
    879. António Mesquita
    880. Osama Metwally
    881. Agnès Meybeck
    882. Dan Meyer
    883. Alison M Meynert
    884. Vanina Meysonnier
    885. Amina Meziane
    886. Mehdi Mezidi
    887. Céline Michelanglei
    888. Isabelle Michelet
    889. Efstathia Mihelis
    890. Vladislav Mihnovit
    891. Hugo Miranda-Maldonado
    892. Asma Moin
    893. David Molina
    894. Elena Molinos
    895. Brenda Molloy
    896. Mary Mone
    897. Agostinho Monteiro
    898. Claudia Montes
    899. Giorgia Montrucchio
    900. Sarah Moore
    901. Shona C Moore
    902. Lina Morales Cely
    903. Lucia Moro
    904. Catherine Motherway
    905. Ana Motos
    906. Hugo Mouquet
    907. Clara Mouton Perrot
    908. Julien Moyet
    909. Jimmy Mullaert
    910. Fredrik Muller
    911. Karl Erik Müller
    912. Aisling Murphy
    913. Aisling Murphy
    914. Lorna Murphy
    915. Marlène Murris
    916. Srinivas Murthy
    917. Himed Musaab
    918. Dimitra Melia Myrodia
    919. Dave Nagpal
    920. Alex Nagrebetsky
    921. Mangala Narasimhan
    922. Nadège Neant
    923. Holger Neb
    924. Raul Neto
    925. Emily Neumann
    926. Bernardo Neves
    927. Pauline Yeung Ng
    928. Anthony Nghi
    929. Duc Nguyen
    930. Orna Ni Choileain
    931. Niamh Ni Leathlobhair
    932. Alistair Nichol
    933. Prompak Nitayavardhana
    934. Stephanie Nonas
    935. Marion Noret
    936. Lisa Norman
    937. Alessandra Notari
    938. Mahdad Noursadeghi
    939. Karolina Nowicka
    940. Saad Nseir
    941. Jose I Nunez
    942. Nurnaningsih Nurnaningsih
    943. Elsa Nyamankolly
    944. Fionnuala O Brien
    945. Annmarie O Callaghan
    946. Annmarie O'Callaghan
    947. Giovanna Occhipinti
    948. Derbrenn OConnor
    949. Max O'Donnell
    950. Tawnya Ogston
    951. Takayuki Ogura
    952. Tak-Hyuk Oh
    953. Sophie O'Halloran
    954. Katie O'Hearn
    955. Shinichiro Ohshimo
    956. Agnieszka Oldakowska
    957. João Oliveira
    958. Larissa Oliveira
    959. Piero L Olliaro
    960. David SY Ong
    961. Wilna Oosthuyzen
    962. Anne Opavsky
    963. Peter Openshaw
    964. Claudia Milena Orozco-Chamorro
    965. Jamel Ortoleva
    966. Javier Osatnik
    967. Linda O'Shea
    968. Miriam O'Sullivan
    969. Nadia Ouamara
    970. Rachida Ouissa
    971. Clark Owyang
    972. Eric Oziol
    973. HM Upulee Pabasara
    974. Justine Pages Maïder Pagadoy
    975. Amanda Palacios
    976. Mario Palacios
    977. Massimo Palmarini
    978. Giovanna Panarello
    979. Prasan Kumar Panda
    980. Mauro Panigada
    981. Nathalie Pansu
    982. Aurélie Papadopoulos
    983. Melissa Parker
    984. Briseida Parra
    985. Jérémie Pasquier
    986. Bruno Pastene
    987. Fabian Patauner
    988. Luís Patrão
    989. Patricia Patricio
    990. Juliette Patrier
    991. Lisa Patterson
    992. Christelle Paul
    993. Mical Paul
    994. Jorge Paulos
    995. William A Paxton
    996. Jean-François Payen
    997. Miguel Pedrera Jiménez
    998. Giles J Peek
    999. Florent Peelman
    1000. Nathan Peiffer-Smadja
    1001. Vincent Peigne
    1002. Mare Pejkovska
    1003. Paolo Pelosi
    1004. Ithan D Peltan
    1005. Rui Pereira
    1006. Daniel Perez
    1007. Luis Periel
    1008. Thomas Perpoint
    1009. Antonio Pesenti
    1010. Vincent Pestre
