Epidemiological description of COVID-19 related mortality
Based on public released national data, a total of 3769 SARS-COV-2 related deaths were identified among 81470 confirmed cases up until 29 May 2020, an overall CFR of 4.6% (95% CI 4.5%-4.8%). Among the 3769 fatal cases, 2428 (64.4%) were male and 3089 (82.0%) were 60 years of age or older. The median age of all fatal cases was 70 (IQR 63-78) years.
Fatal cases were reported from 26 provinces, with most (CFR) from Hubei Province (97.1%, 3651/3769), especially from Wuhan (79.9%, 3011/3769), followed by Xiaogan city (3.4%, 129/3769), Huanggang city (3.3%,125/3769), and Ezhou city (1.6%, 59/3769) in Hubei Province (Figure S1; Table S1). Provinces other than Hubei collectively reported 2.9% (118/3769) of the fatal cases, with the highest CFR (4.0%, 95%CI 0.8%-11.1%) recorded in Xinjiang Uygur Autonomous Region, followed by Hainan Province (3.6%, 95% CI 1.3%-7.6%) and Heilongjiang Province (2.7%, 95% CI 1.4%-4.5%), with the lowest CFR (0.1%, 95% CI 0-0.4%) in Zhejiang Province. No deaths were reported in Jiangsu, Shanxi, Ningxia, Qinghai and Tibet Province. The CFR outside Hubei Province (0.9%, 95% CI 0.7%-1.0%, 118/13669) was significantly lower than that of Hubei Province (5.4%, 95% CI 5.2%-5.6%, 3651/67801, P<0.001, Figure S1; Table S1). The mortality rate in Wuhan city was 28 (95% CI 27-29) per 100,000, higher than other cities in Hubei Province (P<0.001). The CFR of Qingshan District in Wuhan city was the highest (8.3%, 95% CI 7.3%-9.4%, 225/2964), while the mortality rate was the highest [94 (95%CI 79-112) per 100,000] in Hannan District.
Detailed information was obtained from all the 118 fatal patients outside Hubei Province, which represents all of the fatal cases reported by the government. Among these 63.0% were male and the median age was 73 (65-80) years, older than that of all fatal cases nationwide. For comparison, we used 443 fatal cases selected from six designated hospitals in Wuhan, with a comparable age (73 [IQR 64-81] years) and sex (61.6% male) distribution as those of fatal patients outside Hubei Province (Table 1).
For both groups, the CFR increased dramatically with age with a highly similar trend, i.e. starting from no deaths under 20 years of age, to a very low CFR <40 years of age, steadily increasing to >10% in 50-60 years old, and then higher in those ≥60 years of age (Table 1). Longer interval from disease onset to diagnosis and higher proportion of the interval >10 days from disease onset to hospital admission were observed for fatal patients from Wuhan than those from outside Hubei (10 vs. 6 days, P<0.001 and 33.4% vs. 17.8%, P=0.003, respectively). It is noteworthy that a significantly higher proportion of fatal cases entered hospital with mild disease in Wuhan than outside Hubei Province (48.8% vs. 5.1%, P<0.001), which was also related to a shorter clinical course (14 [IQR 9-22] vs. 17 [IQR 10-25] days, P=0.028).
Age- and sex-specific CFR that differed between Wuhan and outside Hubei
We further made a precise estimation of the age/sex CFR that was calculated in Wuhan, and outside Hubei Province in China. The CFR following SARS-COV-2 infection appeared to be higher in males, and increased with age; the highest CFR observed at the male patients ≥70 years old in both regions (Figure 1, Table S2). The outcome of COVID-19 patients was generally worse for males, but the magnitude of this difference was higher in Wuhan than in other provinces outside Hubei, with RR of death for males calculated as 1.94 (95% CI 1.59-2.36) vs. 1.49 (95% CI 1.03-2.17), respectively, compared to females (Table S3). CFR risk ratios also varied across age groups, which was seen in both regions. For example, in Wuhan city, the overall male to female CFR risk ratio in all age groups was 1.94 (95% CI 1.59-2.36), with the greatest RR 2.94 (95% CI 1.60-5.38) observed for the 50-60 years of age group and the lowest RR 1.48 (95% 1.12-1.95) for the ≥70 years of age group. In a similar way, for outside Hubei Provinces, the overall male to female CFR risk ratio in all age groups was 1.49 (95% CI 1.03-2.17), with the greatest RR 3.06 (95% CI 0.83-11.34) observed in the 50-60 years of age group, and the lowest RR 1.35 (95% CI 0.83-2.21) for the ≥70 years of age group (Table S3).