    1011. Lenka Petrou
    1012. Ventzislava Petrov-Sanchez
    1013. Frank Olav Pettersen
    1014. Gilles Peytavin
    1015. Scott Pharand
    1016. Michael Piagnerelli
    1017. Walter Picard
    1018. Olivier Picone
    1019. Mariade Piero
    1020. Carola Pierobon
    1021. Carlos Pimentel
    1022. Raquel Pinto
    1023. Catarina Pires
    1024. Isabelle Pironneau
    1025. Lionel Piroth
    1026. Riinu Pius
    1027. Simone Piva
    1028. Laurent Plantier
    1029. Daniel Plotkin
    1030. Julien Poissy
    1031. Ryadh Pokeerbux
    1032. Maria Pokorska-Spiewak
    1033. Sergio Poli
    1034. Georgios Pollakis
    1035. Diane Ponscarme
    1036. Jolanta Popielska
    1037. Andra-Maris Post
    1038. Douwe F Postma
    1039. Pedro Povoa
    1040. Diana Póvoas
    1041. Jeff Powis
    1042. Sofia Prapa
    1043. Sébastien Preau
    1044. Christian Prebensen
    1045. Jean-Charles Preiser
    1046. Anton Prinssen
    1047. Mark G Pritchard
    1048. Gamage Dona Dilanthi Priyadarshani
    1049. Lucia Proença
    1050. Sravya Pudota
    1051. Oriane Puéchal
    1052. Bambang Pujo Semedi
    1053. Gregory Purcell
    1054. Luisa Quesada
    1055. Vilmaris Quinones-Cardona
    1056. Víctor Quirós González
    1057. Else Quist-Paulsen
    1058. Mohammed Quraishi
    1059. Christian Rabaud
    1060. Aldo Rafael
    1061. Marie Rafiq
    1062. Fernando Rainieri
    1063. Nagarajan Ramakrishnan
    1064. Blandine Rammaert
    1065. Ritika Ranjan
    1066. Christophe Rapp
    1067. Aasiyah Rashan
    1068. Thalha Rashan
    1069. Menaldi Rasmin
    1070. Indrek Rätsep
    1071. Cornelius Rau
    1072. Ali Raza
    1073. Andre Real
    1074. Stanislas Rebaudet
    1075. Sarah Redl
    1076. Brenda Reeve
    1077. Liadain Reid
    1078. Liadain Reid
    1079. Dag Henrik Reikvam
    1080. Renato Reis
    1081. Jordi Rello
    1082. Jonathan Remppis
    1083. Martine Remy
    1084. Hongru Ren
    1085. Hanna Renk
    1086. Liliana Resende
    1087. Anne-Sophie Resseguier
    1088. Matthieu Revest
    1089. Oleksa Rewa
    1090. Luis Felipe Reyes
    1091. Tiago Reyes
    1092. Maria Ines Ribeiro
    1093. David Richardson
    1094. Denise Richardson
    1095. Laurent Richier
    1096. Jordi Riera
    1097. Ana L Rios
    1098. Asgar Rishu
    1099. Patrick Rispal
    1100. Karine Risso
    1101. Nicholas Rizer
    1102. Chiara Robba
    1103. André Roberto
    1104. Stephanie Roberts
    1105. David L Robertson
    1106. Olivier Robineau
    1107. Ferran Roche-Campo
    1108. Paola Rodari
    1109. Simão Rodeia
    1110. Julia Rodriguez Abreu
    1111. Bernhard Roessler
    1112. Pierre-Marie Roger
    1113. Emmanuel Roilideseira
    1114. Amanda Rojek
    1115. Juliette Romaru
    1116. Roberto Roncon-Albuquerque
    1117. Mélanie Roriz
    1118. Manuel Rosa-Calatrava
    1119. Michael Rose
    1120. Dorothea Rosenberger
    1121. Andrea Rossanese
    1122. Matteo Rossetti
    1123. Bénédicte Rossignol
    1124. Patrick Rossignol
    1125. Stella Rousset
    1126. Carine Roy
    1127. Benoît Roze
    1128. Desy Rusmawatiningtyas
    1129. Clark D Russell
    1130. Maeve Ryan
    1131. Maria Ryan
    1132. Steffi Ryckaert
    1133. Aleksander Rygh Holten
    1134. Isabela Saba
    1135. Musharaf Sadat
    1136. Valla Sahraei