The CFR following SARS-COV-2 infection appeared to increase with age, but with a magnitude greater in other provinces outside Hubei than in Wuhan. In Wuhan city, the age specific RR increased from 4.72 (95% CI 2.74-8.13) in the 50-60 years age group to 10.59 (95% CI 6.42-17.47) in the 60-70 years age group, to the highest RR of 40.00 (95% CI 24.68-64.84) in the ≥70 years age group. In contrast, for outside Hubei Province, a more significant effect of older age was observed: the RR increased from 7.88 (95% CI 2.77-22.40) in the 50-60 years age group, to 31.33 (95% CI 12.14-80.85) in the 60-70 years age group and to 147.37 (95% CI 59.34-366.00) in the ≥70 years age group (Table S3). The mortality rate in other provinces outside Hubei had a similar trend as that of CFR, which was low in both sex under 40 years old, and then increased along with the increase of age, and to higher extents in female (Figure 1C).
Risk factor analysis for fatal outcome
In univariate and multivariate logistic regression model, the odds of a fatal outcome of COVID-19 patients was significantly associated with age, sex and the interval from symptom onset to admission (Table S4 and Figure 2A). When data on fatal patients from Wuhan and other provinces outside Hubei Province were separately analyzed, the interval from symptom onset to admission were found in addition to be associated with a fatal outcome, but only in other provinces outside of Hubei and not in Wuhan. A long interval from symptom onset to admission was associated with higher CFR, with the OR estimated to be 4.68 (95% CI 2.49-8.82) for >10 days delay and 5.04 (95% CI 3.06-8.31) for 6-10 days compared to 1-5 days (Figure 2A and Table S4). In all three multivariate models, age remained the strongest risk factor. Patients ≥70 years old outside Hubei had the highest risk of death, with >167-fold increase compared to patients <50 years old (OR=167.05, 95% CI 65.78-424.18).
Survival probability over time for fatal cases and related factors for clinical course
The cumulative incidence probability of survival over time for fatal cases with SARS-COV-2 infection is presented by age, sex, and the interval from symptom onset to admission (Figure 2B; Table S5). The median clinical course of the fatal cases (duration from onset of symptoms to death) was 15 (IQR 9-22) days. About 32.1% (130/561) and 38.3% (215/561) patients died within 1-10 days and 11-20 days after symptom onset, respectively. The median clinical course of those cases which survived (duration from symptom onset to discharge) was 23 (IQR 18-30) days. Significant differences were observed in fatal outcome for cases with different age groups and sex (both P<0.001, Figure 3).
By fitting a multivariate ordinal logistic regression model to fatal case data, severe patients or the patients in Wuhan city when diagnosed were associated with a shorter clinical course, indicating a very rapid progress to a fatal outcome. A longer interval from symptom onset to hospital admission were all associated with a longer clinical course (Table S6).
Characteristics of fatal patients with mild pneumonia
There was 39.6% (222/561) of fatal cases who were admitted to hospital with mild pneumonia; the remainder 60.4% were admitted with severe pneumonia (Table S7). The median age of fatal cases admitted as mild was 75 (IQR 65-82) years, comparable with 72 (IQR 64-80) years for severe cases. The sex distribution was also comparable. Both the interval from onset of symptoms to diagnosis and the interval from onset of symptoms to admission were comparable between fatal cases with mild pneumonia and those with severe pneumonia.
The temporal pattern of CFR
The daily cumulative CFR was estimated by using the reported number of total fatal cases divided by the daily number of total confirmed cases reported (Figure 4). For Wuhan fatal cases, the CFR in January was significantly higher than during the later phase of the epidemic (Figure 4A). There was a clear peak in CFR during the week of 23-29 January 2020, followed by an obvious reduction thereafter to a stable level until recently. The national temporal trend was dominated by that of Wuhan city where both the total number of cases and fatal cases outweighs the other provinces (Figure 4A). In contrast, for provinces other than Hubei, there was an obvious low CFR observed during the first week of February, a period which corresponded to one week after Chinese spring festival travel rush; a tender increase was followed, bringing the CFR to the peaking level at the end of February (Week 6 of the epidemic); a decreasing trend was then observed after February, where the CFR was maintained at a stable level till the latest observation (Figure 4A).
The longitudinal CFR profiles were compared among different age groups and between Wuhan and outside Hubei Province. Generally, the CFR curve of three groups showed similar dynamic trends, but with much greater magnitude in the ≥70 years group than the other two groups. For example, both the decrease of CFR at the end of February in Wuhan and the increase of CFR at the early February in other provinces outside Hubei were obvious and notable in the ≥70 years group than the other two age groups, remarkably contributing to the overall trend of CFR in both regions (Figure 4B, C).