    1137. Nadia Saidani
    1138. Maximilien Saint-Gilles
    1139. Pranya Sakiyalak
    1140. Leonardo Salazar
    1141. Gabriele Sales
    1142. Stéphane Sallaberry
    1143. Charlotte Salmon Gandonniere
    1144. Hélène Salvator
    1145. Olivier Sanchez
    1146. Angel Sanchez-Miralles
    1147. Vanessa Sancho-Shimizu
    1148. Gyan Sandhu
    1149. Zulfiqar Sandhu
    1150. Pierre-François Sandrine
    1151. Oana Sandulescu
    1152. Marlene Santos
    1153. Shirley Sarfo-Mensah
    1154. Bruno Sarmento Banheiro
    1155. Iam Claire E Sarmiento
    1156. Benjamine Sarton
    1157. Sree Satyapriya
    1158. Rumaisah Satyawati
    1159. Egle Saviciute
    1160. Parthena Savvidou
    1161. Justin Schaffer
    1162. Tjard Schermer
    1163. Arnaud Scherpereel
    1164. Marion Schneider
    1165. Stephan Schroll
    1166. Michael Schwameis
    1167. Janet T Scott
    1168. James Scott-Brown
    1169. Nicholas Sedillot
    1170. Tamara Seitz
    1171. Caroline Semaille
    1172. Malcolm G Semple
    1173. Eric Senneville
    1174. Claudia Sepulveda
    1175. Filipa Sequeira
    1176. Tânia Sequeira
    1177. Pablo Serrano Balazote
    1178. Ellen Shadowitz
    1179. Mohammad Shamsah
    1180. Shaikh Sharjeel
    1181. Pratima Sharma
    1182. Catherine A Shaw
    1183. Victoria Shaw
    1184. Haixia Shi
    1185. Mohiuddin Shiekh
    1186. Takuya Shiga
    1187. Nobuaki Shime
    1188. Hiroaki Shimizu
    1189. Keiki Shimizu
    1190. Naoki Shimizu
    1191. Sally Shrapnel
    1192. Hoi Ping Shum
    1193. Nassima Si Mohammed
    1194. Jeanne Sibiude
    1195. Atif Siddiqui
    1196. Louise Sigfrid
    1197. Piret Sillaots
    1198. Catarina Silva
    1199. Maria Joao Silva
    1200. Rogério Silva
    1201. Wai Ching Sin
    1202. Budha Charan Singh
    1203. Punam Singh
    1204. Pompini Agustina Sitompul
    1205. Vegard Skogen
    1206. Sue Smith
    1207. Benjamin Smood
    1208. Coilin Smyth
    1209. Michelle Smyth
    1210. Michelle Smyth
    1211. Morgane Snacken
    1212. Dominic So
    1213. Monserrat Solis
    1214. Joshua Solomon
    1215. Tom Solomon
    1216. Emily Somers
    1217. Agnès Sommet
    1218. Myung Jin Song
    1219. Rima Song
    1220. Tae Song
    1221. Jack Song Chia
    1222. Michael Sonntagbauer
    1223. Alberto Sotto
    1224. Edouard Soum
    1225. Ana Chora Sousa
    1226. Marta Sousa
    1227. Maria Sousa Uva
    1228. Vicente Souza-Dantas
    1229. Alexandra Sperry
    1230. BP Sanka Ruwan Sri Darshana
    1231. Shiranee Sriskandan
    1232. Sarah Stabler
    1233. Thomas Staudinger
    1234. Stephanie-Susanne Stecher
    1235. Ymkje Stienstra
    1236. Birgitte Stiksrud
    1237. Eva Stolz
    1238. Amy Stone
    1239. Adrian Streinu-Cercel
    1240. Anca Streinu-Cercel
    1241. Samantha Strudwick
    1242. Ami Stuart
    1243. David Stuart
    1244. Gabriel Suen
    1245. Jacky Y Suen
    1246. Prasanth Sukumar
    1247. Asfia Sultana
    1248. Charlotte Summers
    1249. Dubravka Supic
    1250. Magdalena Surovcová
    1251. Konstantinos Syrigos
    1252. Jaques Sztajnbok
    1253. Konstanty Szuldrzynski
    1254. Shirin Tabrizi
    1255. Fabio S Taccone
    1256. Lysa Tagherset
    1257. Ewa Talarek
    1258. Sara Taleb
    1259. Jelmer Talsma
    1260. Maria Lawrensia Tampubolon
    1261. Le Van Tan
    1262. Hiroyuki Tanaka
    1263. Taku Tanaka
    1264. Hayato Taniguchi
    1265. Coralie Tardivon
    1266. Pierre Tattevin
    1267. MAzhari Taufik
    1268. Hassan Tawfik
    1269. Richard S Tedder
    1270. João Teixeira
    1271. Sofia Tejada
    1272. Marie-Capucine Tellier
    1273. Vanessa Teotonio
    1274. François Téoulé
    1275. Pleun Terpstra
    1276. Olivier Terrier
    1277. Nicolas Terzi
    1278. Hubert Tessier-Grenier
    1279. Adrian Tey
    1280. Vincent Thibault
    1281. Simon-Djamel Thiberville
    1282. Benoît Thill
    1283. Shaun Thompson
    1284. David Thomson
    1285. Emma C Thomson
    1286. Duong Bich Thuy
    1287. Ryan S Thwaites
    1288. Paul Tierney
    1289. Vadim Tieroshyn
    1290. Jean-François Timsit
    1291. Noémie Tissot
    1292. Maria Toki
    1293. Timo Tolppa
    1294. Kristian Tonby
    1295. Antoni Torres
    1296. Margarida Torres
    1297. Hernando Torres-Zevallos
    1298. Michael Towers
    1299. Théo Treoux
    1300. Huynh Trung Trieu
    1301. Cécile Tromeur
    1302. Ioannis Trontzas
    1303. Tiffany Trouillon
    1304. Jeanne Truong
    1305. Christelle Tual
    1306. Sarah Tubiana
    1307. Helen Tuite
    1308. Jean-Marie Turmel
    1309. Lance CW Turtle
    1310. Pawel Twardowski
    1311. Makoto Uchiyama
    1312. PGIshara Udayanga
    1313. Roman Ullrich
    1314. Alberto Uribe
    1315. Asad Usman
    1316. Cristinava Vajdovics
    1317. Luís Val-Flores
    1318. Ana Luiza Valle
    1319. Amélie Valran
    1320. Stijn Van de Velde
    1321. Marcelvan den Berge
    1322. Machteld Van der Feltz
    1323. Nicky Van Der Vekens
    1324. Peter Van der Voort
    1325. Sylvie Van Der Werf
    1326. Lauravan Gulik
    1327. Jarne Van Hattem
    1328. Stevenvan Lelyveld
    1329. Carolienvan Netten
    1330. Gitte Van Twillert
    1331. Noémie Vanel
    1332. Henk Vanoverschelde
    1333. Pooja Varghese
    1334. Michael Varrone
    1335. Charline Vauchy
    1336. Helen Vaughan
    1337. Aurélie Veislinger
    1338. Sebastian Vencken
    1339. Sara Ventura
    1340. Annelies Verbon
    1341. José Ernesto Vidal
    1342. César Vieira
    1343. Joy Ann Villanueva
    1344. Judit Villar
    1345. Pierre-Marc Villeneuve
    1346. Andrea Villoldo
    1347. Nguyen Van Vinh Chau
    1348. Benoit Visseaux
    1349. Hannah Visser
    1350. Chiara Vitiello
    1351. Fanny Vuotto
    1352. Chih-Hsien Wang
    1353. Jia Wei
    1354. Katharina Weil
    1355. Sanne Wesselius
    1356. Murray Wham
    1357. Bryan Whelan
    1358. Nicole White
    1359. Paul Henri Wicky
    1360. Aurélie Wiedemann
    1361. Surya Otto Wijaya
    1362. Keith Wille
    1363. Virginie Williams
    1364. Evert-Jan Wils
    1365. Ng Wing Yiu
    1366. Calvin Wong
    1367. Ioannis Xynogalas
    1368. Sophie Yacoub
    1369. Masaki Yamazaki
    1370. Yazdan Yazdanpanah
    1371. Cécile Yelnik
    1372. Stephanie Yerkovich
    1373. Toshiki Yokoyama
    1374. Hodane Yonis
    1375. Obada Yousif
    1376. Saptadi Yuliarto
    1377. Akram Zaaqoq
    1378. Marion Zabbe
    1379. Kai Zacharowski
    1380. Maram Zahran
    1381. Maria Zambon
    1382. Miguel Zambrano
    1383. Alberto Zanella
    1384. Konrad Zawadka
    1385. Hiba Zayyad
    1386. Alexander Zoufaly
    1387. David Zucman
  2. Matthew D Hall

    Big Data Institute, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
    Contribution
    Conceptualization, Formal analysis, Methodology, Software, Visualization, Writing – original draft, Writing – review and editing
    For correspondence
    matthew.hall@bdi.ox.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2671-3864
  3. Joaquín Baruch

    ISARIC Global Support Centre, Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
    Contribution
    Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
  4. Gail Carson

    ISARIC Global Support Centre, Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
    Contribution
    Conceptualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
  5. Barbara Wanjiru Citarella

    ISARIC Global Support Centre, Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
    Contribution
    Data curation, Project administration, Writing – review and editing
    Competing interests
    No competing interests declared
  6. Andrew Dagens

    ISARIC Global Support Centre, Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
    Contribution
    Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
  7. Emmanuelle A Dankwa

    Department of Statistics, University of Oxford, Oxford, United Kingdom
    Contribution
    Formal analysis, Software, Writing – review and editing
    Competing interests
    No competing interests declared
  8. Christl A Donnelly

    1. Department of Statistics, University of Oxford, Oxford, United Kingdom
    2. MRC Centre for Global Infectious Disease Analysis, Abdul Latif Jameel Institute for Disease and Emergency Analytics and Department of Infectious Disease Epidemiology, Imperial College London, London, United Kingdom
    Contribution
    Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
  9. Jake Dunning

    ISARIC Global Support Centre, Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
    Contribution
    Conceptualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
  10. Martina Escher

    ISARIC Global Support Centre, Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
    Contribution
    Writing – review and editing
    Competing interests
    No competing interests declared
  11. Christiana Kartsonaki

    MRC Population Health Research Unit, Clinical Trials Service Unit and Epidemiological Studies Unit, Nuffield Department of Population Health, University of Oxford, Oxford, United Kingdom
    Contribution
    Formal analysis, Methodology, Software, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
  12. Laura Merson

    1. ISARIC Global Support Centre, Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
    2. Infectious Diseases Data Observatory, Centre for Tropical Medicine and Global Health, University of Oxford, Oxford, United Kingdom
    Contribution
    Conceptualization, Funding acquisition, Project administration, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4168-1960
  13. Mark Pritchard

    ISARIC Global Support Centre, Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
    Contribution
    Formal analysis, Software, Writing – review and editing
    Competing interests
    No competing interests declared
  14. Jia Wei

    Big Data Institute, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
    Contribution
    Writing – review and editing
    Competing interests
    No competing interests declared
  15. Peter W Horby

    ISARIC Global Support Centre, Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
    Contribution
    Conceptualization, Writing – review and editing
    Competing interests
    No competing interests declared
  16. Amanda Rojek

    1. ISARIC Global Support Centre, Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
    2. Royal Melbourne Hospital, Melbourne, Australia Centre for Integrated Critical Care, University of Melbourne, Melbourne, Australia
    Contribution
    Conceptualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
  17. Piero L Olliaro

    ISARIC Global Support Centre, Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
    Contribution
    Conceptualization, Funding acquisition, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared

Funding

Wellcome Trust (215091/Z/18/Z)

  • Joaquín Baruch
  • Jake Dunning
  • Martina Escher
  • Jia Wei
  • Peter W Horby
  • Piero Luigi Olliaro

Bill and Melinda Gates Foundation (OPP1209135)

  • Joaquín Baruch
  • Jake Dunning
  • Martina Escher
  • Jia Wei
  • Peter W Horby
  • Piero Luigi Olliaro

University of Oxford's COVID-19 Research Response Fund (0009146)

  • Laura Merson

H2020 European Research Council (101003589)

  • ISARIC Clinical Characterisation Group

European Clinical Research Alliance on Infectious Diseases (ECRAID) (965313)

  • ISARIC Clinical Characterisation Group

Health Research Board of Ireland (CTN-2014–12)

  • ISARIC Clinical Characterisation Group

National Institute for Health Research (CO-CIN-01)

  • ISARIC Clinical Characterisation Group

Medical Research Council (MC_PC_19059)

  • ISARIC Clinical Characterisation Group

Public Health England (200907)

  • ISARIC Clinical Characterisation Group

Wellcome Trust (205228/Z/16/Z)

  • ISARIC Clinical Characterisation Group

Respiratory Infections at Imperial College London with PHE (200927)

  • ISARIC Clinical Characterisation Group

NIHR Biomedical Research Centre at Imperial College London (IS-BRC1215–20013)

  • ISARIC Clinical Characterisation Group

Research Council of Norway (312780)

  • ISARIC Clinical Characterisation Group

Liverpool Experimental Cancer Medicine Centre (C18616/A25153)

  • ISARIC Clinical Characterisation Group

CIHR Coronavirus Rapid Research Funding Opportunity (OV2170359)

  • ISARIC Clinical Characterisation Group

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was supported by the UK Foreign, Commonwealth and Development Office and Wellcome [215091/Z/18/Z] and the Bill & Melinda Gates Foundation [OPP1209135]; CIHR Coronavirus Rapid Research Funding Opportunity OV2170359; Grants from Rapid European COVID-19 Emergency Response research (RECOVER) [H2020 project 101003589] and European Clinical Research Alliance on Infectious Diseases (ECRAID) [965,313]; The Imperial NIHR Biomedical Research Centre; The Cambridge NIHR Biomedical Research Centre; and Endorsed by the Irish Critical Care- Clinical Trials Group, co-ordinated in Ireland by the Irish Critical Care- Clinical Trials Network at University College Dublin and funded by the Health Research Board of Ireland [CTN-2014–12]. This work uses Data / Material provided by patients and collected by the NHS as part of their care and support #DataSavesLives. The Data / Material used for this research were obtained from ISARIC4C. The COVID-19 Clinical Information Network (CO-CIN) data was collated by ISARIC4C Investigators. Data and Material provision was supported by grants from: the National Institute for Health Research (NIHR; award CO-CIN-01), the Medical Research Council (MRC; grant MC_PC_19059), and by the NIHR Health Protection Research Unit (HPRU) in Emerging and Zoonotic Infections at University of Liverpool in partnership with Public Health England (PHE), (award 200907), Wellcome Trust [Turtle, Lance-fellowship 205228/Z/16/Z], NIHR HPRU in Respiratory Infections at Imperial College London with PHE (award 200927), Liverpool Experimental Cancer Medicine Centre (grant C18616/A25153), NIHR Biomedical Research Centre at Imperial College London (award IS-BRC-1215–20013), and NIHR Clinical Research Network providing infrastructure support. This work was possible due to the dedication and hard work of the Norwegian SARS-CoV-2 study team, and supported by grants from Research Council of Norway grant no 312780, and a philanthropic donation from Vivaldi Invest A/S owned by Jon Stephenson von Tetzchner; The dedication and hard work of the Groote Schuur Hospital Covid ICU Team, and supported by the Groote Schuur nursing and University of Cape Town registrar bodies coordinated by the Division of Critical Care at the University of Cape Town; and supported by the COVID clinical management team, AIIMS, Rishikesh, India. Matthew Hall and Christophe Fraser were supported by a Li Ka Shing Foundation award to Christophe Fraser.

Ethics

Human subjects: The study was approved by the World Health Organization Ethics Review Committee (RPC571 and RPC572). Local ethics approval was obtained for each participating country and site according to local requirements. All necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived. All necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived.

Senior Editor

  1. Miles P Davenport, University of New South Wales, Australia

Reviewing Editor

  1. Talía Malagón, McGill University, Canada

Publication history

  1. Preprint posted: June 3, 2021 (view preprint)
  2. Received: June 3, 2021
  3. Accepted: November 18, 2021
  4. Accepted Manuscript published: November 23, 2021 (version 1)
  5. Version of Record published: January 26, 2022 (version 2)
  6. Version of Record updated: March 14, 2022 (version 3)

Copyright

© 2021, ISARIC Clinical Characterisation Group et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,326
    Page views
  • 298
    Downloads
  • 4
    Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Crossref, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. ISARIC Clinical Characterisation Group
  2. Matthew D Hall
  3. Joaquín Baruch
  4. Gail Carson
  5. Barbara Wanjiru Citarella
  6. Andrew Dagens
  7. Emmanuelle A Dankwa
  8. Christl A Donnelly
  9. Jake Dunning
  10. Martina Escher
  11. Christiana Kartsonaki
  12. Laura Merson
  13. Mark Pritchard
  14. Jia Wei
  15. Peter W Horby
  16. Amanda Rojek
  17. Piero L Olliaro
(2021)
Ten months of temporal variation in the clinical journey of hospitalised patients with COVID-19: An observational cohort
eLife 10:e70970.
https://doi.org/10.7554/eLife.70970
  1. Further reading

Further reading

    1. Medicine
    2. Microbiology and Infectious Disease
    3. Epidemiology and Global Health
    4. Immunology and Inflammation
    Edited by Jos WM van der Meer et al.
    Collection

    eLife has published articles on a wide range of infectious diseases, including COVID-19, influenza, tuberculosis, HIV/AIDS, malaria and typhoid fever.

    1. Epidemiology and Global Health
    2. Medicine
    Qing Shen, Huan Song ... Unnur Valdimarsdóttir
    Research Article Updated

    Background:

    The association between cardiovascular disease (CVD) and selected psychiatric disorders has frequently been suggested while the potential role of familial factors and comorbidities in such association has rarely been investigated.

    Methods:

    We identified 869,056 patients newly diagnosed with CVD from 1987 to 2016 in Sweden with no history of psychiatric disorders, and 910,178 full siblings of these patients as well as 10 individually age- and sex-matched unrelated population controls (N = 8,690,560). Adjusting for multiple comorbid conditions, we used flexible parametric models and Cox models to estimate the association of CVD with risk of all subsequent psychiatric disorders, comparing rates of first incident psychiatric disorder among CVD patients with rates among unaffected full siblings and population controls.

    Results:

    The median age at diagnosis was 60 years for patients with CVD and 59.2% were male. During up to 30 years of follow-up, the crude incidence rates of psychiatric disorder were 7.1, 4.6, and 4.0 per 1000 person-years for patients with CVD, their siblings and population controls. In the sibling comparison, we observed an increased risk of psychiatric disorder during the first year after CVD diagnosis (hazard ratio [HR], 2.74; 95% confidence interval [CI], 2.62–2.87) and thereafter (1.45; 95% CI, 1.42–1.48). Increased risks were observed for all types of psychiatric disorders and among all diagnoses of CVD. We observed similar associations in the population comparison. CVD patients who developed a comorbid psychiatric disorder during the first year after diagnosis were at elevated risk of subsequent CVD death compared to patients without such comorbidity (HR, 1.55; 95% CI, 1.44–1.67).

    Conclusions:

    Patients diagnosed with CVD are at an elevated risk for subsequent psychiatric disorders independent of shared familial factors and comorbid conditions. Comorbid psychiatric disorders in patients with CVD are associated with higher risk of cardiovascular mortality suggesting that surveillance and treatment of psychiatric comorbidities should be considered as an integral part of clinical management of newly diagnosed CVD patients.

    Funding:

    This work was supported by the EU Horizon 2020 Research and Innovation Action Grant (CoMorMent, grant no. 847776 to UV, PFS, and FF), Grant of Excellence, Icelandic Research Fund (grant no. 163362-051 to UV), ERC Consolidator Grant (StressGene, grant no. 726413 to UV), Swedish Research Council (grant no. D0886501 to PFS), and US NIMH R01 MH123724 (to PFS).